Datasets:

timaeus

/

pile-pubmed_central

like 0

Follow

Timaeus 12

1. Introduction =============== Acetaminophen (APAP) is one of the most common analgesic and antipyretic drug and it has been used as a powerful tool to study mechanisms of hepatotoxicity. Although it is safe at therapeutic doses, overdoses of APAP can cause severe liver injury or acute liver failure and during the last decade, APAP indeed became the most frequent cause of acute liver failure in many countries \[[@B1-molecules-18-03467]\]. APAP-induced reactive metabolites deplete glutathione (GSH) and subsequently cause protein binding, as a critical event in the toxicity \[[@B2-molecules-18-03467],[@B3-molecules-18-03467],[@B4-molecules-18-03467]\]. Based on this mechanism, *N*-acetylcysteine (NAC) has been used to treat patients with APAP-induced toxicity \[[@B5-molecules-18-03467]\], and NAC is the most popular therapeutic agent for this application \[[@B6-molecules-18-03467]\]. The effective therapy against APAP overdose-induced hepatotoxicity is GSH replacement in order to scavenge reactive metabolites such as *N*-acetyl-*p*-benzoquinone imine (NAPQI), which is accomplished with NAC and other sulfhydryl donors. However, NAC needs to be given within 12 to 24 h of APAP ingestion. Patients presenting later may benefit from increased metabolic flux, but the likelihood of a positive outcome is notably decreased \[[@B7-molecules-18-03467],[@B8-molecules-18-03467]\]. [l]{.smallcaps}-2-oxothiazolidine-4-carboxylate (C~4~H~5~NO~3~S, FW 147.16; OTC) is a prodrug of cysteine. It is a substrate of 5-oxoprolinase, an ubiquitous intracellular enzyme, which generates cysteine from OTC intracellularly \[[@B9-molecules-18-03467],[@B10-molecules-18-03467]\]. OTC exhibits ameliorative properties suggesting a number of potential clinical applications in *in vitro* and *in vivo* experiments \[[@B11-molecules-18-03467]\]. In other studies, NAC and OTC replenished cellular glutathione stores in situations in which glutathione is acutely depleted due to the detoxification of a toxic drug metabolite in animal models \[[@B12-molecules-18-03467],[@B13-molecules-18-03467],[@B14-molecules-18-03467]\]. OTC can also increase intracellular concentrations of GSH above physiological concentrations \[[@B10-molecules-18-03467],[@B15-molecules-18-03467],[@B16-molecules-18-03467],[@B17-molecules-18-03467]\]. One of the major functions of GSH is the detoxification of reactive species and toxic oxygen metabolites generated by the endogenous and exogenous metabolism pathways \[[@B18-molecules-18-03467]\]. Glutathione peroxidase mediates scavenging of intermediates such as hydrogen peroxide and hydroperoxides at the expense of GSH \[[@B11-molecules-18-03467]\]. Therefore modulation of GSH and GSH-px is increasingly important in oxidative stress related diseases and their care. To the best of our knowledge, the action of OTC against APAP-induced liver injury in mice has not been demonstrated systematically. Hence, the present study focused on evaluating the ameliorative effects of OTC against APAP-induced liver damage and its mechanism(s) of action in a murine model. 2. Results and Discussion ========================= 2.1. OTC Treatment was Effective in Preventing APAP-induced Liver Damage ------------------------------------------------------------------------ Male BALB/c mice treated with a 300 mg/kg of APAP showed evidence of significant liver damage at 12 h, as indicated by the notable increase of serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels ([Figure 1](#molecules-18-03467-f001){ref-type="fig"}). ALT and AST were significantly increased in the APAP administrated group compared to the normal control group (*p* \< 0.001). However, OTC administration (50 mg/kg and 100 mg/kg) significantly reduced serum ALT and AST levels ([Figure 1](#molecules-18-03467-f001){ref-type="fig"}). NAC (100 mg/kg), positive control to APAP-induced liver damage, also show a similar reductive effect towards APAP-induced elevation of serum levels of ALT and AST. {#molecules-18-03467-f001} 2.2. OTC Treatment had Effects on GSH and GSH-Peroxidase Recovery ----------------------------------------------------------------- OTC and NAC treatments resulted in a complete recovery of GSH and GSH-px levels in liver tissues ([Figure 2](#molecules-18-03467-f002){ref-type="fig"}). APAP treatment resulted in depletion of liver GSH levels at 12 h, however the liver GSH content was further increased in a dose dependent manner by OTC compared to APAP alone ([Figure 2](#molecules-18-03467-f002){ref-type="fig"}A). Administration of 50 mg/kg and 100 mg/kg OTC significantly elevated the liver GSH levels compared to the APAP only group (*p* \< 0.05), however 25 mg/kg OTC produced no relevant recovery of the liver GSH levels. NAC also recovered the liver GSH levels with an effect similar to that of OTC in our APAP-induced liver damage model. OTC administration recovered the GSH-px levels depleted with APAP treatment in a dose dependent manner ([Figure 2](#molecules-18-03467-f002){ref-type="fig"}B). Decrease of the liver GSH-px level were significantly recovered by 50 mg/kg and 100 mg/kg OTC treatment (*p* \< 0.05), however 25 mg/kg OTC showed no relevant recovery of the liver GSH and GSH-px level. NAC also recovered the liver GSH and GSH-px level similar to effect of OTC in the APAP-induced liver damage model. {#molecules-18-03467-f002} 2.3. MDA and 4-HNE Level ------------------------ Twelve hours after APAP administration, MDA and 4-HNE levels were increased 2.1- and 3-fold, respectively, in the APAP group compared with the normal group ([Figure 3](#molecules-18-03467-f003){ref-type="fig"}). OTC treatment significantly alleviated APAP-induced MDA ([Figure 3](#molecules-18-03467-f003){ref-type="fig"}A) and 4-HNE ([Figure 3](#molecules-18-03467-f003){ref-type="fig"}B) production in a dose dependent manner. {#molecules-18-03467-f003} In mice receiving 50 mg/kg and 100 mg/kg of OTC in the APAP-induced hepatotoxicity model, MDA levels were significantly reduced compared to the APAP alone group, and levels of 4-HNE were also significantly reduced compared to the APAP alone group. 2.4. Histologic Observation and Caspase-3 Activity -------------------------------------------------- Histopathological analysis of the APAP alone treated mice showed severe centrilobular necrosis, lymphocyte infiltration and nitrotyrosine development. The vacuolization, cell swelling and abnormal architecture around the centrilobular veins documents that the cells were undergoing necrosis ([Figure 4](#molecules-18-03467-f004){ref-type="fig"}A) and nitrotyrosine formation was significantly increased in the group administrated APAP ([Figure 4](#molecules-18-03467-f004){ref-type="fig"}B). However, OTC treatment significantly alleviated the APAP-induced liver injuries in a dose dependent manner. NAC also recovered the histological parameters similar to the effect of OTC in the APAP-induced liver damage model. {#molecules-18-03467-f004} The extent of necrosis was correlated with massive DNA fragmentation as demonstrated by the TUNEL assay ([Figure 5](#molecules-18-03467-f005){ref-type="fig"}A). Treatment with OTC markedly reduced the area of apoptosis, the number of TUNEL-positive cells, and nitrotyrosine protein adducts. Treatment with NAC also attenuated the areas of apoptosis, the extent of DNA damage and nitrotyrosine staining. Therefore, these results suggest that treatment with OTC after APAP administration was effective in protecting the liver compared to NAC. To determine whether or not this anti-apoptosis effect of OTC could be involved in the APAP-induced liver damage, caspase-3 activity was measured in the liver tissue lysates ([Figure 5](#molecules-18-03467-f005){ref-type="fig"}B). Caspase-3 activity was clearly elevated in the APAP administrated group compared with the no treatment normal group. Administration of OTC significantly reduced caspase-3 activation compared to the APAP only group in a dose dependent manner, however 25 mg/kg OTC showed no relevant recovery of the liver GSH levels. NAC also recovered histological parameters with an effect similar to that of OTC in the APAP-induced liver damage model. These results suggested that APAP-induced liver damage was related to the mechanism of cell death through apoptosis. {#molecules-18-03467-f005} 2.5. Discussion --------------- The goals of the present study were to examine the effectiveness of the [l]{.smallcaps}-cysteine pro-drug OTC in its ability to prevent APAP-induced liver damage. In this study, although 25 mg/kg OTC showed no relevant hepatoprotective effects in the APAP-induced hepatotoxicity model mice, above 50 mg/kg OTC had protective effects. OTC administration reduced APAP-induced serum AST and ALT levels and showed recovery of GSH, GSH-px levels in APAP-damaged liver tissue. Administration of OTC decreased centrilobular necrosis, MDA and 4-HNE formations, TUNEL positive area and caspase-3 activation in liver tissues of APAP-induced hepatotoxic model mice. OTC was synthesized by reaction of [l]{.smallcaps}-cysteine with phenyl chloroformate \[[@B17-molecules-18-03467]\]. OTC is useful in intracellular enzyme studies because it is an intracellular delivery system for [l]{.smallcaps}-cysteine \[[@B9-molecules-18-03467],[@B10-molecules-18-03467],[@B18-molecules-18-03467]\]. OTC is very soluble and stable in solution, and converted intracellulary to [l]{.smallcaps}-cysteine by 5-oxo-[l]{.smallcaps}-prolinase via the unstable intermediate, *S*-carboxycysteine \[[@B9-molecules-18-03467],[@B19-molecules-18-03467]\]. Effects of OTC by replenishment of tissue GSH levels were also reported in an endotoxin-induced acutely septic mouse model \[[@B20-molecules-18-03467]\], cardiac dysfunction \[[@B21-molecules-18-03467]\], hepatotoxicity model \[[@B22-molecules-18-03467]\] and asthma \[[@B23-molecules-18-03467]\]. OTC administration ameliorated APAP-induced leakage of cellular enzymes into plasma which is a sign of hepatic tissue damage. Measurement of serum ALT and AST levels are used as important diagnostic markers to indicate liver injury due to hepatotoxins \[[@B24-molecules-18-03467],[@B25-molecules-18-03467]\]. Pretreatment with OTC decreased serum ALT and AST levels in APAP-induced hepatotoxicity model mice ([Figure 1](#molecules-18-03467-f001){ref-type="fig"}) and liver damages in the form of changed liver architectures were recovered in a dose dependent manner ([Figure 4](#molecules-18-03467-f004){ref-type="fig"}). MDA is a product of polyunsaturated fatty acid (PUFA) peroxidation and is a good marker of lipid peroxidation \[[@B26-molecules-18-03467]\], which is related to APAP-induced tissue damage. MDA levels were significantly increased in plasma \[[@B27-molecules-18-03467],[@B28-molecules-18-03467]\], hepatocytes \[[@B29-molecules-18-03467]\] and liver tissues, and reduced by natural compound extracts. Therefore, administration of OTC broke the chain reaction of lipid peroxidation evidenced, for example, by MDA formation in the APAP-induced hepatotoxicity model. Thus, we suggest that the therapeutic potential of OTC is dependent on an antioxidant mechanism. 4-HNE-protein adducts may be considered a particularly good marker of lipid oxidation during liver injury. Indeed, the demonstrated adduct formation reaction of 4-HNE with important signaling proteins strongly suggests a pathogenic role for lipid aldehyde in the progression of liver diseases \[[@B30-molecules-18-03467],[@B31-molecules-18-03467]\]. In this study, levels of 4-HNE in liver tissue were higher in mice that were given APAP alone, whereas OTC administration reduced 4-HNE adduction in APAP-induced hepatotoxicity model mice. This result suggested that the OTC may be partly involved in 4-HNE metabolism to counteract the APAP-induced hepatotoxicity. Nitration of tyrosine (*i.e*., formation of nitrotyrosine) has been shown to be an excellent biomarker of peroxynitrite formation \[[@B32-molecules-18-03467],[@B33-molecules-18-03467]\] and it was shown that nitrotyrosine occurs in the centrilobular cells of the liver of acetaminophen-treated mice. Peroxynitrite formation is formed by a rapid reaction between nitric oxide and superoxide, and peroxynitrile production was increased under APAP-toxicity conditions \[[@B34-molecules-18-03467]\]. It is normally detoxified by GSH and GSH px, and GSH is depleted in acetaminophen toxicity \[[@B35-molecules-18-03467]\]. GSH px is a key enzyme in this defense mechanism \[[@B36-molecules-18-03467]\]. In some reports, neither OTC nor *N*-acetylcysteine resulted in statistically significant increases in plasma GSH in normal healthy volunteers at 8 h \[[@B14-molecules-18-03467]\]. In this study, OTC administration significantly reduced nitrotyrosine levels in a dose dependent manner in liver tissues of APAP-induced hepatotoxicity model mice. This result suggest that GSH and GSH px activation is induced by OTC administration and this is consistent with other parameters illustrating the hepatoprotective nature of OTC. The APAP-induced hepatotoxicity serves as an example of the interrelationship between apoptotic and necrotic cell death and their common origin in mitochondrial dysfunction, and overdoses of APAP are a frequent cause of acute drug-induced liver failure \[[@B37-molecules-18-03467]\]. As in other mechanisms of liver failure, the roles played by apoptosis and oncotic necrosis in APAP-induced liver failure have been controversial. Caspase-3 protein is a member of the cysteine-aspartic acid protease (caspase) family \[[@B38-molecules-18-03467]\] and caspase-3 shares many of the typical characteristics common to all currently-known caspases \[[@B39-molecules-18-03467]\]. Acetaminophen triggered the release of cytochrome c from mitochondria into the cytosol, activation of caspase-3, 8, and 9, cleavage of poly(ADP-ribose) polymerase, and degradation of lamin B1 and DNA \[[@B40-molecules-18-03467]\]. Several reports have shown that APAP-induced liver apoptosis was inhibited by repression of mitochondrial apoptotic signaling \[[@B41-molecules-18-03467]\], preventing down-regulation of Bcl-2 and up-regulation of Bax \[[@B42-molecules-18-03467]\], regulation of connexin \[[@B43-molecules-18-03467]\], and cleaved caspase-3 expression \[[@B43-molecules-18-03467]\] caused by maintaining of hepatic glutathione homeostasis \[[@B44-molecules-18-03467]\]. In present study, we observed APAP-induced apoptotic signaling with TUNEL staining of liver tissue and cleavage of caspase-3, and this was significantly reduced by OTC administration. Therefore, these results suggested that OTC is an effective substance against liver injuries. 3. Experimental =============== 3.1. Materials -------------- APAP, OTC and other chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA). ALT and AST levels in serum were measured using ALT and AST assay kits (ASAN Pharmaceutical. Seoul, Korea). GSH and GSH-px assay kits were obtained from United States Biological (Salem, MA, USA). Enzyme linked immunosorbent assay (ELISA) kits for MDA and 4-HNE were purchased from Cell Biolabs, Inc. (San Diego, CA, USA). Caspase-3 activity and TUNEL staining kits were purchased from R&D Systems (Minneapolis, MN, USA). Specific antibody against nitrotyrosin for immunohistochemistry was obtained from Specific antibody against nitrotyrosin for immunohistochemistry was obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). All assay kits were used according to the corresponding manufacturer's instructions. 3.2. Animals ------------ Male BALB/c mice were purchased from The Orient Bio (SungNam, Korea) and housed in an environmentally controlled room with a 12-hour light and dark cycle and *ad libitum* access to food and water. Mice were starved overnight and orally administrated OTC or NAC. After 2 h, 300 mg/kg APAP in warm saline was administered in a single i.p. injection and then animals were sacrificed by cervical dislocation 12 h later. Blood was collected from the inferior vena cava, placed in clean Eppendorf tubes for 3 h at room temperature, and then the clotted blood was centrifuged at 14,000--20,000 × g for 15--20 min at 4 °C. Livers were excised, and small sections were fixed in 10% phosphate-buffered formalin for histological analysis. 3.3. Caspase-3 Activity Assay ----------------------------- The caspase-3 activity assay was performed using a colorimetric activity assay kit according to the manufacturer's instructions. In brief, assays were performed by incubating protein from tissue lysate (200 μg) in reaction buffer (100 μL) containing caspase-3 substrate (4 mM DEVD-pNA. 5 μL) in 96-well plates. The reaction buffer contained 1% NP-40, 20 mM Tris-HCl (pH 7.5), 137 mM *N*-acetyl-cysteine, and 10% glycerol. Lysates were incubated at 37 °C for 2 h. Samples were incubated in the dark and measured with a microplate reader (Molecular Devices, Sunnyvale, CA, USA) at an absorbance of 405 nm. 3.4. Histology and Immunohistochemistry --------------------------------------- Mice were anesthetized with diethyl ether and the liver was incised. Livers were then removed from the mice and fixed overnight in a cold 10% formalin solution. Fixed tissues were processed routinely for paraffin embedding, and 5 μm sections were used for hematoxylin-eosin staining. Stained morphology was analyzed using a microscope (Leica, Wetzlar, Germany). For immunohistochemistry, tissue sections are deparaffinized and stained with anti-nitrotyrosine antibody and TUNEL reagent. The sections were reacted overnight at 4 °C or for the time/temperature indicated in the manufacturer's instruction, rinsed with PBS and then covered with cover-slips. They were examined by fluorescence microscopy (Olympus, Tokyo, Japan) to assess the molecular expressions or distributions. 3.5. Statistical Analysis ------------------------- Differences in data among the groups were analyzed by one-way ANOVA, and all values were expressed as mean ± S.E.M. The differences between groups were considered to be significant at *p* \< 0.05. 4. Conclusions ============== In summary, this study demonstrates the ameliorative effects of OTC administration in APAP-induced hepatotoxicity model mice. Serum AST/ALT levels were reduced by OTC administration and recovery of GSH, GSH-px in liver tissue was shown. Administration of OTC decreased centrilobular necrosis, MDA/4-HNE formations, TUNEL positive area and caspase-3 activation in liver tissues of APAP-induced hepatotoxic model mice. These results suggest that OTC has a potential use for treatment against APAP-induced hepatotoxiccity related diseases. This work was supported by the Small and Medium Business Administration in 2012 (S2004797). *Sample Availability*: Samples of the compounds are available from the authors. [^1]: These authors contributed equally to this work.

1.. Introduction ================ The demand for safety-critical applications (e.g., civil aviation, aircraft landing) using GNSSs has gained extensive and increased interest in recent years. With the advent of the new GNSSs, such as the European Union\'s Galileo system \[[@b1-sensors-15-09404]\], the USA\'s modernized GPS \[[@b2-sensors-15-09404]\] and China\'s Beidou/Compass \[[@b3-sensors-15-09404]\], new radio navigation signals will be broadcast, and more attention has been devoted to the design of the signal structure, which is expected to increase the accuracy, availability, integrity and continuity of service, especially in the field of the safety of life (SOL) applications (e.g., the accuracy needed during the landing of an aircraft). High precision positioning and reliable SOL services represent the main challenges for the upcoming satellite navigation systems. The presence of disturbing signals, such as spurious, harmonic and electromagnetic interferences, will result in serious performance degradation for GNSS receivers. Radio frequency interference (RFI) is one of the biggest threats for satellite navigation systems. Although the satellite navigation system has a certain capability to be immune from interference, since the direct sequence spread spectrum (DSSS) is utilized, due to the low power of the received GNSS signals, the presence of intentional or unintentional disturbing signals, such as spurious, harmonic and electromagnetic interferences, will result in serious performance degradation for GNSS receivers. Among all of the different error sources that can potentially corrupt GNSS signals, RFI is particularly harmful, since, in some cases, it cannot be mitigated by a simple correlation process \[[@b4-sensors-15-09404]\]. The jamming environment is threatening for satellite navigation systems. Many systems rely on the transmission of radio frequency (RF) energy in the L-band. For example, the European Galileo E5a and E5b radio bands, located within 1164--1214 MHz, occupy frequencies already allocated for aeronautical radio navigation services (ARNS), such as tactical air navigation (TACAN), distance measuring equipment (DME) and secondary surveillance radar (SSR) \[[@b5-sensors-15-09404]\]. In addition, other RF transmissions, such as continuous wave (CW) signals originating from the European Digital Video Broadcast Terrestrial (DVB-T) service, can be considered as the main threat for the GNSS signals, since they appear in Radio Navigation Satellite System (RNSS) frequency bands. The presence of RFI and other channel impairments can heavily degrade the reception of useful GNSS signals, which results in poor navigation accuracy or complete loss of tracking for GNSS receivers. Currently, interference detection and mitigation (ID&M) have become very important issues for GNSS applications. Interference can be detected and mitigated through various means. On the hardware side, specialized instrumentation, such as choke rings or active beam-forming antennas, can be used to suppress interference and improve the reception of line of sight (LOS) satellite signals. The main drawbacks of these antenna techniques are their requirements for particular hardware configurations and their computational complexity \[[@b6-sensors-15-09404]--[@b8-sensors-15-09404]\]. In the GNSS receiver, a special front-end architecture design adopts pulse blanker or automatic gain control (AGC) to reduce interference. For example, a specific AGC and analog-to-digital converter (ADC) design performing digital pulse blanking has been implemented in the GPS L5 receiver \[[@b9-sensors-15-09404]\]. In addition, a nonuniform ADC controlled by a digital AGC can be adaptively adjusted based on the interference power strength to obtain an optimized conversion gain of the ADC (*i.e.*, the SNR loss in ADC) and improved receiver performance \[[@b10-sensors-15-09404],[@b11-sensors-15-09404]\]. However, these approaches require digital access to the feedback control of the AGC device, which is not a common output of generic analog front-ends used in the GNSS receivers. In the literature, several interference mitigation techniques have been proposed and investigated, and each of them differs according to the operate domain (time, frequency or space). These techniques can be classified according to the specific processing domain. Considering time domain techniques, temporal filtering can be usually adopted. This method can be implemented at the digital intermediate frequency (IF) level after the ADC in the front-end of the GNSS receiver, which is effective only against narrow band RFI sources, because a wide band interference cannot be easily discriminated from the thermal noise by adopting temporal filtering. Frequency domain techniques are generally based on spectral estimation of the incoming signal, which is obtained by applying signal processing techniques, such as the discrete Fourier transform (DFT). These frequency domain techniques are typically performed by comparing the spectrum of the received signal with a theoretical threshold, which is usually determined according to a statistical model representing the received signal \[[@b12-sensors-15-09404]\]. Recently, the research topic on transformed domain techniques (e.g., TF transform) has obtained increasing attention in the ID&M for GNSS receivers \[[@b13-sensors-15-09404]--[@b16-sensors-15-09404]\]. These techniques allow one to observe the received GNSS signals in a joint domain. In many cases, interferences may appear for a limited time and present a very variable behavior in frequency. In comparison to the GNSS signal, interferences are extremely different in terms of time and frequency characteristics. The presence of an interfering signal is limited to a region of the two-dimensional TF plane, and the adoption of the TF analysis is allowed to detect different types of disturbing signals. TF representations (TFRs) map a one-dimensional time signal into a two-dimensional function of time and frequency \[[@b17-sensors-15-09404]\], which have found significant applications in non-stationary signal analysis. An interference mitigation technique based on the TFR approach has been described \[[@b18-sensors-15-09404]\], where the TFR of the received GNSS signal is obtained by performing an orthogonal-like Gabor expansion on the samples at the output of the ADC of the GNSS receiver front-end. Another class of ID&M algorithm is reported aiming at obtaining a representation of the received signal in a different domain by making use of the time-scale transformation, which can be performed by means of wavelet transform \[[@b19-sensors-15-09404]\]. Interfering signals are usually concentrated in a limited area of the TF plane, while noise is spread over the entire plane. The TF analysis approaches are very appealing as countermeasures in the detection of a large variety of interfering signals for GNSS receivers. There are different tools representing TF distributions, and the commonly-used TFRs in interference detection for GNSS applications include spectrogram, Wigner--Ville distribution (WVD) \[[@b13-sensors-15-09404],[@b16-sensors-15-09404]\] and Choi--Williams transform (CWT) \[[@b16-sensors-15-09404]\]. The spectrogram and WVD have been considered in the interference detection for GNSS applications \[[@b13-sensors-15-09404]\]. The spectrogram approach presents the TF resolution trade-off problems according to the uncertainty principle, providing poor TF localization properties. In order to overcome the TF resolution trade-off problems of the spectrogram, WVD has been used in interference detection for GNSS receivers \[[@b13-sensors-15-09404],[@b16-sensors-15-09404]\]. WVD is well known, since it provides nearly the best TF resolution among all of the TF distributions and also satisfies a large number of good properties \[[@b20-sensors-15-09404]\], but it presents very severe cross-interfering terms without any physical meaning between true signal components (auto-terms), due to the interaction of different frequency components \[[@b16-sensors-15-09404]\]. Most of the previous works on interference detection for GNSS receivers based on TF analysis are concerned with various TFR plots. Nevertheless, the quadratic TF distribution is usually a biased estimator for signal instantaneous frequency due to the presence of cross-term problems or makes a trade-off of temporal and frequency resolution due to the limitation of the uncertainty principle. In order to overcome or attenuate the cross-interfering terms present in the the quadratic TF distributions, the Choi--Williams distribution has been proposed to detect the sweep interference for GNSS receivers \[[@b16-sensors-15-09404]\]. Appropriate image processing techniques can be also used for detection and parameter estimation of chirp signals by line detection in an image \[[@b21-sensors-15-09404]\]. In addition, several kernel design methods have been proposed to mitigate the cross-term effect \[[@b22-sensors-15-09404]--[@b24-sensors-15-09404]\]. Unfortunately, these techniques need heavy computational complexity when they are applied in a real-time context. A reasonable method is to to introduce a window function in the time domain to reduce the undesired cross-term effects; therefore, the concept of the pseudo Wigner--Ville distribution (PWVD) is educed \[[@b20-sensors-15-09404]\]. Consequently, the window function in PWVD can partially suppress the cross-terms to some ex0tent; the disadvantage of the filtering window operation is the degradation of the resolution, particularly in the frequency domain. In PWVD, the time window operation is equal to frequency filtering in WVD, which can reduce the number of cross-interfering terms by suppressing the interferences between signal components sufficiently separated in time. In order to obtain a better readable result, the cross-terms between components in the frequency domain should be also minimized. Thus, an additional window function is added in order to perform a smoothing in time independently of the frequency smoothing. Therefore, the smoothed version of PWVD, namely the smoothed pseudo Wigner--Ville distribution (SPWVD), can be obtained \[[@b25-sensors-15-09404]\]. The SPWVD is characterized by a separable kernel, which allows the time and frequency smoothing to be adjusted independently, which becomes one of the most versatile of Cohen\'s class TF distributions. The smoothing windows can be adopted to reduce the cross-terms significantly; unfortunately, the SPWVD method also smears localized components, leading to less accurate localization of the signal auto-components in the TF plane compared to the WVD approach. Therefore, a reassignment method can be advantageously applied to improve TF localization properties in SPWVD \[[@b25-sensors-15-09404],[@b26-sensors-15-09404]\]. In this way, the reassigned smoothed pseudo Wigner--Ville distribution (RSPWVD) can be obtained. The RSPWVD method is used to compensate for faults in mapping the TF energy distribution by relocating the value of the neighboring energy to the gravity center rather than the geometric center. In the RSPWVD method, by the adoption of the two-dimensional low-pass filtering smoothing function, the cross-term artifacts present in the quadratic TF distribution can be efficiently eliminated; meanwhile, by the use of the reassignment, the TF localization and aggregation properties of the auto-terms of the signal can be significantly improved. In this paper, an improved TF analysis method by adopting RSPWVD has been proposed in interference detection for GNSS receivers. To the best of our knowledge, this interference detection technique based on joint TF analysis by adopting RSPWVD in the interference detection units for GNSS receivers is new. The performance of the proposed method has been deeply evaluated in comparison with the existing TF analysis approaches. Different localization properties and cross-term effects in the TF plane have been well investigated and compared among the aforementioned TF distributions adopted in interference detection for GNSS receivers. In order to prove the effectiveness of the proposed TF analysis method by adopting RSPWVD in interference detection for GNSS receivers, an experiment is accomplished in the GPS L1 signal, which is characterized in additive white Gaussian noise (AWGN) corrupted by linearly-modulated sweep interference (chirp disturbance). The analysis results show that the proposed joint TF analysis by using RSPWVD eliminates the cross-terms significantly and preserves the high resolution of time and frequency in the TF plane at the same time. This developed improved TF analysis technique by adopting RSPWVD in interference detection makes the spectral characteristic of the interfering term sharply distinguishable among the received GNSS signal, which provides improved readability in the TF plane and enhanced detection performance for GNSS receivers with respect to the state-of-the-art TF analysis approaches. 2.. Signal and System Model =========================== The signal at the input of a GNSS receiver, in a noisy environment with RFI, can be written as: $$y_{RF}\left( t \right) = \sum\limits_{i = 1}^{N_{s}}{r_{RF,i}\left( t \right) + \eta_{RF}\left( t \right)}$$*i.e.*, the sum of *N~s~* useful signals emitted by *N~s~* different satellites and of a disturbing term η*~RF~*(*t*) and *N~s~* is the number of satellites in view. The expression of the signal in space (SIS) transmitted by the *i*-th satellite and received at the GNSS receiver antenna with a propagation delay τ*~i~* is usually assumed as the following structure: $$r_{RF,i}\left( t \right) = A_{i}c_{i}\left( {t - \tau_{i}} \right)d_{i}\left( {t - \tau_{i}} \right)\cos\left\lbrack {2\pi\left( {f_{RF} + f_{d,i}} \right)t + \varphi_{RF,i}} \right\rbrack$$where: *A~i~* is the amplitude of the *i*-th useful signal;τ*~i~* is the code phase delay introduced by the transmission channel;*c~i~*(*t* − τ*~i~*) is the pseudo random noise (PRN) code sequence, which is assumed to take a value in the set {−1,1};*d~i~*(*t* − τ*~i~*) is the bit stream of the navigation message, binary phase-shift keying (BPSK) modulated, including satellite data; and each binary unit is called a bit;*f~d,i~* is the Doppler frequency shift affecting the *i*-th useful signal, and φ*~RF,i~* is the initial carrier phase offset;*f~RF~* is the carrier frequency, and it depends on the GNSS signal band under analysis; in the case of the GPS L1 signal, *f~RF~* = *f~L~*~1~ = 1575.42 MHz. In general, the disturbing term η*~RF~*(*t*) can be expressed as: $$\eta_{RF}\left( t \right) = j_{RF}\left( t \right) + w_{RF}\left( t \right)$$where *j~RF~*(*t*) is a non-stationary RFI and *w~RF~*(*t*) is a zero-mean stationary AWGN process. The interfering signal *j~RF~*(*t*) can assume different forms depending on the jammer that generates it. Several efforts have been devoted to the analysis and characterization of civilian GNSS jammers; despite significant differences, the transmitted jamming signal is usually frequency modulated with an almost constant amplitude. In this paper, the interference term *j~RF~*(*t*) is assumed to be in the class of sweep interference (linear chirp). Sweep interference is one of the main classes of the interfering signals, and its corresponding time-domain function represented by sinusoids can be written as follows: $$j_{RF}\left( t \right) = A_{\textit{inst}}\left( t \right)\cos\left\lbrack {2\pi f_{\textit{inst}}\left( t \right)t + \varphi_{0}} \right\rbrack$$where *A~inst~*(*t*) is the interfering signal amplitude, *f~inst~*(*t*) denotes its instantaneous frequency and φ~0~ is the initial phase of the interference (at time *t* = 0), which can be assumed to be a random variable with a uniform distribution in the range \[−*π*, +*π*). In a linear chirp, the instantaneous frequency *f~inst~*(*t*) of the interfering signal evolves linearly with time over the interval \[*f~RF~*+∆*f*~0~, *f~RF~*+∆*f*~1~\], where *f~RF~* is the GNSS signal center frequency. Therefore, the instantaneous frequency *f~inst~*(*t*) can be expressed as: $$\begin{matrix} {f_{\textit{inst}}\left( t \right) = f_{0} + kt} & {0 \leq t \leq t_{j}} \\ \end{matrix}$$where *f*~0~ is the starting frequency (at time *t* = 0), *f*~0~ = *f~RF~* + *∆f*~0~, *t~j~* is the frequency sweep period of the jamming signal and *k* is the rate of frequency increase or chirp rate, written as: $$k = \frac{f_{1} - f_{0}}{t_{j}} = \frac{\Delta f_{1} - \Delta f_{0}}{t_{j}}$$where *f*~1~ is the final frequency during a specific frequency sweep period, *f*~1~ = *f~RF~* + ∆*f*~1~ and ∆*f*~1~ − ∆*f*~0~ stands for the frequency sweep. The input signal *y~RF~*(*t*) defined in [Equation (1)](#FD1){ref-type="disp-formula"} is received by the receiver antenna, down-converted and filtered by the receiver front-end. Then, the received signal before the analog to digital (A/D) conversion can be written as: $$\begin{array}{ll} {y\left( t \right)} & {= \underset{i = 1}{\overset{N_{s}}{\text{∑}}}{r_{i}\left( t \right) + \eta\left( t \right)}} \\ & {= \underset{i = 1}{\overset{N_{s}}{\text{∑}}}{A_{i}{\widetilde{c}}_{i}\left( {t - \tau_{i}} \right)d_{i}\left( {t - \tau_{i}} \right)\cos\left\lbrack {2\pi\left( {f_{IF} + f_{d,i}} \right)t + \varphi_{i}} \right\rbrack + \eta\left( t \right)}} \\ \end{array}$$where *f~IF~* is the receiver intermediate frequency (IF). The term *c̃~i~*(*t* − τ*~i~*) represents the spreading sequence after filtering of the front-end, and here, the simplifying condition: $${\widetilde{c}}_{i}\left( t \right) \approx c_{i}\left( t \right)$$is assumed and the impact of the front-end filter is neglected. η(*t*) is the down-converted and filtered disturbing component, η\[*t*\] = η\[*t*\] + *w*\[*t*\]. Considering the interference term *j*\[*t*\], the mean power of the sweep interference can be defined as: $$J = \text{Var}\left\{ {j\left\lbrack t \right\rbrack} \right\}$$ The jammer-to-noise ratio (JNR) is defined as follows: $$\frac{J}{N} = \frac{J}{\sigma_{IF}^{2}} = \frac{J}{N_{0}B_{IF}}$$ In order to avoid the cross-terms resulting from the interaction between the positive and negative frequency parts of the spectrum, the analytic representation of the received signal is adopted, provided as follows: $$y_{a}\left\lbrack t \right\rbrack = y\left\lbrack t \right\rbrack + jŷ\left\lbrack t \right\rbrack$$where the analytic signal *y~a~*\[*t*\] has a real part *y*\[*t*\] and an imaginary part *y*\[*t*\], which contains the Hilbert transform of *y*\[*t*\]. The imaginary part is a version of the original real part with a 90° phase shift. The use of the analytic signal has two advantages: first, interference between positive and negative frequencies can be eliminated, as the analytic signal *y~a~*\[*t*\] has components belonging only to the half plane of positive frequencies; second, even though this filtering suppresses negative frequencies and the sampling rate is reduced, it does not introduce any loss of information. 3.. Time Frequency Transforms ============================= The classical method for analyzing a signal with time-varying frequency content is to split the time-domain signal into many segments. The signal to be transformed is multiplied by a window function, which is nonzero for only a short period of time, and then, take the Fourier transform of each segment as the window slid along the time axis, resulting in a two-dimensional representation of the signal. This is known as the short-time Fourier transform (STFT) operation, which is the most widely-used method for analyzing non-stationary signals. Additionally, simply, in the continuous time case, it is defined as: $$\textit{STFT}\left( {t,\omega} \right) = {\int_{- \infty}^{+ \infty}{y_{a}\left( \tau \right)h\left( {\tau - t} \right)e^{- j\omega\tau}d\tau}}$$where *h*(*t*) is the analysis window, which is commonly a real and even window function centered on zero, *y~a~*(*t*) is the defined analytical signal to be transformed and *STFT*(*t*, ω) is a function of *t* and ω, which is linear and depends on the chosen window *h*. In order to understand the time properties at a particular frequency, the definition of the STFT can also be expressed in the frequency domain by manipulating [Equation (12)](#FD12){ref-type="disp-formula"}, obtaining the following result: $$\textit{STFT}\left( {t,\omega} \right) = \frac{1}{2\pi}e^{- j\omega t}{\int_{- \infty}^{+ \infty}{Y_{a}{(\omega^{\prime})}H{({\omega^{\prime} - \omega})}e^{j\omega^{\prime}t}d\omega^{\prime}}}$$where *H*(ω) is the frequency window function, which is the Fourier transform of *h*(*t*). The dual relationship between [Equation (12)](#FD12){ref-type="disp-formula"} and [(13)](#FD13){ref-type="disp-formula"} is apparent; the TFR can be generated via a moving window in time or a moving window in frequency. In the STFT analysis, one intends to achieve both high time and frequency resolution if possible. However, the resolution in the time domain is limited by the width of the window function *h*(*t*); similarly, the resolution in the frequency domain is limited by the width of the frequency window *H*(ω). As a result, the choice of a window to represent the signal by its spectrogram imposes a compromise between the conversation of temporal localization and that of frequency localization. This compromise is due to Heisenberg\'s uncertainty principle, which states that the window width in time and the window width in frequency are inversely proportional to each other. Therefore, choosing a small time window leads to good resolution in time and necessarily implies poor resolution in frequency; conversely, a long time window yields poor time resolution, but good frequency resolution. The length of the window plays a fundamental role in this TF compromise. The squared magnitude of the STFT, denoted by *SPEC*(*t*, ω), is called spectrogram, which can be written as follows: $$\textit{SPEC}\left( {t,\omega} \right) = \left| {\textit{STFT}\left( {t,\omega} \right)} \right|^{2}$$where *STFT*(*t*, ω) is the STFT defined in [Equation (12)](#FD12){ref-type="disp-formula"}. The spectrogram of the signal has poor TF localization properties due to the presence of analysis window function. In order to avoid the TF resolution trade-off problem of the spectrogram, WVD is adopted in interference detection for GNSS receivers. In the WVD, a time-dependent instantaneous auto-correlation function is chosen as: $$R\left( {t,\tau} \right) = y_{a}(t + \frac{\tau}{2})y_{a}^{*}(t - \frac{\tau}{2})$$ The WVD of *y~a~*(*t*) is then defined as the Fourier transform of this time-dependent instantaneous auto-correlation function \[[@b27-sensors-15-09404]\], written as follows: $$\textit{WVD}\left( {t,\omega} \right) = \frac{1}{2\pi}{\int_{- \infty}^{+ \infty}{R\left( {t,\tau} \right)e^{- j\omega\tau}\text{d}\tau = \frac{1}{2\pi}{\int_{- \infty}^{+ \infty}{y_{a}(t + \frac{\tau}{2})y_{a}^{*}(t - \frac{\tau}{2})e^{- j\omega\tau}\text{d}\tau}}}}$$where *y~a~*(·) denotes the signal to be analyzed, *t* is the time, ω represents the angular frequency, τ is called the lag variable and (\*) denotes the complex conjugate. In addition, this class of bilinear (or quadratic) TF distributions can be most easily understood in terms of the ambiguity function. If the inverse Fourier transform of the instantaneous auto-correlation function *R*(*t*, τ) is taken with respect to *t* instead of τ, the ambiguity function can be obtained as follows: $$AF\left( {\tau,\theta} \right) = {\int_{- \infty}^{+ \infty}{R\left( {t,\tau} \right)e^{j\theta t}\text{d}t = {\int_{- \infty}^{+ \infty}{y_{a}(t + \frac{\tau}{2})y_{a}^{*}(t - \frac{\tau}{2})e^{j\theta t}\text{d}t}}}}$$ The ambiguity function can be used to monitor the disturbing effect in the received GNSS signals. The WVD has a number of desirable properties that make it a good indicator of how the energy of the signal can be viewed as a function of time and frequency. First, the WVD of any signal is always real. Second, it satisfies the time marginal condition: $$\int_{- \infty}^{+ \infty}{\textit{WVD}\left( {t,\omega} \right)\text{d}\omega = \left| {y_{a}\left( t \right)} \right|^{2}}$$ That is, by summing the TF distribution over all frequencies, the instantaneous energy of the signal at a particular time instant can be obtained. Similarly, the WVD also satisfies the frequency marginal condition: $$\int_{- \infty}^{+ \infty}{\textit{WVD}\left( {t,\omega} \right)\text{d}t = \left| {Y_{a}\left( \omega \right)} \right|^{2}}$$ In this case, by summing the TF distribution over all time, the power spectrum of the signal at a particular frequency can be obtained. Although WVD has many good properties and provides nearly the best resolution among all of the TF techniques, its main drawback comes from undesirable cross-term interference. The WVD is said to be bilinear, because the analyzed signal enters twice in its calculation. Consider the signal *y*(*t*) = *y*~1~(*t*) + *y*~2~(*t*), where *y*(*t*), *y*~1~(*t*) and *y*~2~(*t*) are analytic. Expanding the instantaneous auto-correlation function of *y*(*t*), we can obtain: $$R_{y}\left( {t,\tau} \right) = R_{y_{1}}\left( {t,\tau} \right) + R_{y_{2}}\left( {t,\tau} \right) + R_{y_{1}y_{2}}\left( {t,\tau} \right) + R_{y_{2}y_{1}}\left( {t,\tau} \right)$$where *R~y~*~~1~~*~y~*~~2~~(*t*, τ) and *R~y~*~~2~~*~y~*~~1~~(*t*, τ) are the instantaneous cross-correlation functions (e.g., $\left. R_{y_{1}y_{2}}\left( {t,\tau} \right) = y_{1}\left( {t + \frac{\tau}{2}} \right)y_{2}^{*}\left( {t - \frac{\tau}{2}} \right) \right)$. Taking Fourier transforms of [Equation (20)](#FD20){ref-type="disp-formula"} with respect to τ, it is easy to know that: $$WVD_{y}\left( {t,\omega} \right) = WVD_{y_{1}}\left( {t,\omega} \right) + WVD_{y_{2}}\left( {t,\omega} \right) + 2\operatorname{Re}\left\{ {WVD_{y_{1}y_{2}}\left( {t,\omega} \right)} \right\}$$where *WVD~y~*~~1~~(*t*, ω) and *WVD~y~*~~2~~(*t*,ω) are the WVDs of *y*~1~(*t*) and *y*~2~(*t*), respectively, and the last term is the cross-WVD (XWVD) between *y*~1~(*t*) and *y*~2~(*t*), provided as: $$WVD_{y_{1}y_{2}}\left( {t,\omega} \right) = \frac{1}{2\pi}{\int_{- \infty}^{+ \infty}{y_{1}(t + \frac{\tau}{2})y_{2}^{*}(t - \frac{\tau}{2})e^{- j\omega\tau}\text{d}\tau}}$$ Thus, the WVD of the sum of two signals is not the sum of their corresponding WVDs, but also of their XWVDs. This means that the spectrum energy density of the sum of two signals does not reduce to the sum of the individual densities (unless the signals are spectrally disjoint). If *y*~1~(*t*) and *y*~2~(*t*) are mono-component signals, *WVD~y~*~~1~~(*t*,ω) and *WVD~y~*~~2~~(*t*,ω) are the auto-terms, while 2Re{*WVD~y~*~~1~~*~y~*~~2~~(*t*, ω)} is a cross-term. As a result, if a signal contains more than one component, in the TF plane, its WVD suffers from spurious features containing cross-terms that occur halfway between each pair of auto-terms. The magnitude of these oscillatory cross-terms can be twice as large as the auto-terms, and they do not possess any physical meaning. As an example, the time-domain signal, which contains four Gaussian components, is shown in [Figure 1](#f1-sensors-15-09404){ref-type="fig"}. The WVD of this signal is provided in [Figure 2a](#f2-sensors-15-09404){ref-type="fig"}, and correspondingly, the contour of the computed WVD is presented in [Figure 2b](#f2-sensors-15-09404){ref-type="fig"}. From [Figure 2](#f2-sensors-15-09404){ref-type="fig"}, there exist four peaks, which respectively denote the corresponding auto-terms of the four Gaussian components in the TF plane; in addition, the WVD also presents six cross-terms, which occur between each pair of Gaussian signal components, and among them, two cross-terms overlap in the diagonal intersection point of the rectangle connected by the four vertices (*i.e.*, four Gaussian components in the joint TF plane). These extra cross-terms have large oscillating amplitudes due to the interaction of the different signal components. The magnitudes of the oscillatory cross-terms can be twice as large as the auto-terms, but in the diagonal intersection position, the intensity of the corresponding cross-term can be four-times as large as the auto-terms, since two cross-terms overlap in this intersection point. In [Figure 2b](#f2-sensors-15-09404){ref-type="fig"}, the serious cross-terms apparently present in the regions of the TF plane where we expect no energy at all, which make proper interpretation impossible. This is the main drawback of the WVD approach. A feasible method to depress the effect of the cross-terms is to introduce windows in WVD; therefore, the concept of PWVD is educed. The WVD weighs equally all times of the future and past for a given time, but in the practical calculation of the distribution for a time instant *t*, we may concentrate the properties of the signal near the time of interest rather than from minus to plus infinity. Therefore, if we want to emphasize the signal around time *t*, a window function *h*(τ) can be multiplied with the instantaneous ambiguity function $y_{a}\left( {t + \frac{\tau}{2}} \right)y_{a}^{*}\left( {t - \frac{\tau}{2}} \right)$ in order to define the PWVD, which is the WVD windowed in the time direction \[[@b20-sensors-15-09404]\], written as follows: $$\textit{PWVD}\left( {t,\omega} \right) = \frac{1}{2\pi}{\int_{- \infty}^{+ \infty}{h\left( \tau \right)y_{a}(t + \frac{\tau}{2})y_{a}^{*}(t - \frac{\tau}{2})\text{e}^{- j\omega\tau}\text{d}\tau}}$$where *PWVD*(*t*,ω) is controlled by the window *h*(τ), which is an even function and peaked around *τ* = 0. Since the WVD is highly nonlocal, the window function is adopted in the PWVD to make it local. The overlapped window function contributes the cross-term suppression for multi-component signals. The operation of time-window overlay is equal to the frequency filtering in WVD; consequently, the interference between time-shifted signals is usually attenuated. An example of PWVD of the four Gaussian components signal is provided in [Figure 3a](#f3-sensors-15-09404){ref-type="fig"}, and correspondingly, the contour of the computed PWVD is presented in [Figure 3b](#f3-sensors-15-09404){ref-type="fig"}. From the results, it is easy to find that only two cross-terms remain in the TF plane, and the number of the extra cross-terms decreases from six to two, since the the cross-interfering term between each pair of time-shifted signal components is attenuated. In addition, from [Figure 3b](#f3-sensors-15-09404){ref-type="fig"}, it is clear to observe that the TF localization properties of the auto-terms of the analyzed signal are deteriorated in comparison with the WVD approach. In the PWVD, the cross-interfering terms have been partially mitigated by the adoption of the window function. The time window operation is equal to frequency filtering in WVD, which can reduce the number of cross-interfering terms by suppressing the interferences between signal components sufficiently separated in time. However, these advantages are achieved at the price of a blurring of the auto-terms of the signal and a loss of many desirable theoretical properties. 4.. Interference Detection Based on Joint TF Analysis by Using RSPWVD ===================================================================== Although the WVD has many good properties and provides nearly the best resolution among all of the TF analysis techniques, due to its intrinsic quadratic nature, it suffers from cross-term interference when it is applied to multi-component signals. This drawback severely hinders the usefulness of the WVD for detecting RFI characteristics in the TF plane. In Section 3, the PWVD has been used to suppress the cross-terms for multi-component signals; however, many desirable properties of the WVD, such as marginals and instantaneous frequency, are annihilated, and the TF concentration property is also attenuated. In order to eliminate the cross-terms present in the quadratic TF distribution and to preserve good time and frequency resolution at the same time, a reassigned SPWVD method has been proposed in interference detection for GNSS receivers. 4.1.. SPWVD ----------- The unsatisfactory results obtained with the existing TF distributions justify the search for better tools; one way of achieving this is to start from the general form of quadratic representations. All of these existing TF distributions could be written in a generalized form, which can be used to facilitate the design of desirable TF transforms. This class of transform is known as Cohen\'s class \[[@b20-sensors-15-09404]\], and the definition of Cohen\'s class of bilinear (or quadratic) TF distributions can be written as follows: $$\begin{array}{ll} {C\left( {t,\omega} \right)} & {= \frac{1}{4\pi^{2}}{\int_{- \infty}^{\infty}{\int_{- \infty}^{\infty}{AF\left( {\tau,\theta} \right)g\left( {\tau,\theta} \right)\textit{e}^{- j{({\theta t + \omega\tau})}}\text{d}\theta\text{d}\tau}}}} \\ & {= \frac{1}{4\pi^{2}}{\int_{- \infty}^{\infty}{\int_{- \infty}^{\infty}{\int_{- \infty}^{\infty}{y_{a}(\textit{s} + \frac{\tau}{2})y_{a}^{*}(\textit{s} - \frac{\tau}{2})g\left( {\tau,\theta} \right)\textit{e}^{- j{\lbrack{\theta{({t - s})} + \omega\tau}\rbrack}}\text{d}s\text{d}\theta\text{d}\tau}}}}} \\ \end{array}$$where *AF*(τ, θ) is the ambiguity function defined in [Equation (17)](#FD17){ref-type="disp-formula"}; and *g*(τ, θ) is a two-dimensional parametrization function defined in the ambiguity function domain, which is called the kernel function of Cohen\'s class. This kernel function determines the properties of the bilinear TF distribution, which is often a low-passing function and normally serves to mask out the interference in the original Wigner--Ville representation. When considering *g*(τ, θ) = 1, WVD can be obtained. Cohen\'s class can be also rewritten as the double convolution of the WVD of the signal *y~a~*(*t*) and a two-dimensional smoothing function, provided as follows: $$\begin{array}{ll} {C{({t,\omega;\Pi})}} & {= W_{ya}{({t,\omega})}**\Pi{({t,\omega})}} \\ & {= \frac{1}{4\pi^{2}}{\int_{- \infty}^{\infty}{\int_{- \infty}^{\infty}{\prod\left( {t - s,\omega - \theta} \right)W_{y_{a}}\left( {s,\theta} \right)\text{d}s\text{d}\theta}}}} \\ \end{array}$$ $$\prod\left( {t,\omega} \right) = {\int_{- \infty}^{\infty}{\int_{- \infty}^{\infty}{g\left( {\tau,\theta} \right)e^{- j{({\omega\tau + \theta t})}}\text{d}\tau\text{d}\theta}}}$$where (\*\*) denotes the double convolution operation, Π(*t*, ω) is a two-dimensional smoothing function and *W~ya~*(*s*,θ) is the WVD of the signal *y~a~*(*t*). Different TFRs can be obtained from the fundamental WVD by applying a different smoothing function Π(*t*, ω). When considering WVD, Π(*t*, ω) = δ(*t*)δ(ω). If a separable smoothing function is considered, it can be written as the product of windows from both time and frequency domains: $$\prod\left( {t,\omega} \right) = g\left( t \right)H\left( {- \omega} \right)$$where *H*(ω) is the Fourier transform of the window function *h*(*t*), which allows the smoothing of the cross-interfering terms oscillating in parallel with the frequency axis (frequency smoothing); and the window function *g*(*t*) allows the smoothing of the cross-interfering terms oscillating in parallel with the temporal axis (temporal smoothing). Obviously, the wider the spreading, the more the smoothing. Therefore, the smoothing in the ambiguity function domain combined with the parametrization function allows both the suppression of the cross-terms and the preservation of the auto-ambiguity terms of the analyzed signal. Since the cross-terms with the WVD are strongly oscillating, the most effective way of removing cross-term interference is to apply two-dimensional low-pass filtering in the ambiguity domain. The resulting two-dimensional convolution of the WVD in [Equation (25)](#FD25){ref-type="disp-formula"} defines the smoothed version of PWVD, that is SPWVD, which can be written as follows: $$\textit{SPWVD}\left( {t,\omega;g,h} \right) = {\int_{- \infty}^{\infty}{h\left( \tau \right){\int_{- \infty}^{\infty}{g{({t^{\prime} - t})}y_{a}(t^{\prime} + \frac{\tau}{2})y_{a}^{*}(t^{\prime} - \frac{\tau}{2})e^{- j\omega\tau}\text{d}t^{\prime}\text{d}\tau}}}}$$ Therefore, an independent and progressive control can be applied to the WVD in both time and frequency directions. The independency of *h*(*t*) and *g*(*t*) makes SPWVD more flexible to reduce the cross-terms present in WVD. 4.2.. Reassignment Method ------------------------- Compared with WVD, SPWVD can be used to effectively depress the influence of the cross-interfering terms of a multi-component signal, but its TF concentration and localization properties decrease somewhat. In order to improve the TF aggregation properties in SPWVD, a reassignment method has been considered. From [Equation (25)](#FD25){ref-type="disp-formula"}, we can know that the two-dimensional smoothing function Π(*t* − *s*, ω − θ) determines a certain TF region at the neighborhood nearby the point (*t*, ω), inside which a weighted average of the WVD *W~ya~*(*s*, θ) of the signal *y~a~*(*t*) is performed. However, these mean values may not be symmetrically distributed around a certain point (*t*, ω), which is the geometrical center of the TF region. Consequently, the point (*t*, ω) is not truly representative for such a region. In contrast, the energy gravity center of such a region is more approximate to represent the local energy distribution of the analyzed signal. The local energy distribution Π(*t* − *s*, ω − θ)*C*(*s*, θ; Π) of Cohen\'s class TF distribution can be assumed as the distribution of mass, and it is better to assign the total mass to the gravity center rather than to its geometrical center. In this paper, the reassignment method is considered to relocate each value of Cohen\'s class distribution *C*(*t*, ω) at any point (*t*, ω) to another point (*t̂*, ω̂), which is the gravity center of the signal\'s energy distribution around the point (*t*, ω). Then, the reassigned Cohen\'s class TF distribution can be defined as follows: $$C^{(r)}{({t^{\prime},\omega^{\prime};\prod})} = {\int_{- \infty}^{\infty}{\int_{- \infty}^{\infty}{C{({t,\omega;\prod})}\delta{({t^{\prime} - \hat{t}{({y_{a};t,\omega})}})}\delta{({\omega^{\prime} - \hat{\omega}{({y_{a};t,\omega})}})}\text{d}t\text{d}\omega}}}$$where: $$\hat{t}\left( {y_{a};t,\omega} \right) = \frac{\int_{- \infty}^{\infty}{\int_{- \infty}^{\infty}{s\prod{\left( {t - s,\omega - \theta} \right)W_{y_{a}}\left( {s,\theta} \right)\text{d}s\text{d}\theta}}}}{\int_{- \infty}^{\infty}{\int_{- \infty}^{\infty}{\prod{\left( {t - s,\omega - \theta} \right)W_{y_{a}}\left( {s,\theta} \right)\text{d}s\text{d}\theta}}}}$$ $$\hat{\omega}\left( {y_{a};t,\omega} \right) = \frac{\int_{- \infty}^{\infty}{\int_{- \infty}^{\infty}{\theta\prod{\left( {t - s,\omega - \theta} \right)W_{y_{a}}\left( {s,\theta} \right)\text{d}s\text{d}\theta}}}}{\int_{- \infty}^{\infty}{\int_{- \infty}^{\infty}{\prod{\left( {t - s,\omega - \theta} \right)W_{y_{a}}\left( {s,\theta} \right)\text{d}s\text{d}\theta}}}}$$*C*^(^*^r^*^)^(*t′*, ω′; Π) is Cohen\'s class TF distribution after reassignment. Therefore, if a suitable smoothing kernel is selected, all of the bilinear TF distributions after reassignment are able to mitigate the cross-terms and keep high TF concentration properties at the same time. In particular, when the reassignment method is applied to SPWVD, RSPWVD can be obtained, which is calculated as follows: $$\textit{RSPWVD}{({t^{\prime},\omega^{\prime};g,h})} = {\int_{- \infty}^{\infty}{\int_{- \infty}^{\infty}{\textit{SPWVD}{({t,\omega;g,h})}\delta{({t^{\prime} - \hat{t}{({y_{a};t,\omega})}})}}}} \cdot \delta{({\omega^{\prime} - \hat{\omega}{({y_{a};t,\omega})}})}\text{d}t\text{d}\omega$$where: $$\hat{t}{({y_{a};t,\omega})} = t - \frac{\textit{SPWVD}{({t,\omega;\tau_{g},h})}}{2\pi\textit{SPWVD}\left( {t,\omega;g,h} \right)}$$ $$\hat{\omega}{({y_{a};t,\omega})} = \omega + j\frac{\textit{SPWVD}\left( {t,\omega;g,D_{h}} \right)}{2\pi\textit{SPWVD}\left( {t,\omega;g,h} \right)}$$with τ*~g~* = *g*(*t*) and $D_{h}(t) = \frac{\text{d}h(t)}{\text{d}t}$. The RSPWVD *RSPWVD*(*t′*, ω′; *g*, *h*) can be used to eliminate the cross-term artifacts inherent in the quadratic TF distributions, which presents good resolution in both the time and frequency domains; therefore, in this work, the joint TF analysis based on RSPWVD has been firstly proposed in interference detection for GNSS receivers, which is clearly illustrated in [Figure 4](#f4-sensors-15-09404){ref-type="fig"}. The double convolution of the WVD *W~y~a~~*(*t*, ω) of the analytical signal and the two-dimensional smoothing function Π(*t*, ω) is performed to obtain the *SPWVD*(*t*, ω; *g*, *h*); after reassignment, the cross-term-free TF distribution *RSPWVD*(*t′*, ω′; *g*, *h*) has good readability, which is suitable for multi-component signal analysis. The proposed joint TF analysis by using RSPWVD makes the spectral characteristic of the interfering signal sharply distinguishable among the received GNSS signal; therefore, the instantaneous frequency of the disturbing term in the received GNSS signals can be effectively estimated by detecting the peaks of the proposed joint TF distribution. 5.. Performance Evaluation ========================== In this section, the performance of the proposed interference detection method based on joint TF analysis by using RSPWVD is analyzed. In particular, this proposed new algorithm, which adopts RSPWVD to detect sweep interference in GNSS receivers, is compared with the conventional interference detection approaches in the disturbing scenario. The mentioned interference detection approaches are tested on real GPS data collected by using the GNSS software receiver developed at Beihang University \[[@b28-sensors-15-09404]\]. The scheme of the test is reported in [Figure 5](#f5-sensors-15-09404){ref-type="fig"}, while an image of the experimental setup adopted for collecting the GPS data corrupted by sweep interference is depicted in [Figure 6](#f6-sensors-15-09404){ref-type="fig"}. The real GPS samples are collected by using the GNSS software receiver connected to the Trimble Zephyr Geodetic 2 antenna placed on the roof of the new main building at Beihang University in an open-sky static condition; the developed software interferer is adopted for generating the sweep interfering signal for GNSS applications. The generated sweep interference is added to the GPS samples collected by the GNSS receiver front-end. In the experiment, the scenario adopted for the test is characterized by the setting parameters provided in [Table 1](#t1-sensors-15-09404){ref-type="table"}, representing the real GPS L1 signal in zero mean Gaussian noise corrupted by a constant amplitude linearly frequency modulated interference (linear chirp), which is commonly considered as a test bench in interference detection for GNSS receivers. In the experiment settings, the JNR value of the sweep interference is set to be −1 dB; in a linear chirp, the instantaneous frequency *f~inst~* of the interfering signal evolves linearly with time over the interval \[*f*~L1~ + ∆*f*~0~, *f*~L1~ + ∆*f*~1~\], where *f*~L1~ is the GPS L1 signal center frequency, ∆*f*~0~ = +7 MHz and ∆*f*~1~ = −9 MHz. In [Figure 7a](#f7-sensors-15-09404){ref-type="fig"}, the ambiguity function of the GPS L1 signal without interference is shown as a spike. In the case with the presence of sweep interference, the ambiguity function of the interfered GPS L1 signal is depicted in [Figure 7b](#f7-sensors-15-09404){ref-type="fig"}, where the disturbing effect can be clearly observed. The adoption of the ambiguity function of the received interfering signal is beneficial to better monitor the interference contribution in the received GNSS signals. In [Figure 8](#f8-sensors-15-09404){ref-type="fig"}, the spectrogram of the GPS L1 signal with sweep interference is depicted, where the Hamming window function is chosen. In [Figure 8a](#f8-sensors-15-09404){ref-type="fig"}, the window size is 63 samples, where the disturbing term more or less emerges in the TF plane, but very poor TF localization properties are obtained with this approach. In order to evaluate the TF characteristics of the spectrogram by increasing the length of the analysis window, the spectrogram of the interfered GPS L1 signal is presented in [Figure 8b](#f8-sensors-15-09404){ref-type="fig"}, where the window length is increased to 127 samples. It is clear that a long time window leads to improved frequency resolution and inevitably yields poor resolution in time. The spectrogram is nonlinear, but this nonlinearity results from the operation of squared magnitude and, therefore, does not lead to undesirable cross-terms present in the WVD. In practice, the spectrogram approach cannot be used to provide the instantaneous frequency estimate for the interfering signal, although sometimes, a good approximation can be achieved. In [Figure 9a](#f9-sensors-15-09404){ref-type="fig"}, the WVD of the GPS L1 signal with the sweep interference is depicted, and correspondingly, the contour of the computed WVD is presented in [Figure 9b](#f9-sensors-15-09404){ref-type="fig"}. The interfering term presents a linear behavior in frequency, which is well localized in the restricted portion of the TF plane. The price that is paid for high TF resolution with the WVD approach is the undesirable cross-terms. Due to the interaction of the different signal components, the presence of the serious cross-terms is apparently observed in [Figure 9b](#f9-sensors-15-09404){ref-type="fig"}, which makes interpretation quite difficult and brings much error for the estimation of the instantaneous frequency of the interfering signal. In [Figure 10a](#f10-sensors-15-09404){ref-type="fig"}, the PWVD of the GPS L1 signal with the sweep interference is depicted, and correspondingly, the contour of the computed PWVD is presented in [Figure 10b](#f10-sensors-15-09404){ref-type="fig"}. It is clear that although the sweep interfering term presents a linear behavior in the TF plane, the PWVD presents less accurate TF localization precision, since the TF concentration property of the PWVD is attenuated. In addition, the cross-terms can be still observed in the TF plane by using the PWVD approach. In order to mitigate the cross-interfering terms present in the WVD, a two-dimensional smoothing function is used to smooth out the interfering signal in both the time and frequency domains. In [Figure 11a](#f11-sensors-15-09404){ref-type="fig"}, the SPWVD of the interfered GPS L1 signal is provided, where the peaks of the SPWVD are clearly observed, denoting the sweep interference contribution. The corresponding contour of the SPWVD is provided in [Figure 11b](#f11-sensors-15-09404){ref-type="fig"}, and it is easy to find that the undesired sweep interference shows a linear behavior in frequency, which is localized in the linear portion of the TF plane. From [Figure 11b](#f11-sensors-15-09404){ref-type="fig"}, it is very clear that the undesired cross-interfering terms are partially mitigated by the adoption of the two-dimensional smoothing function in the SPWVD. Although the adoption of the smoothing window in SPWVD is beneficial to suppress the cross-terms in the TF plane, it smears localized components, leading to less accurate TF localization of the auto-terms of the signal, as compared to the WVD approach. The disadvantage of this filtering window operation is that it limits the original excellent TF resolution features. Therefore, in order to remove the cross-terms present in the quadratic TF distribution and to preserve the good TF energy concentration properties at the same time, an improved TF distribution by adopting RSPWVD has been proposed in interference detection for GNSS receivers. This proposed joint TF distribution of the interfered GPS L1 signal is depicted in [Figure 12a](#f12-sensors-15-09404){ref-type="fig"}, and the corresponding contour of the calculated TF distribution of RSPWVD is provided in [Figure 12b](#f12-sensors-15-09404){ref-type="fig"}. From the results, it is clear that the undesired sweep interfering term presents a very strict linear behavior in the TF plane with the proposed joint TF distribution based on RSPWVD, which provides representations that are easy to interpret in interference detection for GNSS receivers. This proposed improved TF analysis technique effectively eliminates the cross-term artifacts present in the quadratic TF distributions, which shows good readability in the TF plane. In the reassigned method, the smoothing window is moved from the geometrical center to the energy gravity center of the TF distribution. Therefore, with the proposed RSPWVD method, the disturbing interference is highly localized in the restricted region of the TF plane, and a much improved TF localization property can be achieved in comparison to the SPWVD approach. The proposed joint TF distribution based on RSPWVD has been proven to be valid and effective in interference detection for GNSS receivers in jamming environments. This proposed joint TF analysis technique based on RSPWVD effectively solves the cross-term problem of the bilinear TF distribution and keeps good time and frequency resolution in the TF plane at the same time, which has been proven to be effective to adopt in interference detection for GNSS receivers. With this developed interference detection technique, the interfering signal can be correctly identified, and its instantaneous frequency can be precisely estimated. In the following interference excision/mitigation unit (anti-jamming device), the estimated instantaneous frequency of the interfering signal will be further used to control the coefficients of an excision filter (notch filter) that adaptively removes the disturbing interference signal. 6.. Conclusions =============== In this paper, an improved TF analysis method based on RSPWVD has been proposed to detect sweep interference for GNSS receivers. In order to prove the advantages and effectiveness of the developed technique, a comprehensive performance comparison has been carried out compared with the existed TF analysis approaches. The experiments have been performed on GPS L1 signals in the disturbing scenario in order to support the theoretical analysis among the aforementioned TF distributions adopted in interference detection for GNSS receivers. From the analysis results, the spectrogram approach presents the TF resolution trade-off problems and provides poor localization properties in the TF plane; the interference detection based on WVD presents a severe cross-interfering term problem due to the interaction of different frequency components, which makes proper disturbance interpretation impossible. Then, the SPWVD method is adopted to partially mitigate the cross-terms present in the bilinear TF distribution, since a two-dimensional smoothing function is added in both the time and frequency domains, but the disadvantage of using filtering windows is the degradation of time and frequency resolution. In order to eliminate the cross-term problem and preserve good time and frequency resolution in the TF plane at the same time, an improved joint TF distribution by using RSPWVD has been firstly proposed in interference detection for GNSS receivers. The interference detection based on joint TF distribution by using RSPWVD efficiently combines the removal of the cross-interfering terms provided by a two-dimensional smoothing kernel function and an increased TF concentration of the auto-terms of the signal components achieved by the reassignment. From the analysis results, the proposed interference detection method based on RSPWVD successfully overcomes the cross-term problem and, meanwhile, presents good TF localization properties, which provide much improved interference detection performance in comparison with the existing TF analysis approaches. The developed interference detection technique based on joint TF analysis by adopting RSPWVD has been proven to be suitable to improve the interference detection performance in the jamming environments, which is promising for adoption in the anti-interference GNSS receiver design for civil aviation and military purposes, particularly in disturbing environments. This research work is supported by the Scientific Research Foundation for Returned Scholars, Ministry of Education of China. Kewen Sun conceived the work, designed the algorithms, performed the experiment, analyzed the experiment results and wrote the manuscript. Tian Jin helped during the phase of GPS data collection. Dongkai Yang commented on the work. The authors declare no conflict of interest. {#f1-sensors-15-09404} {#f2-sensors-15-09404} {#f3-sensors-15-09404} {#f4-sensors-15-09404} {#f5-sensors-15-09404} {#f6-sensors-15-09404} {#f7-sensors-15-09404} {#f8-sensors-15-09404} {#f9-sensors-15-09404} {#f10-sensors-15-09404} {#f11-sensors-15-09404} {#f12-sensors-15-09404} ###### Experimental setting parameters. **Parameter** **Value** ------------------------------------ ------------ Carrier-to-noise ratio, *C*/*N*~0~ 46 dB-Hz Sampling frequency, *f~s~* 24 MHz Intermediate frequency, *f~IF~* 40.42 MHz Code length 1023 chips Sweep period 1 ms Spectrogram analysis window Hamming

Introduction {#rjw174s1} ============ Human amniotic membrane consists of an epithelial monolayer, a thick basement membrane, and an avascular connective tissue matrix. Placental donations from healthy mothers delivering via elective cesarean section provide the source. The first use was in 1910 and then subsequently utilized in a variety of reconstruction procedures, as a surgical dressing and in treatment of diabetic and burn wounds \[[@rjw174C1]\]. Concerns about possible transmission of disease led to abandonment of human amniotic tissue through the 1980s. Its use regained popularity followed reassurance of clinical safety, extending to use in ocular surface disease and spine surgery \[[@rjw174C2]--[@rjw174C5]\]. Literature review provided no references of human amniotic membrane used as a biologic dressing to aid reconstruction of the nose. Evidence demonstrating efficacy in epithelialization and pain reduction along with anti-inflammatory and anti-scarring properties pointed to human amniotic membrane as an ideal option for nasal reconstruction \[[@rjw174C6]--[@rjw174C8]\]. This case report presents a patient in which dehydrated human amniotic membrane (EpiFix; MiMedex Group, Inc., Kennesaw, GA) was used to facilitate healing of the lower third of the nose. Case Presentation {#rjw174s2} ================= A 56-year-old male presented status post assault with human bites to the tip of his nose, left ala, right ear and first digit of his left hand. He denied nasal obstruction and no facial fractures were associated. He denied of consciousness and reported no other trauma. Past medical history was significant for hypertension controlled with Lisinopril, diet and exercise. Patient denied tobacco and illicit drug use but reported occasional alcohol use. Physical exam revealed a well-developed, well-nourished adult Caucasian male in no acute distress. There was a 22 × 18 mm full thickness skin and subcutaneous tissue defect present on the nasal tip. The columella and left ala were similarly involved. The perichondrium and nasal cartilage remained intact and undamaged. Nasopharyngoscopy with a zero degree nasopharyngoscope revealed a slight right septal deviation without evidence of intranasal trauma, discharge or mucosal bleeding. Involvement of the patient\'s right auricular pinna was very superficial and treatment was deferred. No facial lacerations, bony abnormalities or palpatory tenderness was appreciated. Facial weakness and neurologic deficits were also no appreciated. Bilateral involvement of the hands and upper extremities also appeared very superficial and treatment was deferred. Debridement of the nasal tip and adjacent structures was performed using standard sterile technique and local anesthesia. The defect measured 25 × 20 mm following debridement and the perichondrium and nasal cartilage was confirmed intact and without evidence of disease or injury. Dehydrated human amniotic membrane was fashioned to the dimensions of the defect and affixed to the wound. The allograft did not overly the native integument. Suture fixation was deferred. Sterile wet to dry dressing was applied and the patient was instructed to avoid recurrent trauma to, and washing of, the graft site. The patient was instructed to maintain a wet to dry dressing over the graft site at all times and to avoid displacement of the amniotic graft during dressing changes. Total procedure time was less than 20 minutes. The patient returned on postoperative date \#3 (POD \#3) for wound check. The amniotic graft was undisturbed from initial placement and the wound bed was moist and clean. Patient provided wet to dry dressing changes continued for another 4 days at which time a simple bandage was applied. Again, the patient was instructed to avoid manipulation of the graft. Follow-up continued regularly at 10--14 day intervals for the duration of treatment. The wound remained clean and had no signs of infection during the healing period. No additional surgical or resurfacing procedures were required. The patient did not require medical assistance for wound care which included twice daily wet to dry dressing changes for 7 days postoperatively. Incorporation of the graft matrix at 7 days postoperatively was considered sufficient to defer further wet to dry dressing changes with conversion to simple bandage coverage with daily changes. Complete healing was noted at 3-month follow-up. Results demonstrated minimal nasal tip deviation and secondary scar contraction. Patient satisfaction was absolute and he denies any nasal obstruction. Discussion {#rjw174s3} ========== This case suggests that dehydrated human amniotic membrane provides an alternative to local tissue transfer and skin grafting for traumatic injuries involving the nose. Other studies have demonstrated dehydrated human amniotic membrane to be effective as an allograft in ocular surface reconstruction and in healing chronic wounds and burns \[[@rjw174C2]--[@rjw174C5], [@rjw174C9]\]. Human amniotic membranes functions through multiple properties. The extracellular matrix components provide scaffolding for cell migration. Cytokines within the extracellular matrix include epidermal growth factor, keratinocyte growth factor, transforming growth factor and basic fibroblast growth factor. These cytokines promote re-epithelialization through support of keratinocyte proliferation and differentiation \[[@rjw174C10]\]. Anti-inflammatory properties limits fibrosis and scar formation. Several theories have been proposed to explain the anti-inflammatory effect of human amniotic membrane including down-regulation of transforming growth factor beta signaling, suppression of pro-inflammatory cytokines, and reduced polymorphonuclear cell infiltration and subsequent keratinocyte death \[[@rjw174C5]--[@rjw174C8]\]. Application of dehydrated human amniotic membrane appears a cost-effective strategy for treating wounds involving of the nose not involving the cartilage and perichondrium when compared to reconstruction with local or regional tissue transfer. Dehydrated human amniotic membrane grafts can be applied using local anesthesia and avoiding the costs and risks associated with intravenous sedation or general anesthesia. Application of dehydrated human amniotic membrane grafts are a simple single-stage procedure for treatment of the wound demonstrated in this case report. Pedicled tissue transfer by contrast is a minimum two-stage procedure and skin grafting entails donor site morbidity and possible complications. Dehydrated human amniotic membrane appears to be a simple, efficacious, cosmetically acceptable and cost-effective strategy for treating traumatic wound of the nose without cartilage or perichondrium involvement. Future studies are warranted to compare the dehydrated human amniotic membrane to local tissue transfer, skin grafting and healing by secondary intention. Study outcomes should consider costs, time to resolution and cosmesis. Conflict of Interest Statement {#rjw174s4} ============================== None declared.

Introduction {#S1} ============ Investigating the interactions between components as a network provides a common platform to uncover signatures of complex systems^[@R1]-[@R5]^. While this approach has been recently exploited to investigate biological systems of different scales, ranging from interactions between molecules to interactions between species, application of network theory at the level of individual cells has been rather limited. In this study, we present a network based approach to understand general principles in the organisation of cells in an organism (*i.e.*, epithelia). Early in animal development, cells in an epithelia begin to divide and alter their position, shape and size in stereotypical ways^[@R6]-[@R15]^. Despite being a highly dynamic process, this results in ordered, robust structure that ultimately leads to the formation of mature organs with cellular organisation suited to their specialised functions. Though we have an understanding of the contribution of genetic mechanisms (*e.g.*, external signals and the associated gene regulatory pathways^[@R6],\ [@R14]^) and cellular mechanics (*e.g.*, intrinsic patterns arising due to the rate of cell division^[@R8],\ [@R9],\ [@R11],\ [@R15]^ or cell re-arrangement due to anisotropy of cortical forces within individual cells^[@R6],\ [@R16]-[@R20]^) to the development of epithelial architecture in various model systems, we lack the means to objectively characterise and quantify the similarities and differences in the organisation of epithelia. Previous studies have offered insights into epithelial organisation by focusing primarily on geometric characteristics of individual cells such as the cell area and the number of contacts^[@R8],\ [@R9],\ [@R18],\ [@R21]^ and have led to the formulation of empirical relationships such as Aboav-Weaire's law and Lewis' law^[@R22],\ [@R23]^. This has largely emphasized the similarity between different epithelia. However, a more comprehensive view of epithelial organisation can be achieved if one considers the higher order organisation of cells such as the patterns in the network of interactions between cells that typically characterises an epithelium. The ability to do this would provide a way to describe objectively an epithelium, facilitate the investigation of fundamental questions about its organisation and dynamics, and establish an objective basis for comparative studies of epithelia from different sources. Importantly, the network characteristics of epithelial organisation (as opposed to geometric features) are not readily assessed by eye. This implies that higher-order organisation may not be accounted for in our current understanding of epithelial architecture. In this work we present an approach we term GNEO ([G]{.ul}eometric and [N]{.ul}etwork representation of [E]{.ul}pithelial [O]{.ul}rganisation) which by combining network and geometric measures of epithelial organisation, addresses these issues. We show that our approach is able to capture a defining signature that distinguishes epithelia from different organs, species, developmental stages and genetic conditions. In this way, GNEO permits characterization, quantification and classification of normal and perturbed epithelia in an objective manner. Results {#S2} ======= The GNEO method for characterising epithelial organisation {#S3} ---------------------------------------------------------- In order to capture information about the spatial organisation of cells and the global features of an epithelium, we generated network representations of confocal images of epithelia based on the cell-cell contacts. This allows principles from graph theory and complex networks^[@R2],\ [@R3],\ [@R5],\ [@R24]^ to be used to investigate short- and long-range patterns in epithelial organisation. In such a network, the centre of each cell is treated as a node and two nodes are linked if the two cells are neighbors (*i.e.*, physically contact each other) in the epithelium (Methods; [Fig. 1a](#F1){ref-type="fig"} and [Supplementary Fig. S1](#SD7){ref-type="supplementary-material"}). For each image, we generated a 'feature vector' consisting of eight features (Box 1): the means and standard deviations of the cell area, degree (*i.e.*, number of neighbors), clustering coefficient (the amount of interconnectedness among a cell's immediate neighbors) and average degree of neighbors (the average number of neighbors of a cell's neighbor). Thus, the mean values of the features in the feature vector represent information about the cell shape (area and degree) and the pattern of cell-to-cell contacts (degree, clustering coefficient and average degree of neighbors). While the degree is informative of the short-range pattern of contacts (the immediate neighbors of cell), the clustering coefficient and average degree of the neighbors represents the cell's context and surrounding, thus reflecting higher order organization. In turn, the standard deviation values are indicative of the cell-to-cell variability (*i.e.*, heterogeneity) of a feature in an epithelium. For each type of epithelium, we collected a set of images from several different individuals and extracted the feature vectors in each case ([Supplementary Table S1](#SD7){ref-type="supplementary-material"}). This allows us to compare epithelia, for example from different developmental stages, tissues and species ([Fig. 1b](#F1){ref-type="fig"} and [Supplementary Fig. S2](#SD7){ref-type="supplementary-material"}-[S3](#SD7){ref-type="supplementary-material"}), and the natural variation (*i.e.*, individual-to-individual variability) in epithelial organisation. To compare different epithelia, we used multivariate statistical methods to identify the contribution of the different feature in the feature vector that best separate different epithelial types. We took advantage of an unsupervised and a supervised method, namely Principal Component Analysis (PCA) and Discriminant Analysis (DA)^[@R4],\ [@R24],\ [@R25]^ (Box 1 and Methods). Both these methods provide information about the relative contribution of the features that distinguish different epithelia, termed 'feature weights' ([Fig. 2a](#F2){ref-type="fig"} and [Supplementary Table S2](#SD7){ref-type="supplementary-material"}). The statistical significance of the separation of the different epithelia was assessed using the MANOVA test (Methods). Comparison of epithelia from different organisms {#S4} ------------------------------------------------ To validate the approach, we first used the feature vector to compare visually distinct epithelia. For this we took advantage of the neural tube (samples cNT1 to cNT12) and embryonic ectoderm (cEE1 to cEE14) from chicken embryos, and the *Drosophila* wing imaginal disc from the prepupal stage (dWP1 to dWP16). Discriminant Analysis and PCA revealed that the epithelia from the two different organisms could be clustered into two distinct groups ([Fig. 2b](#F2){ref-type="fig"}; cNT and dWP; DA; *p*=9.17×10^−18^; [Supplementary Fig. S4](#SD7){ref-type="supplementary-material"}; cEE and dWP; DA; *p*=2.41×10^−21^, [Supplementary Fig. S5](#SD7){ref-type="supplementary-material"}) demonstrating the efficacy of the method. The greater importance of the average degree of the neighbours (N) and the standard deviation of the degree (D) in the dWP-cNT separation ([Fig. 2a,b](#F2){ref-type="fig"} and [Supplementary Table S1](#SD7){ref-type="supplementary-material"}) suggests that these two network features capture a certain defining signature that is independent of the cell area (which is comparable for these epithelial types). Comparison of different epithelial types {#S5} ---------------------------------------- To verify whether we could discriminate between different types of epithelia from the same organism, we compared the columnar neuroepithelium (cNT) of the chick to the squamous embryonic ectoderm (cEE). PCA and Discriminant Analysis of the feature vectors of these two epithelia revealed that they form two distinct groups ([Fig. 2c](#F2){ref-type="fig"}; DA; *p*=4.49×10^−16^; [Supplementary Fig. S6](#SD7){ref-type="supplementary-material"}; PCA; *p*=4.27×10^−10^). In this case, the cEE dataset was more spread out, reflecting the greater heterogeneity among these samples. However, each sample was clearly separated from cNT epithelia. In addition, the method separated the *Drosophila* wing pouch (dWP), the chick neuroepithelium (cNT) and embryonic ectoderm (cEE) epithelia into three groups ([Fig. 2d](#F2){ref-type="fig"}), demonstrating that it is sensitive to different types of epithelial organization. Multiple features contributed to the separation of these epithelia, suggesting that it was the combination of features that allowed the discrimination ([Fig. 2a](#F2){ref-type="fig"}). This underscores the importance of global structure of the network. In particular, we found that the standard deviation of the non-geometric (*i.e.*, network) features were more important in separating the different epithelial types, suggesting that, in this case, GNEO is able to capture patterns in epithelial organisation which are not visually apparent and are independent of cell area. Comparison of epithelia from different developmental stages {#S6} ----------------------------------------------------------- We next compared more closely related epithelia. Discriminant Analysis on the feature vectors of the wildtype (WT) epithelia of the wing pouch (which develops into the adult wing blade; dWP1 to dWP16) and the notum (which develops into the adult thorax; dNP1 to dNP12) of the wing imaginal disc from the prepupal stage showed that these samples were only partially separable. This indicates a similarity in the overall organisation of both epithelia at this stage during development ([Fig. 2e](#F2){ref-type="fig"}; DA; *p*=0.001), consistent with these regions comprising different areas of the same epithelial sheet. Moreover, a comparison of the prepupal wing pouch (dWP1 to dWP16) with the third instar larva wing pouch (dWL1 to dWL15) produced a discriminant graph with two groups, which also overlapped but were better separated ([Fig. 2f](#F2){ref-type="fig"}; DA; *p*=3.45×10^−5^). Together these data suggest that the global organization of the wing epithelia change gradually during development and across an epithelial sheet. The GNEO approach provides a way to assess the relatedness of different epithelia through approaches that calculate distances between data points using standard approaches such as hierarchical clustering. Consistent with our observation, a comparison of four epithelial types (Discriminant Analysis and PCA of the *Drosophila* dWP and dNP and the chick cEE and cNT) showed that dWP and dNP forms an overlapping cluster whereas the cEE and cNT form separate clusters ([Fig. 2g](#F2){ref-type="fig"}). These observations confirm that the epithelial organisation of the prepupal notum and wing are similar whereas the neural tube and embryonic ectoderm are not. Comparison of epithelia from different tissues {#S7} ---------------------------------------------- While the wing pouch and the notum are closely related, we tested if our approach can separate distantly related epithelia in *Drosophila*, by comparing the third instar larval wing disc epithelia (dWL) with the eye epithelia (dEL1 to dEL5). Unlike the wing disc epithelium, the third instar eye disc contains a gradient of apical constriction induced by myosin II activation^[@R26],\ [@R27]^, leading to apparently different cell shapes and sizes across the epithelial sheet. Discriminant Analysis of the feature vectors clearly demonstrated that the eye and wing disc epithelia are significantly different and that most of the features, except the average clustering coefficient, contribute to this separation ([Supplementary Fig. S8](#SD7){ref-type="supplementary-material"}; DA; *p*=1.11×10^−5^). Natural variation in epithelial organisation {#S8} -------------------------------------------- The observation that the network features of the feature vectors were important in discriminating different types of epithelial tissues raised the possibility that local cell packing (resulting from cell shape and distribution) gives rise to the characteristic long-range patterns of an epithelium. In this view, the global organization is a property of the epithelium that emerges from the collective behaviour of the cells. Consistent with this idea, the individual-to-individual variation (measured as coefficient of variation (C.V.), see [Table 1](#T1){ref-type="table"} and Methods) of mean values of the network features between the different individuals is several fold smaller than that of cell area ([Table 1](#T1){ref-type="table"}). In addition, the individual-to-individual variation (C.V.) of the mean network features was much smaller than the C.V. of the cell-to-cell variability (*i.e.*, the standard deviation of the features in the feature vector) across individuals. These observations suggest that there is a reproducible long-range epithelial structure, which is, to a large extent, independent of variations in cell size. Comparison of wild-type and genetically perturbed epithelia {#S9} ----------------------------------------------------------- What regulates this reproducible long-range organization of epithelia? Several factors have been implicated in controlling the behaviour of individual cells and consequently epithelial architecture^[@R6],\ [@R8],\ [@R9],\ [@R11],\ [@R15]-[@R18]^. However, how global epithelial structure is determined by, for example the effect of the cytoskeleton of the cells within the epithelium is not understood. Therefore, we applied GNEO to quantify objectively the effect in the wing disc of removing myosin II heavy chain using RNAi^[@R28],\ [@R29]^, a genetic manipulation that robustly and uniformly disrupts the cytoskeletal organization and epithelial architecture (Methods). Both PCA and DA were clearly able to separate wildtype (WT) discs from those in which myosin II had been reduced ([Fig. 3a](#F3){ref-type="fig"} and [Supplementary Fig. S9a](#SD7){ref-type="supplementary-material"}). Interestingly, while all WT wings formed one distinct tight cluster, the mutant wings were more broadly spread over the graph ([Fig. 2a](#F2){ref-type="fig"}; PCA; *p*=4.27×10^−10^). This is consistent with a visual inspection of the data, which showed that reducing the levels of myosin II by RNAi knockdown disrupted epithelial organisation to different extents in different wing discs. In order to provide an objective score for the severity of the mutant phenotype, we calculated the Euclidean distance between each mutant wing and the center of mass of the WT wings ([Supplementary Table S3](#SD7){ref-type="supplementary-material"}). The coefficient of variation of the distances was \~26%, which most likely reflects the variability of the RNAi efficiency among the individuals. Individual mutant samples were between 15 and 35 times further from the center of mass than the average of the distances of the dWP samples ([Supplementary Table S3](#SD7){ref-type="supplementary-material"}). We then investigated how the inclusion of other epithelia during the comparison affected our analysis of myosin II reduction. Addition of the prepupal notum (dNP) showed that both wildtype epithelia cluster together whereas the mutant wing pouch epithelia were still scattered ([Fig. 3b](#F3){ref-type="fig"}; PCA; *p*=1.57×10^−13^ and [Supplementary Fig. S9b](#SD7){ref-type="supplementary-material"}; DA; *p*=1.89×10^−18^). Thus GNEO can objectively recognise that the *Drosophila* WT samples are similar to each other but distinct from the mutant ones. A PCA that includes the chick embryonic ectoderm and the neural tube, either together or alone, revealed that each group forms distinct clusters (PCA: [Fig. 3c](#F3){ref-type="fig"}; *p*=2.98×10^−31^; [Fig. 3d](#F3){ref-type="fig"}; p=3.40×10^−22^ and DA: [Supplementary Fig. S9c](#SD7){ref-type="supplementary-material"}; *p*=2.03×10^−44^; [Supplementary Fig. S9d](#SD7){ref-type="supplementary-material"}; *p*=1.36×10^−34^). This suggests that the *Drosophila* mutant epithelia are clearly distinct from the WT *Drosophila* and chick epithelia. Discriminant Analysis indicated that the area, the degree, the degree of neighbours and the standard deviation of the degree are the most important features ([Fig. 2a](#F2){ref-type="fig"}) for obtaining this separation ([Supplementary Fig. S9a-d](#SD7){ref-type="supplementary-material"}). This is consistent with an effect of myosin II on regulating cell shape^[@R26],\ [@R27],\ [@R30],\ [@R31]^, whose levels when altered affects cell area, the number of neighbours and the "regularity" of the pattern of cell contacts. The increased variability of the network features when myosin II is knocked down demonstrates an important biological conclusion of this work: long-range constancy of epithelial packing is regulated by cytoskeletal organisation within individual cells. This is also reflected by the higher variation between discs from different individuals in the myosin II knock-down compared to WT discs ([Table 1](#T1){ref-type="table"}). Together, these data suggest that the long range organization of an epithelium is determined, at least in part, by the cytoskeleton of the cells comprising the tissue affecting interactions across the field of cells. Discussion {#S10} ========== The network-based approach, GNEO, introduced here, captures epithelial organisation by accounting for patterns of cell-contacts which cannot be quantified either by visual inspection or by using geometric features that describe individual cells alone. In particular, GNEO can objectively quantify differences between epithelia from different tissues and organisms, even when the size and shape of the cells comprising these epithelial appear visually indistinguishable. First, we show that epithelia from different organs and species have distinct, reproducible and quantifiable differences in their structure. Second, a surprising result is that differences in cell area play a relatively minor role in distinguishing wing disc and neural tube epithelia -- two epithelia that come from different species which produce very different tissues. This indicates that there is unexpected consistency in both the average size and range of sizes of cells across epithelia from different species. Third, we provide evidence that the non-geometric features of the epithelia are most informative in distinguishing them. This shows that the topological organization of the epithelium differs strongly between tissues and between species. Together with the finding that the structure of the same epithelia from different individuals is highly reproducible, our analysis indicates an unexpected level of genetic control over the long range organization of epithelia. To the best of our knowledge this has not been previously reported. Lewis's and Aboav-Weaire's laws^[@R23]^ define relationships between area, degree and neighbour degree, and place the emphasis on the universal connection between these features. By contrast our work examines how measurements of these features across a population of cells can be used to build a quantitative and objective description of an epithelium. The experiment in which we disrupt myosin II provides an indication of the mechanism by which population level features of an epithelium emerge from the collective behaviour of individual cells. Moreover, we provide evidence that GNEO can operate as a reliable classifier to differentiate mutant and WT epithelia, and to quantify precisely the severity of mutant phenotypes. While there are other ways of constructing networks from images of epithelia, this representation of cell-to-cell contacts offers a simple and readily applicable method to analyze epithelia objectively. Representing epithelia as feature vectors opens up the possibility of applying artificial intelligence (*e.g.*, pattern recognition) algorithms to classify them in an objective manner and can be extended to include more sophisticated network features. Since the method is automatable, adaptations of this approach can be used in high-throughput experiments aimed at identifying pathways and quantifying the effects of mutations in functional genomics screens. Of particular value, GNEO allows characterisation of subtle phenotypes undetectable by visual inspection. This approach can also be adapted to other biological samples such as nerve-cell connections, muscle cells attachments, and tumours. Methods {#S11} ======= Genetic strains and confocal imaging of the epithelia {#S12} ----------------------------------------------------- Flies were grown by employing standard culture techniques. The following lines were used: Arm-GFP (WT), *C765-Gal4* (<http://flybase.bio.indiana.edu>) and *U-zip-RNAi* (Vienna *Drosophila* RNAi Center collection). Imaginal discs from the prepupa and third instar larvae were dissected in PBS and fixed with 4% paraformaldehyde in PBS for 35 min. The samples were washed six times for 10 min with PBT (PBS, 0.3% triton) and 3 times for 5 min with PBS. Imaginal discs were mounted using Fluoromount-G (Southern Biotech). Images were taken with a BioRad Radiance 2100 laser scanning confocal microscope. All the images were captured using 63× immersions objective with 3 times zoom and exported as a 1024 × 1024 pixel TIFF file. The area of 1 pixel is 3.78×10^−3^ μm^2^. The regions of the imaginal discs that appear in the images were selected with the following criteria: For the prepupal and larval wing disc, images from dorsal compartments (leaving out the D/V boundary region) of the wing pouch region were obtained. The notum images were taken from the anterior part of the disc ([Supplementary Fig. S2](#SD7){ref-type="supplementary-material"}). For the eye disc, the images were taken locating the morphogenetic furrow at the side with an additional margin of three to five rows of cells (to include the first and second rows of clusters of photoreceptors). For the chicken images, Hamburger and Hamilton (HH) ^[@R32]^ stage 10 and 17 chick embryos ([Supplementary Fig. S3](#SD7){ref-type="supplementary-material"}), were fixed for 1 hour in 4% paraformaldehyde. HH st. 17 embryos were subsequently transferred to methanol then rehydrated. Immunostaining was performed with a ZO1 antibdoy (Zymed labs) and embryos were flat mounted. Images of neural tube epithelium, at intermediate dorsoventral positions, were obtained at the level of somite 5 of HH st. 17 embryos. Embryonic ectoderm was imaged adjacent to the most recently formed somite of HH st. 10 embryos. Imaging conditions were as for imaginal discs, image orientation: anterior, left; dorsal, up. Image processing and generation of the epithelial network {#S13} --------------------------------------------------------- The acquired images were converted into their respective 2-pixel BMP files. First, the confocal images were imported using Adobe Photoshop CS2. Colours were inverted to obtain a dark signal over a light background. Epithelial cells were identified from the processed images by using a semi-automated framework. The images first had their illumination corrected through polynomial fitting followed by thresholding, and were then manually checked and edited to remove artefactual connections between cells. The cells and their boundaries were defined using a foreground tracking program^[@R4]^. Adjacent cells were identified by transforming their images into an epithelial network where the centroid of each cell is represented by a node and the neighborhood relations are mapped into weighted edges ([Supplementary Fig. S1](#SD7){ref-type="supplementary-material"}). The methodology involves the following steps: (a) all cells have their borders detected^[@R4]^, (b) for each pixel *p* of the border of each cell *i*, all border pixels belonging to other cells and falling within the circle of radius *r* centered at *p* are identified and counted (we use *r*=6) (c) the node corresponding to this cell is connected to other cells by edges whose weights, *w*, correspond to the total number of neighboring pixels found during step (b). We established empirically that if the weight is greater than 40% of the minimum equivalent radius of the areas of each adjacent pair of cells, those cells can be considered to be neighbours. The equivalent radius of a cell with area *A* is defined as corresponding to $\sqrt[{}]{\left. A\slash\pi \right.}$. We selected an area within every image in order to have all the networks presented with similar boundary conditions. The squares were drawn to obtain the maximum possible surface without including the centroids of the cells in the border of the images. The features of the cells falling within this area were calculated. The cells outside were only used in order to provide neighbors to the cells analyzed in the network. Calculation of geometric and network features {#S14} --------------------------------------------- The area was measured by counting the number of pixels inside each cell^[@R4]^. Each epithelial image was represented as a network where each node corresponds to one of the cells and the links between nodes reflect the spatial adjacency between the epithelial cells ([Fig. 1a](#F1){ref-type="fig"}). Several measurements can be estimated from these networks^[@R24]^ in order to provide useful characterisation and respective biological interpretations. Let the graph be represented in terms of its adjacency matrix *K*, such that the presence of a connection between nodes *i* and *j*, with 1 ≤ *i*, *j* ≤ *N* , implies *K(i,j)* = *K(j,i)* = 1, with *K(i,j)* = *K(j,i)* = 0 being enforced otherwise. In this work, we employed the following topological characteristics: Degree of a node *i*, which corresponds to the number of edges attached to it, *i.e.*, $k\left( i \right) = \sum\limits_{v = 1}^{N}K\left( v,i \right)$. Clustering coefficient of a node *i*, which is obtained by dividing the number of edges between the neighbours of *i*, represented as *n*(*i*), by the maximum possible number of connections between those nodes. This measurement can be calculated as $c\left( i \right) = 2\frac{n\left( i \right)}{k\left( i \right)\left( k\left( i \right) - 1 \right)}$. Therefore, the clustering coefficient varies from 0 (no interconnections between the neighbours of *i*) to 1 (the neighbours are fully interconnected). Average degree of the neighbours of a node, calculated as the average of the number of edges that are attached to the neighbours of node *i*. This measurement can be calculated as $b\left( i \right) = \frac{1}{k\left( i \right)}\sum\limits_{v = \mathit{\text{neighbors\ of\ i}}}k\left( v \right)$. For every feature, both the average and standard deviation values across all cells in an epithelium were estimated and were used to characterise epithelial organisation. Cell elongation was initially considered as an independent parameter but was found not to contribute further to the classification and separation. We believe that this is because variations in the elongation tend to affect the degree, clustering coefficient and average degree of neighbours, therefore becoming correlated with those measurements. Feature vector and multivariate statistical analysis {#S15} ---------------------------------------------------- The geometric and network analyses of epithelial images yield a large number of features or measurements ([Supplementary Table S1](#SD7){ref-type="supplementary-material"}). More precisely, a total of 8 features, corresponding to the mean and standard deviation of the area of cells, as well as the degree, clustering coefficient and average degree of a node are obtained. Thus, for every image of an epithelium, a feature vector of eight dimensions was obtained. We apply an unsupervised and supervised multivariate statistical method, namely Principal Component Analysis (PCA) and Discriminant Analysis (DA)^[@R4],\ [@R24],\ [@R25]^ (see [Supplementary Methods](#SD7){ref-type="supplementary-material"} for explanation). Standardisation^[@R4]^ of the measurements is performed in order to eliminate the effect of the magnitude of the measurements on the respective separation between categories. More specifically, the average and standard deviation for every feature across all individuals in each of the epithelial type was calculated. For every feature in a feature vector representing an individual epithelium, the calculated average value was subtracted and divided by the standard deviation. Thus, the components of the eigenvectors associated to the largest eigenvalues provide a quantification of the degree of contribution of each original measurement in maximising the dispersion of the projection (in the case of PCA) and optimising the separation between the categories (in the case of discriminant analysis). Since both these methods output a weighted linear combination of all the features in the feature vector to obtain the final projection (the new axes), the loading values associated to each feature provide a direct measure of the contribution of that particular feature towards the projection onto the smaller-dimensional space. For our analysis, the maximum between the magnitudes of the values of the components of each feature (*i.e.*, measurement) of the first two eigenvector (*i.e.*, the loadings associated with the first two component axes) was defined as the weight of that respective measurement. Estimation of statistical significance {#S16} -------------------------------------- The MANOVA (multivariate analysis of variance) test, which is a reference statistical test for probing the hypotheses that two or more populations, characterized in terms of two or more dependent variables, are or not distinct, was used to assess the statistical significance of the obtained separation. The tested (null) hypothesis H0 is that the two samples come from the same population, with H1 indicating different populations. The p-values are calculated in the standard way^[@R33],\ [@R34]^, after PCA or DA. Both PCA and DA provide new random variables that are linear combinations of the previous ones. In the case of PCA, the new variables are completely uncorrelated. In the case of the DA, the method does take into account the known categories of the cells and therefore enhances the separation between the categories. In this case, our interest was focused on the contribution of the measurements on the separation, not on the separation itself. The p-values in this case reflect the effect of the informed cell categories and should be treated as such. The assumptions required for MANOVA were verified even though after application of PCA and DA, as the respective clusters remained largely normal and with similar variances. Supplementary Material {#SM} ====================== The authors acknowledge the MRC for funding and would like to thank A Baonza, A Wuster, A Pombo, C Chothia, E Levy, F Velazquez, G Chalancon, J Casanova, J Gsponer, J Modolell, P Cicuta, S Balaji, S Munro, S Teichmann and T Lecuit for providing helpful comments on this work. MMB acknowledges Darwin College, EMBO YIP and Schlumberger Ltd for support. LME is funded by the Marie Curie and the EMBO fellowships. LdFC is grateful to FAPESP (05/00587-5), CNPq (301303/06-1) for financial support. Part of this work was performed during a Visiting Scholarship to LDFC from St. Catharine's College, University of Cambridge. JB is supported by the MRC (UK) and AK by a FEBS fellowship. **Author contributions** MMB and LME designed the study with help from LdFC and MF. LME and AK obtained the fly and chicken images, respectively. LME processed the images. LdFC wrote the software, generated the epithelial network and performed all the comparisons and statistical analysis. All authors participated in the interpretation of results, discussions and the development of the project. LME and MB wrote the manuscript with input from all authors. **Supplementary Information** accompanies this paper at <http://www.nature.com/naturecommunications>. **Competing financial interests:** The authors declare no competing financial interests. ###### Qualitative definition of the geometric and network features and the statistical approaches used {#F1} {#F2} {#F3} -------------------------------------------------------------------------------------------- Coefficient\ average standard deviation of Variation\ x100 --------------- --------- -------------------- ------ ------ ------- ------- ------- ------- 36.21 0.55 0.42 0.64 37.59 6.49 7.28 4.90 40.25 0.47 0.86 0.96 34.75 4.19 6.86 7.01 27.83 0.34 0.50 0.41 33.62 4.15 5.20 3.60 30.81 1.94 2.84 2.33 51.69 11.66 15.75 10.78 18.59 0.64 2.77 1.53 24.30 8.93 12.70 10.74 32.57 3.11 8.99 6.20 31.42 16.41 14.69 17.88 -------------------------------------------------------------------------------------------- Area (A) Degree of a node (D) Clustering coefficient (C) Average degree of neighbours (N) [^1]: Current address: Instituto Biomedicina Sevilla (IBiS), Universidad de Sevilla/ CSIC/ Hospital Virgen del Rocío. Seville, Spain.

Introduction ============ The efficient and refined division of labor among social insects, such as termites, wasps, ants, and bees, plays a key role in their ecological success ([@B40]). A normal honeybee swarm is usually composed of three types of individuals, namely diploid females that include a queen and worker bees and haploid males (drones). Marked differences exist between the queen and worker bees in terms of morphology, behavior, reproductive ability, function, and life span ([@B39]). This phenomenon in social insects is known as caste determination and is affected by environmental factors ([@B37]; [@B13]). Although both the queen and worker bees develop from fertilized eggs, the larvae gradually develop to a queen and worker bees, depending on the environment ([@B10]). Plasticity is one of the key characteristics of the division of labor in social insects ([@B31]), and it regulates the process of caste determination. Caste determination is not usually inherited but is mediated by external factors. Fertilized eggs and larvae selectively develop into a queen and worker bees depending on local nutrition and environmental factors. If worker larvae of *Apis mellifera* are transplanted within 3 days of age to the queen cells of a bee colony and incubated, they can grow and develop into queen bees with mature ovaries ([@B20]). In China, a standard practice in commercial beekeeping is to raise queens by transplanting eggs or young larvae into artificial queen cells ([@B11]; [@B7]), triggering the development of a queen that produces royal jelly ([@B42]). In commercial queen rearing practices, there is variation in the age at which worker larvae are transplanted to queen cells to be raised as queens. Some reports have shown that queens reared from older worker larvae have decreased body sizes, a smaller spermatheca, and fewer ovarioles compared with those of the queens reared from younger worker larvae ([@B29]; [@B27]). Therefore, the queens reared from older worker larvae have decreased reproductive ability, and the colony produces a significantly smaller worker comb and drone comb and has lower stored food. The quality of the queen is the most important economic trait of a colony. Although it is well known that environmental factors affect the quality of a queen, the underlying regulatory mechanism is unclear. In this study, the transcriptomes of 4-day-old larvae transferred at different ages to the queen cell were analyzed by RNA-sequencing (RNA-Seq) to monitor changes in the expression profile of larvae that developed into queens possessing different qualities. The differential regulation of development led to caste determination, eventually leading to reproductive division of labor; thus, our study provides valuable information on the regulatory mechanisms of environmental factors affecting the quality of a queen and amplifying the understanding of caste differentiation in bees. Materials and Methods {#s1} ===================== Honeybee Larvae and Experimental Procedures ------------------------------------------- To minimize noise in the genetic background, *A. mellifera carnica* bees were derived from a single drone-inseminated queen. The colonies were raised at the Honeybee Research Institute of Yangzhou University in Yangzhou, China. The queens were confined for 6 h in a blanket honeycomb to lay eggs. Worker larvae aged 1, 2, and 3 days that developed from the eggs, which the queen laid in worker cells, were transplanted to the queen cells and reared to the age of 4 days; the three groups were named L1, L2, and L3, respectively. Each group had three biological replicates, and each replicate included three larvae. The experimental design is shown in Figure [1](#F1){ref-type="fig"}. All samples were immediately flash-frozen in liquid nitrogen. RNA from each sample was extracted and subjected to RNA-Seq analysis at Shanghai OE Biotech Company. {#F1} RNA Isolation, Library Preparation, and Sequencing -------------------------------------------------- Total RNA was extracted from three larvae using the mirVana^TM^ miRNA Isolation Kit (Ambion-1561) and was pooled by following the protocol of the manufacturer. The integrity of RNA was evaluated using the Bioanalyzer 2100 RNA-6000 Nano Kit (Agilent Technologies). Samples with an RNA integrity number (RIN) ≥ 7 were analyzed further. Complementary DNA (cDNA) libraries were constructed from mRNA that was isolated using the TruSeq Stranded Total RNA Sample Prep Kit, based on enrichment with oligo (dT) magnetic beads and fragmentation (approximately 200--700 nucleotides) in fragmentation buffer. Briefly, random hexamer primers were used to amplify and prepare the library. Double-stranded cDNA products were purified using the QiaQuick PCR Extraction Kit (Qiagen) and eluted in EB buffer for end repair and poly (A) addition. These libraries were then sequenced on a HiSeqTM 2500 instrument (Illumina), and 150-bp paired-end reads were generated. Mapping Reads to the Predicted Coding Sequence (CDS) of Genes in the Reference Genome ------------------------------------------------------------------------------------- The raw reads of all nine samples studied were processed using the Trimmomatic software ([@B23]). Reads containing poly-N regions and low-quality reads (defined as 50% of bases in a read with a quality value ≤ 5) were removed to obtain clean reads. The clean reads were then mapped to the predicted coding sequences (CDSs) of the corresponding genes of the reference *A. mellifera* genome, using hisat2 ([@B30]). Expression Annotation and DEG Analysis -------------------------------------- The fragments per kilobase million (FPKM) value of each transcript (protein_coding) was calculated using bowtie2 ([@B33]) and eXpress ([@B1]). Differentially expressed genes (DEGs) were identified using the DESeq functions estimateSizeFactors and nbinomTest ([@B15]). A *P*-value \< 0.05 and a fold change \> 2 or fold change \< 0.5 was set as the threshold for significant differential expression. Hierarchical cluster analysis of the DEGs was performed to explore the expression patterns of the transcripts. Functional Analysis of the DEGs ------------------------------- All identified DEGs were matched to the gene ontology (GO) terms ([@B2]) using BLAST2GO ([@B9]) by following the standard procedure to perform the Basic Local Alignment Search Tool (BLAST) searches for each gene (BLASTn, DFCI database). Enrichment analysis ([@B21]) revealed whether the DEGs have related functions. Validation of the RNA-Seq Data by Quantitative Real-Time PCR (qRT-PCR) ---------------------------------------------------------------------- A total of seven DEGs were randomly selected and examined in qRT-PCR experiments performed with three biological replicates. The reactions were performed using the ABI 7500 system with SYBR Green. β-actin (AB023025), a reference gene, was used as the internal control ([@B24]). The qRT-PCR data were expressed relative to the expression of β-actin. The sequences of the gene-specific primers are shown in Supplementary File [S1](#FL1){ref-type="supplementary-material"}. Results ======= RNA-Seq and Analysis of the Raw Data ------------------------------------ After filtering the adaptor sequences (the regions containing poly-N and low-quality sequences), over 86.36 million clean reads were produced in each library. An overview of the sequencing statistics is displayed in Table [1](#T1){ref-type="table"}. The percent of clean reads among the raw reads in each library ranged from 96.19 to 97.05%, and Q30 (percent of bases with a Phred score \> 30 among the raw bases) was more than 93.34% (Table [1](#T1){ref-type="table"}), suggesting that the high-quality RNA-Seq data obtained could be used for further analysis. An average of 68.41 million reads per sample was mapped to the predicted CDS of the corresponding genes in the honeybee (*A. mellifera*) genome (64.93--71.66 million). Of the total reads, the rate of reads that matched was \> 75% (Table [2](#T2){ref-type="table"}). The correlation value of the three biological replicates for each sample was \> 0.99 (*R*^2^ \> 0.99) based on the values (Supplementary Figure [S1](#FS1){ref-type="supplementary-material"}). The reported sequencing data has been approved and assigned to the Sequence Read Archive (SRA) database (SRA accession number: [SRP158315](SRP158315)). ###### Throughput and quality of RNA sequencing data. Sample Raw reads Clean reads Percent of clean reads GC percent (%) Q30 percent (%) ------------- ----------- ------------- ------------------------ ---------------- ----------------- Sample_L1_1 89589486 86769356 0.968521641 41.50% 93.94% Sample_L1_2 86459394 83172042 0.961978082 41.50% 93.34% Sample_L1_3 87437832 84118264 0.962035106 41.00% 93.39% Sample_L2_1 89173290 86269582 0.967437469 40.00% 94.07% Sample_L2_2 87318644 84739966 0.970468185 41.00% 94.36% Sample_L2_3 86361798 83220106 0.963621739 40.00% 93.68% Sample_L3_1 87818532 84839102 0.966072879 41.00% 93.93% Sample_L3_2 89477462 86620466 0.968070216 41.00% 94.13% Sample_L3_3 89125256 86298048 0.968278262 41.00% 94.17% ###### Summary of reads mapped to the reference genome of *Apis mellifera.* Sample Sample_L1_2 Sample_L1_1 Sample_L1_3 Sample_L2_2 Sample_L2_1 Sample_L3_3 Sample_L2_3 Sample_L3_1 Sample_L3_2 ------------------------------ ------------------- ------------------- ------------------- ------------------- ------------------- ------------------- ------------------- ------------------- ------------------- Total reads 83172042 86769356 84118264 84739966 86269582 86298048 83220106 84839102 86620466 Total mapped reads 69676988 (83.77%) 70823525 (81.62%) 67621141 (80.39%) 68114357 (80.38%) 71661953 (83.07%) 68670887 (79.57%) 65399143 (78.59%) 64927181 (76.53%) 68870251 (79.51%) Multiple mapped 880546 (1.06%) 902330 (1.04%) 848781 (1.01%) 883477 (1.04%) 885178 (1.03%) 909010 (1.05%) 808646 (0.97%) 825601 (0.97%) 914938 (1.06%) Uniquely mapped 68796442 (82.72%) 69921195 (80.58%) 66772360 (79.38%) 67230880 (79.34%) 70776775 (82.04%) 67761877 (78.52%) 64590497 (77.61%) 64101580 (75.56%) 67955313 (78.45%) Read-1 34329259 (41.27%) 34873412 (40.19%) 33331119 (39.62%) 33630071 (39.69%) 35429434 (41.07%) 33918281 (39.30%) 32323040 (38.84%) 32070186 (37.80%) 34004794 (39.26%) Read-2 34467183 (41.44%) 35047783 (40.39%) 33441241 (39.76%) 33600809 (39.65%) 35347341 (40.97%) 33843596 (39.22%) 32267457 (38.77%) 32031394 (37.76%) 33950519 (39.19%) Reads map to '+' 34430320 (41.40%) 34985553 (40.32%) 33391713 (39.70%) 33649542 (39.71%) 35398775 (41.03%) 33894805 (39.28%) 32297706 (38.81%) 32023469 (37.75%) 33984253 (39.23%) Reads map to '-' 34366122 (41.32%) 34935642 (40.26%) 33380647 (39.68%) 33581338 (39.63%) 35378000 (41.01%) 33867072 (39.24%) 32292791 (38.80%) 32078111 (37.81%) 33971060 (39.22%) Non-splice reads 47803802 (57.48%) 48628225 (56.04%) 45897286 (54.56%) 46439326 (54.80%) 48714867 (56.47%) 46375151 (53.74%) 44519421 (53.50%) 44059225 (51.93%) 46795353 (54.02%) Splice reads 20992640 (25.24%) 21292970 (24.54%) 20875074 (24.82%) 20791554 (24.54%) 22061908 (25.57%) 21386726 (24.78%) 20071076 (24.12%) 20042355 (23.62%) 21159960 (24.43%) Reads mapped in proper pairs 65372332 (78.60%) 66310718 (76.42%) 63153412 (75.08%) 63709484 (75.18%) 66844568 (77.48%) 64018970 (74.18%) 60946450 (73.24%) 60559126 (71.38%) 63793796 (73.65%) Principal component analysis was performed, and the results showed that the samples clustered into three groups (Figure [2](#F2){ref-type="fig"}). These results indicated sufficient reproducibility and rationality of sampling. {#F2} DEGs in Worker Larvae Transplanted at Different Ages ---------------------------------------------------- The RNA-Seq analysis was performed to compare the gene expression levels between the three groups (L1, L2, and L3). The results showed that the number of DEGs in L1 vs. L3 (1798) was significantly higher than that in L1 vs. L2 (1022). Compared with those in L1, more than 60% of the genes in L2 and L3 were downregulated, and the number of downregulated genes increased with the age of the transplanted worker larva (Figure [3](#F3){ref-type="fig"}). We observed 578 genes that were either down or upregulated in L2 and L3 of which 413 genes were downregulated and only 165 were upregulated (Figure [4](#F4){ref-type="fig"}). This finding implied that these shared downregulated genes likely play pivotal roles during larval developmental after transplantation. Among the downregulated genes, a high proportion of genes was involved in metabolism, body development, reproductive ability, and longevity (Supplementary File [S2](#FL2){ref-type="supplementary-material"}). {#F3} {#F4} Functional Annotation and Classification ---------------------------------------- The DEGs within the L1 vs. L2 and L1 vs. L3 groups (*p* \< 0.05) were assigned GO terms related to their cellular components, molecular functions, and biological processes. The 2741 and 3872 GO terms were displayed in the Supplementary File [S3](#FL3){ref-type="supplementary-material"}. Furthermore, GO top-30 enrichment analysis (top 10 enriched genes in terms of molecular function, biological process, and cellular component categories) (Figures [5](#F5){ref-type="fig"}, [6](#F6){ref-type="fig"}) revealed that 24 of the top-30 enriched terms were same between the two groups (Figure [6](#F6){ref-type="fig"}). The results of the GO enrichment analysis showed a similar pattern between L1 vs. L2 and L1 vs. L3. {#F5} {#F6} Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathway Enrichment Analysis of DEGs ---------------------------------------------------------------------------------- The KEGG pathway enrichment analysis was conducted for the DEGs within the L1 vs. L2 and L1 vs. L3 groups. A total of 171 DEGs within the L1 vs. L2 group were enriched for 102 KEGG pathways, whereas 363 DEGs within the L1 vs. L3 group were enriched for 112 pathways. The DEGs within the L1 vs. L2 and L1 vs. L3 groups were enriched for 95 of the same pathways (Supplementary File [S4](#FL4){ref-type="supplementary-material"}) of which 52 pathways were related to synthesis and metabolism, 16 to genetic information processing, 13 to embryonic development, 12 to cellular processes, and 2 to human diseases (Supplementary File [S4](#FL4){ref-type="supplementary-material"}). Among these pathways, several are thought to be related to caste differentiation ([@B4]; [@B38]; [@B3]; [@B12]; [@B14]), including the MAPK signaling pathway (fly), FoxO signaling pathway, mTOR signaling pathway, longevity regulating pathway (multiple species), Wnt signaling pathway, dorsoventral axis formation, Hedgehog signaling pathway (fly), TGF-beta signaling pathway, Hippo signaling pathway (fly), Hippo signaling pathway (multiple species), Toll and Imd signaling pathway, and insect hormone biosynthesis. The number of DEGs (especially downregulated genes) in these pathways increased with the age of the transplanted worker larva (Figure [7](#F7){ref-type="fig"}). Further analysis of the genes of several biological pathways showed that some genes are simultaneously involved in multiple biological pathways (Figure [8](#F8){ref-type="fig"}). Among these genes, transcriptional regulator Myc-B (XM_016914906.1), casein kinase 1 (CK 1)-like (XM_016914939. 1), and S-phase kinase-associated protein 2 (XM_006557702.2) were either upregulated or downregulated in the L2 and L3 groups compared with the L1 group. In particular, CK 1-like participates simultaneously in the Wnt, FoxO, Hedgehog, Hippo, TGF-beta, and longevity regulating pathways and was upregulated in the L2 and L3 groups compared with expression in the L1 group. The transcriptional regulator Myc-B participates in the Wnt, Hippo, and TGF-beta pathways, where S-phase kinase-associated protein 2 participates in FoxO and mTOR pathways simultaneously; both DEGs were downregulated in the L2 and L3 groups when compared with expression in the L1 group. {#F7} {#F8} Confirmation of the RNA-Seq Data by RT-qPCR ------------------------------------------- To validate the accuracy and reproducibility of the transcriptome results, seven DEGs were randomly selected for RT-qPCR verification. Total RNA samples isolated from the L1, L2, and L3 groups were used as templates. The results showed that the expression patterns of candidate genes were consistent with the RNA-Seq data (Figure [9](#F9){ref-type="fig"}), which confirmed that the measured changes in gene expression detected by RNA-Seq indeed reflected transcriptome differences between the different libraries. {#F9} Discussion ========== Honeybee is a major model organism among eusocial insects. Theoretically, its complex social behavior can be interpreted to reflect changes in gene expression. The caste differentiation phenomenon of female bees can be traced to DEGs, which determine the developmental fate of a larva. The worker larvae are usually moved to an artificial queen cell to produce royal jelly. Furthermore, it has been found that worker larvae that are less than 3 days old can develop into queens; however, the transplanted older worker larvae develop into individuals of smaller size and lower weight when compared with those of the adult queens ([@B36]; [@B41]; [@B17]). Worker larvae that are more than 3 days old when transplanted fail to develop into queens ([@B36]). In this study, we explored genetic changes in 4-day-old worker larvae transplanted into queen cells at the age of 1, 2, and 3 days, which could affect the fate of the worker larvae. The results showed that the downregulated DEGs are mostly involved in metabolism, body development, reproductive ability, and longevity, indicating that these processes in a queen may decrease when larvae are transplanted at increasing ages. These functions may be critical for the ability of worker larvae to develop into queens, and when these functions in larvae decreased, the quality of queens decreased. Previous reports showed that the queen quality and colony productivity decreased depending on the queen rearing method ([@B41]; [@B34]; [@B29]). This is likely to be caused by a poor-quality queen, which in turn affects the quality of the swarm. In practice, older worker larvae are chosen for transplantation as they are easier to handle and have a high success rate ([@B8]); however, this results in the poor quality of the queen and colony. Therefore, it is preferable to rear queens from younger larvae to achieve better outcomes in terms of queen performance and colony function. Approximately 400 years ago, people realized that the differentiation into queen and worker bees was due to the different food fed to larvae. Later studies showed that this is due to the differences in both the quantity and quality of food ([@B25]), among which the royal jelly plays a key role in caste differentiation. Throughout the larval period, the queen larvae consume substantial amounts of royal jelly, but worker larvae receive plenty of royal jelly only for the first 3 days, after which the food of the worker larvae is switched to a mixture of royal jelly, honey, and pollen. Studies have shown that there are a few differences in the main ingredients of the royal jelly provided to the queen larvae and worker larvae during the first 3 days. The food that the queen larvae and worker larvae receive differs significantly in terms of fat, sugar, and protein contents after the first 3 days ([@B32]). Therefore, several studies have focused on genetic differences in queen and worker larvae caused by external factors after the age of 3 days ([@B43]; [@B19]), ignoring the influence of the external environment on the development of larvae within the first 3 days. It was found that during queen breeding, the age of the transplanted worker larvae had a significant influence on the development of queen bees, affecting the quality of the queen. Furthermore, the developmental differences occur between queen and worker larvae within the first 3 days of age. Enrichment analysis revealed that 95 pathways were enriched, based on the DEGs in both L1 vs. L2 and L1 vs. L3 groups; the most prominent among these pathways being insect hormone biosynthesis, longevity regulation, dorsoventral axis formation, MAPK, FoxO, mTOR, Hedgehog, TGF-β, Wnt, Hippo, and Toll and Imd signaling pathways. Hormones play a key role in the caste differentiation of bees. During the developmental stage, two increases in the juvenile hormone (JH) titer occur; the JH titers of queen larvae are significantly higher than those of the worker larvae. This phenomenon is closely related to the development of the ovaries. During the first increase (occurring in the first 5 days), there is a threshold of JH titer; if at this critical period, the larvae JH titer exceeds the threshold, then the ovaries are well developed ([@B4]). Furthermore, [@B27] showed that the size of the ovaries decreased when the larvae were transplanted at an older age. In this study, the DEGs were enriched for the insect hormone biosynthesis pathway and were downregulated with the increase in transplanted age, indicating that with transplanted age, hormone synthesis may be subdued. Moreover, 4 days of age is a key period for caste differentiation of the larvae ([@B18]); therefore, a decrease in the JH titer inevitably affects the development of the ovaries, eventually decreasing the queen quality. Previous data also showed that the Hippo, Wnt, TGF-β, and notch signaling pathways together influence the organ size of fruit flies ([@B5]), which supports the possibility that the development of queen ovaries may be subdued with an increase in the transplanted age. Further, TOR plays a key role in the bidirectional development of honeybees ([@B26]), and the Wnt, Hippo, notch, MAPK, and TOR signaling pathways were all involved in the caste differentiation of bees ([@B38]; [@B3]); caste differentiation is closely related to the larval development state. In addition, the FoxO signaling, dorso-ventral axis formation, Hedgehog, and TGF-β signaling pathways were found to jointly affect the growth of embryo ([@B12]; [@B14]). The longevity regulating pathway and Toll and Imd signaling pathway can affect the lifespan and immune function of the larvae, which is an important characteristic that differs between queen and worker bees. All the above-mentioned pathways were enriched in this study, indicating that the developmental direction of worker larvae changed by differing degrees due to the changes in activation of these biological pathways. We further analyzed these biological pathways and found that some genes are involved in several biological pathways simultaneously and that these pathways form a mutual regulatory network through these genes. Casein kinase 1 participated in four pathways, namely the FoxO, Wnt, Hedgehog, and Hippo signaling pathways. It has been proven to be involved in the regulation of mammalian cell proliferation and programmed cell death process ([@B28]; [@B35]). Transcriptional regulator Myc-B and S-phase kinase-associated protein 2 also participate in the regulation of mammalian cell proliferation ([@B22]; [@B6]), and S-phase kinase-associated protein 2 was associated with ovarian development in mice ([@B16]). Thus, decreased cell proliferation induces decreased body size, a smaller spermatheca, and fewer ovarioles in queens reared from older larvae. The results suggest that these pathways cross talked through the network to modify the developmental pathway of larvae; CK 1 is an important liaison. The results provide valuable information on the regulatory mechanism of environmental factors affecting the queen quality, which amplifies the understanding of caste differentiation in bees. Ethics Statement ================ This study was carried out in accordance with the recommendations of 'Animal Welfare Guidelines of Jiangsu Agri-animal Husbandry Vocational College, animal welfare committee of Jiangsu Agri-animal Husbandry Vocational College'. The protocol was approved by the 'animal welfare committee of Jiangsu Agri-animal Husbandry Vocational College'. Author Contributions ==================== LY designed the study and carried out the data analysis. KW participated in drafting the manuscript. LN, HXZ, and YYC participated in sample collection. TJ and GHC provided advice on data analysis and helped draft the manuscript. Conflict of Interest Statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. **Funding.** This work was supported by a grant from the National Natural Science Foundation of China (Grant No. 31502020), the earmarked fund for Modern Agro-industry Technology Research System from the Ministry of Agriculture of China (Grant No. CARS-45-SYZ6), and the fund for scientific research project provided by Jiangsu Agri-animal Husbandry Vocational College (grant number NSFPT201632). We thank research assistants Yunbo Xue and Jinsheng Niu of the Apiculture Science Institute of Jilin Province, China for providing the single drone-inseminated queens. Supplementary Material ====================== The Supplementary Material for this article can be found online at: <https://www.frontiersin.org/articles/10.3389/fgene.2018.00416/full#supplementary-material> ###### Correlation values of the three biological replicates of each sample. The correlation coefficients between the samples were obtained according to the mRNA expression, the closer the correlation coefficient is to 1, the more similar the expression pattern between samples is. The abscissa denotes the sample names, and the ordinate denotes the corresponding sample names. ###### Click here for additional data file. ###### Primer sequences of DEGs identified in various pathways in RNASeq for qRT-PCR validation. ###### Click here for additional data file. ###### The common DEGs in L1 vs. L2 and L1 vs. L3. The information of the DEGs both up/down regulated in L2 and L3 compared with L1, including the change-fold, ID, and the product of the DEGs. ###### Click here for additional data file. ###### GO enrichment analysis of DEGs. The GO terms of molecular function, biological process, and cellular component categories enriched by DEGs in L1 vs. L2 and L1 vs. L3. ###### Click here for additional data file. ###### The common pathways in L1 vs. L2 and L1 vs. L3. The information of the pathways enriched by DEGs in L1 vs. L2 as well as DEGs in L1 vs. L3, including the enrichment score, functional category and DEGs' ID. ###### Click here for additional data file. [^1]: Edited by: Jianke Li, Institute of Apiculture Research (CAAS), China [^2]: Reviewed by: Jinshan Xu, Chongqing Normal University, China; Zeng Zhi Jiang, Jiangxi Agricultural University, China; Dafu Chen, Fujian Agriculture and Forestry University, China [^3]: This article was submitted to Evolutionary and Population Genetics, a section of the journal Frontiers in Genetics

Introduction ============ According to the most recent catalogue of Ichneumonoidea ([@B12]), the tribe Phygadeuontini of the subfamily Cryptinae (Hymenoptera, Ichneumonidae), comprises 123 genera. [@B11] included 14 subtribes within Phygadeuontini (= Gelini of Townes). Two subtribes, Cephalobaridina and Gnypetomorphina, were subsequently synonymized with, respectively, the Phygadeuontina and Hemitelina by [@B2]. Prior to this publication 33 genera and 76 species of Phygadeuontini have been recorded from China ([@B7], [@B9], [@B10]). In this article, one new genus and two new species, collected in Quannan County, Jiangxi Province, China, are described. The new genus belongs to the subtribe Endaseina of the tribe Phygadeuontini. Type specimens are deposited in the Insect Museum, General Station of Forest Pest Management, State Forestry Administration, People's Republic of China. The specimens were collected using the entomological net in the forest of Quannan County, Jiangxi Province (China). The forest of Quannan is a forest composed of mixed deciduous angiosperms and evergreen conifers, mainly including Quercus spp., Castania spp., Castanopsis fabri Hance, Cinnamomum spp., Pinus massoniana (Lamb.). The morphological terminology is mostly that of [@B1]. Wing vein nomenclature is based on [@B5] and the terminology on [@B3], [@B4]). Taxonomy ======== Carinityla ---------- Sheng & Sun gen. n. urn:lsid:zoobank.org:act:8C613E44-3B20-423D-B78A-DA983E05F4CF ### Type species: Carinityla punctulata Sheng & Sun, sp. n. ### Etymology. The name of the new genus is based on the strongly swollen tyloids. The gender is feminine. ### Description. Fore wing length 7.2 to 8.8 mm. Head and thorax with dense and comparatively long hairs. Eye surface with short, sparse hairs. Upper margin of face slightly produced, weakly concave medially. Clypeus slightly convex, median portion of apical margin somewhat arcuate and distinctly raised. Mandible elongate, upper and lower margins almost parallel, upper tooth longer than lower tooth. Apical truncation of scape almost transverse. Apical half of antenna strongly flattened below in female. Flagellomeres 10 to 11 (12) of male with strongly swollen tyloids. Epomia long and strong, from lower-anterior angle of pronotum continuing to its dorsal portion ([Figure 4](#F1){ref-type="fig"}). Notauli present. Posterior edge of mesoscutum with transverse groove, which is unusually conspicuous and complete. Scutoscutellar groove without median longitudinal carina. Epicnemium with short transverse carina opposite lower corner of pronotum. Epicnemial carina strongly curved backward or broken ([Figure 5](#F1){ref-type="fig"}) above sternaulus. Anterior half of sternaulus deep; posterior half weak, reaching to posterior margin of mesopleuron above its lower posterior corner. Areolet pentagonal, receiving vein 2m-cu at or slightly basad of its outer corner ([Figures 1, 7](#F1){ref-type="fig"}, [10, 14](#F2){ref-type="fig"}). Vein 2m-cu subvertical, with one bulla. Hind wing vein 1-cu strongly inclivous, about 3.0 to 4.0 times as long as cu-a. Propodeum completely areolate, carinae very strong. Costula connecting area superomedia in front of its middle. Propodeal spiracle 3.0 to 3.5 times as long as wide. Median dorsal carinae of first tergum absent. Ovipositor compressed, tip very long and gradually tapered, with a weak nodus and very thin and indistinct ridges on ventral valve. ### Remarks. This new genus is similar to Amphibulus Kriechbaumer 1893 and Coptomystax Townes 1970 and can be distinguished from Amphibulus in the notaulus present; epicnemial carina strongly curved backward or secondary carina present above sternaulus; fore wing vein 2m-cu almost reaching 3rs-m ([Figure 1, 7](#F1){ref-type="fig"}, [10, 14](#F2){ref-type="fig"}). In Amphibulus, the notaulus is absent; the epicnemial carina is neither curved backward nor is there a secondary carina above the sternaulus; fore wing vein 2m-cu is usually far from 3rs-m, usually connecting with the areolet near its median portion. The new genus can be distinguished from Coptomystax Townes by the eye surface with short sparse hairs; the upper margin of face without tubercle; a transverse groove at the posterior edge of mesoscutum distinct, complete and unusually conspicuous; the epicnemial carina approaching the subalar prominence, or discontinuous above the sternaulus. In Coptomystax the eye surface is bare; the upper edge of the face has a high and compressed median tubercle; the transverse groove at the posterior edge of the mesoscutum is distinct medially, evanescent laterally; the epicnemial carina ends below the middle of the hind edge of the pronotum. This new genus can also be easily distinguished from the related genera Endasys Förster 1869 and Cisaris Townes 1970 by the following combination of characters: scutoscutellar groove without median longitudinal carina, median dorsal carinae of first tergum absent (Endasys: scutoscutellar groove with a median longitudinal carina, first tergum with distinct median dorsal carinae); fore wing with distinct areolet, posterior edge of mesoscutum with transverse groove (Cisaris: fore wing without areolet, posterior edge of mesoscutum without transverse groove). In [@B11] key to genera, the new genus can be inserted as follows: ---- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------- 7 Median dorsal carinae of first tergite strong, at least in front of spiracle. Prescutellar transverse groove with a strong median longitudinal carina. Holarctic Region 6\. Endasys -- Median dorsal carinae of first tergite weak or absent. Prescutellar transverse groove usually without a strong median longitudinal carina 7' 7' Upper margin of face edge-shaped, without median tubercle. Apical half of antenna strongly flattened below in female. Areolet receiving vein 2m-cu at or slightly basad of its outer corner. Eye surface with sparse hairs Carinityla Sheng & Sun, gen. n. -- Upper margin of face concave, at least not edge-shaped, with a median tubercle. Apical half of antenna not flattened below. Areolet receiving vein 2m-cu near its center. Eye surface with or without hairs 8 ---- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------- ### Carinityla punctulata Sheng & Sun sp. n. urn:lsid:zoobank.org:act:8C613E44-3B20-423D-B78A-DA983E05F4CF [Figures 1--9](#F1){ref-type="fig"} #### Etymology. The name of the new species is based on the dense punctures on the head and thorax. #### Types. *Holotype*, female, CHINA: Quannan County, 628m, Jiangxi Province, 9 June 2010, leg. Shi-Chang Li. *Paratypes*: 7 males, CHINA: Quannan County, 628 to 700m, Jiangxi Province, 16 May to 10 June 2008, leg. Shi-Chang Li. 1 female and 1 male, CHINA: Quannan County, 628 to 700m, Jiangxi Province, 26 to 31 May 2010, leg. shi-Chang Li. #### Diagnosis. Second tergum, hind femur and tibia reddish brown. Notaulus not reaching to center of mesoscutum. Scutellum weakly convex, lateral sides not raised, without carina except at its basal corner. {#F1} #### Description. Female. Body length 9.3 to 9.7 mm. Fore wing length 7.2 to 7.8 mm. Ovipositor length about 2.8 mm. Head and mesosoma with dense punctures and long yellowish brown hairs. #### Head. Face ([Figure 2](#F1){ref-type="fig"}) convex, approximately 2.2 times as wide as long, with dense, irregular punctures, median portion with short longitudinal wrinkles. Clypeal suture vestigial between clypeal foveae. Clypeus slightly convex, basal portion with punctures sparser than on face, subapically with shallow transverse concavity; apical portion smooth and impunctate, distinctly raised medially. Subbasal portion of mandible with short longitudinal wrinkles, apical portion with sparse shallow punctures; upper tooth distinctly longer than lower tooth. Cheek and gena with dense punctures, distance between punctures 0.2 to 0.5 times diameter of puncture. Subocular sulcus distinct. Malar space 0.4 to 0.5 times as long as basal width of mandible. Gena slightly convergent backward, in dorsal view 0.7 to 0.8 times as long as width of eye. Vertex ([Figure 3](#F1){ref-type="fig"}) with dense punctures. Postero-ocellar line about 0.44 times as long as ocular-ocellar line. Frons approximately flat, with even and dense punctures, distance between punctures 0.2 to 0.5 times diameter of puncture. Antenna distinctly shorter than body length, with 27 flagellomeres, ratio of length of flagellomere 1:2:3:4:5 is 3.7:4.1:4.0:4.0:3.9. Flagellomeres 10 to 11 (12) of male with strongly swollen tyloids ([Figure 8](#F1){ref-type="fig"}) on apical half of flagellomere and at most half as long as flagellomere. Occipital carina complete and strong, joining oral carina above base of mandible. #### Mesosoma. Anterior portion of pronotum with weak longitudinal wrinkles and fine punctures; laterally concave and lower portion with oblique transverse wrinkles; upper posterior portion with fine punctures, upper posterior margin shoulder-shaped, raised narrowly. Mesoscutum with dense elongate punctures. Notaulus present on front portion of mesoscutum. Scutoscutellar groove with very weak longitudinal wrinkles. Scutellum slightly convex, with irregular punctures. Postscutellum concave, smooth. Subalar prominence strongly convex. Anterior and upper margins of mesopleuron with fine punctures; median portion of mesopleuron ([Figure 5](#F1){ref-type="fig"}) with irregular transverse punctures; speculum with dense and fine punctures. Epicnemial carina broken above sternaulus, upper end of lower portion connecting with short transverse carina opposite lower corner of pronotum; upper portion of epicnemial carina oblique, upper end reaching about half distance to subalar prominence. Metapleuron with dense punctures, distance between punctures 0.2 to 0.5 times diameter of puncture. Juxtacoxal carina complete. Anterior portion of submetapleural carina strongly lobed. Wings brownish hyaline. Fore wing with vein 1cu-a slightly distal of 1-M by less than vein width. Vein 2-Cu approximately 2.0 times as long as 2cu-a. Hind wing vein 1-cu about 3.0 times as long as cu-a. Legs robust, with dense brown hairs. Hind coxa and femur with distinct fine punctures. Spurs of hind tibia about half length of first tarsomere. Ratio of length of hind tarsomeres 1:2:3:4:5 is 10.0:4.5:3.5:1.8:3.7. Propodeum ([Figure 6](#F1){ref-type="fig"}) with sandy beige long hairs. Area superomedia hexagonal, 1.2 times as wide as long, costula connecting slightly in front of its middle. Area basalis smooth, vaguely punctate. Area externa with distinct punctures. Area superomedia with indistinct longitudinal wrinkles. Area dentipara with oblique longitudinal wrinkles. Area lateralis with oblique transverse wrinkles. Area petiolaris with transverse wrinkles. Propodeal spiracle approximately 3.3 times as long as wide, almost touching lateral longitudinal carina (closer to lateral longitudinal carina than to pleural carina). Propodeal apophysis short and compressed. #### Metasoma. First and second terga smooth and shining, with very sparse and fine punctures. First tergum about 2.3 times as long as apical width. Postpetiole evenly convex. Median dorsal carinae absent. Dorsolateral and ventrolateral carinae complete. Spiracle circular, very small, slightly convex, located at apical 0.4 of first tergum. Second tergum 0.5 to 0.6 times as long as apical width. Remaining terga with short brown hairs. Ovipositor sheath approximately 0.95 times as long as hind tibia. Ovipositor compressed, with weak nodus. #### Color ([Figure 1](#F1){ref-type="fig"}). Black, except the following. Ventral side and apical portion of scape, apical portion of pedicel, ventral side of basal portion (more or less) and flat side of flagellomeres, tegula brown. Dorsal sides of seventh to thirteenth flagellomeres white. Median portion of mandible, dorsal sides of front and mid femora and tibiae brown. Maxillary and labial palpi, fore and mid coxae, trochanters and ventral sides of femora yellowish brown. Fore and mid tarsi dark brown. Hind coxa brown to yellowish brown. Hind trochanter, femur and tibia reddish brown. Apical ends of hind femur and tibia, hind first tarsomere brownish black. Hind second to fifth tarsomeres blackish brown. First and second terga, basal margin of third tergum reddish brown. Posterior margin of third to sixth terga slightly narrowly tinged brown. Main portions of seventh and eighth terga white. Fore wing with stigma brown, veins blackish brown. Hind wing with veins brown. #### Male ([Figure 7](#F1){ref-type="fig"}). Body length 9.5 to 11.0 mm. Fore wing length 7.2 to 8.5 mm. Face 1.7 to 1.8 times as wide as long. Antenna with 26 to 28 flagellomeres. Upper posterior portion of pronotum, in front of tegula, weakly convex. Notaulus present, almost reaching to center (about 0.4) of mesoscutum. Area superomedia inverse trapeziform, 1.9 to 2.1 times as wide as long, costula connecting at its anterior 0.2 ([Figure 9](#F1){ref-type="fig"}). First tergum 2.6 to 2.7 times as long as apical width. Antennae with dorsal profiles of eighth to thirteen flagellomeres white. Apical half of hind first tarsomere and second to fourth tarsomeres buff. First to third terga reddish brown. #### Host. Unknown. ### Carinityla pilosa Sheng & Sun sp. n. urn:lsid:zoobank.org:act:1F3772D3-2D07-45A3-985B-B547E618578C [Figures 10--16](#F2){ref-type="fig"} #### Etymology. The specific name is derived from the long and dense hairs on the body. #### Types. *Holotype*, female, CHINA: Quannan County, 650m, Jiangxi Province, 29 June 2010, leg. Shi-Chang Li. *Paratypes*: 7 males, CHINA: Quannan County, 530 to 628m, Jiangxi Province, 12 May to 10 June 2008, leg. Shi-Chang Li. 3 males, CHINA: Quannan County, 628 to 700m, Jiangxi Province, 31 May to 18 June 2010, leg. Shi-Chang Li. #### Diagnosis. Second tergum, hind femur and tibia black. Notaulus of male reaching beyond center of mesoscutum. Lateral sides of scutellum raised and median portion weakly concave in male. Scutellum with lateral carina extending 0.2 to 0.3 of its length. {#F2} #### Description. Female. Body length about 9.0 mm. Fore wing length about 7.2 mm. Ovipositor length about 2.8 mm. Head and mesosoma with long and dense yellowish brown hairs. #### Head. Face ([Figure 11](#F2){ref-type="fig"}) strongly convex medially, approximately 1.9 times as wide as long, with dense and irregular punctures. Clypeal suture vestigial between clypeal foveae. Clypeus slightly convex, basal portion with sparse and irregular punctures, distance between punctures 1.0 to 3.0 times diameter of puncture; subapical portion with shallow, transverse concavity; apical 0.2 smooth and impunctate, median section of apical margin distinctly raised. Mandible with indistinct longitudinal wrinkles and fine punctures; upper tooth longer than lower tooth. Cheek with elongate punctures. Subocular sulcus indistinct. Malar space approximately 0.5 times as long as basal width of mandible. Gena with dense punctures, distance between punctures 0.2 to 1.0 times diameter of puncture; in dorsal view approximately 0.6 times as long as width of eye. Vertex with irregular punctures, distance between punctures 0.2 to 1.5 times diameter of puncture. Postero-ocellar line about 0.6 times as long as ocular-ocellar line. Frons approximately flat, with regular and dense punctures, distance between punctures 0.2 to 1.0 times diameter of puncture. Antenna distinctly shorter than body in length, with 27 flagellomeres, ratio of length of flagellomeres 1:2:3:4:5 is 3.7:4.7:4.5:4.3:4.1. Flagellomeres 10 to 11 (12) of male with tyloids ([Figure 15](#F2){ref-type="fig"}) similar to those of Carinityla punctulata. Tyloid on flagellomere 11 0.7 to 0.9 times as long as flagellomere. Occipital carina complete and strong, joining oral carina above base of mandible. #### Mesosoma. Anterior portion of pronotum with fine punctures, laterally concave and lower portion with dense, oblique transverse wrinkles; upper posterior portion with fine punctures, distance between punctures 0.2 to 1.0 times diameter of puncture; upper posterior margin slightly and narrowly raised. Mesoscutum with dense punctures, distance between punctures 0.2 to 0.5 times diameter of puncture; posterior median portion with irregular longitudinal wrinkles. Notaulus present anteriorly. Scutoscutellar groove with fine longitudinal wrinkles. Scutellum almost flat, with irregular punctures. Postscutellum smooth and shining, lateral portion strongly convex, anterior-lateral portion deeply concave. Subalar prominence strongly convex. Upper portion of mesopleuron ([Figure 12](#F2){ref-type="fig"}) with dense punctures; lower portion, above sternaulus, with irregular punctures; lower posterior portion with transverse wrinkles. Speculum with dense punctures. Epicnemial carina strongly curved backward above sternaulus, upper end reaching to subalar prominence. Metapleuron with dense and irregular punctures. Juxtacoxal carina complete. Anterior section of submetapleural carina strongly projecting. Wings brownish hyaline. Fore wing with vein 1cu-a almost opposite 1-M. Vein 2-Cu approximately 2.0 times as long as 2cu-a. Hind wing vein 1-cu about 3.0 times as long as cu-a. Legs robust, with long and dense brown hairs. Hind coxae and femora with distinct fine punctures. Spurs of hind tibia approximately half length of first tarsomere. Ratio of length of hind tarsomere 1:2:3:4:5 is 10.0:4.7:3.4:1.6:3.7. Propodeum ([Figure 13](#F2){ref-type="fig"}) with long, brown hairs. Area superomedia hexagonal, approximately 1.15 times as wide as long, costula connecting slightly in front of its middle. Area basalis and area superomedia smooth and shining. Area externa with fine punctures. Area dentipara with indistinct wrinkles. Area spiracularis almost smooth. Area lateralis with oblique transverse wrinkles. Area petiolaris and area posteroexterna with transverse wrinkles. Propodeal apophysis short and compressed. Propodeal spiracle approximately 3.0 times as long as wide, distance to pleural carina approximately 1.3 times as long as distance to lateral longitudinal carina. #### Metasoma. First to third terga smooth and shining. First tergum approximately 2.3 times as long as apical width. Hind section of dorsolateral carina, behind spiracle, indistinct. Lateral margins of petiole almost parallel, only posterior portion slightly broadened. Postpetiole weakly and evenly convex, anterior half with sparse and fine punctures. Spiracle circular, very small, located at about apical 0.3 of first tergum. Second tergum approximately 0.65 times as long as apical width. Remaining terga with short brown hairs and indistinct punctures. Ovipositor sheath approximately as long as hind tibia. Ovipositor compressed, with weak nodus. #### Color ([Figure 10](#F2){ref-type="fig"}). Black, except the following. Ventral profiles of scape and pedicel dark brown. Flat portion of flagellomeres more or less brown. Dorsal profiles of eighth to fourteenth flagellomeres white. Maxillary and labial palpi buff except dark bases. Median portion of mandible crimson. All coxae and trochanters, inner profiles of front and mid femora yellowish brown. Remaining portion of fore legs and mid femora brown. Apices of mid femora, mid tibiae and tarsi puce. Apical portion of first tarsomere of hind tarsi, second to fourth tarsomeres, posterior median portions of sixth and seventh terga, main portion of eighth tergum white. Petiole of first tergum yellowish brown; postpetiole reddish brown. Stigma yellowish brown. veins brownish black. #### Male ([Figure 14](#F2){ref-type="fig"}). Body length 9.5 to 12.0 mm. Fore wing length 7.5 to 8.8 mm. Face 1.7 to 1.8 times as wide as long. Malar space 0.2 to 0.3 times as long as basal width of mandible. Antenna with 27 to 29 flagellomeres. Upper posterior portion of pronotum, in front of tegula, weakly convex. Notaulus long, reaching beyond center of mesoscutum. Lateral sides of scutellum raised, median portion weakly concave; basal 0.2 to 0.3 with lateral carina. Median portion of mesopleuron smooth and shining, impunctate. Area superomedia 1.5 to 1.6 times as wide as long, costula connecting at its anterior 0.3 ([Figure 16](#F2){ref-type="fig"}). Propodeal spiracle 3.0 to 3.5 times as long as wide. First tergum 2.3 to 2.5 times as long as apical width. Dorsal profile of basal flagellomeres brownish black, ventral profile reddish brown; dorsal profiles of seventh to thirteen flagellomeres white; apical flagellomeres brownish black. Stigma and veins brownish black. #### Host. Unknown. Key to species of Carinityla Sheng & Sun ---------------------------------------- ---- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------- 1 Female 2 -- Male 3 2 Epicnemial carina complete, strongly curved backward above sternaulus. First tergum lacking hind portion of dorsolateral carina (behind spiracle). Hind femora and second tergum black Carinityla pilosa Sheng & Sun, sp. n. -- Epicnemial carina broken above sternaulus. Dorsolateral carina of first tergum complete. Hind femora and second tergum reddish brown Carinityla punctulata Sheng & Sun, sp. n. 3 Notaulus reaching beyond center of mesoscutum. Lateral sides of scutellum raised, median portion weakly concave. Area superomedia of propodeum 1.5 to 1.6 times as wide as long. Hind femora and second to third terga black Carinityla pilosa Sheng & Sun, sp. n. -- Notaulus not reaching to center of mesoscutum. Scutellum normal, weakly convex, lateral sides slanting downwards. Area superomedia of propodeum 1.9 to 2.1 times as wide as long. Hind femora and second to third terga reddish brown Carinityla punctulata Sheng & Sun, sp. n. ---- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------- Supplementary Material ====================== ###### XML Treatment for Carinityla The authors are deeply grateful to Dr. Gavin Broad and an anonymous referee, for reviewing this manuscript, and also thank Shi-Chang Li, Jun-Gen Luo and Dong-Sun Ding for their help in the course of exploration in Jiangxi Province. This research was supported by the National Natural Science Foundation of China (NSFC, No. 30671686; No. 30872035). [^1]: Academic editor: Gavin Broad

Infectious diseases ranging from respiratory (influenza, common cold, tuberculosis, the respiratory syncytial virus), vector-borne (plague, malaria, dengue, chikungunya, and Zika) to sexually transmitted (the human immunodeficiency virus, syphilis) have historically affected the human population in profound ways. For example, the Great Plague, well known as the Black Death, was caused by the bacterium *Yersinia pestis* and killed up to 200 million people in Eurasia and about 30--60% of Europe's population during a 5-year span in the fourteenth century. At the time, the plague infection was thought to be due to some "bad air" and it was not discovered that bites of infected fleas were behind the pandemic until late 1890s. If the human civilization had known about the transmission mechanisms behind the plague infections, the epidemic's impact on morbidity and mortality could have been mitigated through basic public health interventions. This is to say that knowledge of the transmission processes and the natural history of infectious diseases in different environments represents invaluable actionable information for thwarting the spread of infectious diseases. Fortunately, over the years human civilization has made great strides in increasing our understanding of the transmission dynamics of emerging and re-emerging infectious diseases. For instance, John Snow, known as the father of modern epidemiology, mapped the location of cholera cases during the 1854 epidemic in Soho, London, and made the link between the spatial distribution of cholera cases and a pump that he hypothesized as the source of the disease (Fig. [1.1](#Fig1){ref-type="fig"}). Following his observations, the pump was removed to avoid further exposures, and the number of cases subsided.Fig. 1.1The number of new cholera cases during the 1854 epidemic in Soho, London One remarkable and definite shift to the germ theory occurred during the "golden bacteriology" era during the second half of the nineteenth century. In fact, the 1889--1990 influenza pandemic is arguably the first influenza pandemic that occurred in a new and progressive state of knowledge about infectious disease transmission. This pandemic is better known as the "Russian Flu" because the rapid global spread of the pandemic virus can be traced back to Saint Petersburg, Russia in October 1889 (Valleron et al. [@CR27]). Moreover, it was the first pandemic to unfold in a world connected by rail and maritime transportation; it spread across Europe in approximately 6 weeks, with an estimated mean speed at 394 km/week (Valleron et al. [@CR27]) and circulated around the world in just 4 months (Valleron et al. [@CR27]). Following the 1889--1990 influenza pandemic, in 1918 a novel influenza virus struck the world and killed 20--100 million people, a figure that easily exceeds the death toll associated with World War I (Johnson and Mueller [@CR17]; Dahal et al. [@CR14]; Mills et al. [@CR22]). In the USA alone, about 675,000 people succumbed to the 1918 pandemic virus (Fig. [1.2](#Fig2){ref-type="fig"}). However, it was not discovered until years later that an influenza virus was responsible for this pandemic. One hundred years after the 1918 pandemic, we not only remember this devastating historic health disaster, it also serves as a stark reminder of the public health impact that the influenza virus continues to exert on the global population. The 1918 "Spanish Flu" pandemic represents one of the most important case studies for pandemic preparedness available today. However, locating death records to reconstruct the mortality impact of this pandemic requires the arduous task of searching for these documents in old cemeteries, public archives, parishes, and church records (Alonso et al. [@CR2]; Simonsen et al. [@CR26]).Fig. 1.2Excess death rate associated with the 1918 influenza pandemic in US cities that exhibited the highest peak excess death rates The application of mathematical and statistical tools to investigate and forecast evolving epidemics and pandemics has increased significantly during the last couple of decades from ∼50 to \>800 publications per year (Fig. [1.3](#Fig3){ref-type="fig"}). The worldwide epidemic of acquired immunodeficiency syndrome (AIDS), caused by the human immunodeficiency virus (HIV), started in the early 1980s and accelerated the applications and developments of mathematical and statistical models. This contributed to the understanding of factors that promote transmission of HIV and of strategies for preventing transmission. While the number of studies that apply mathematical modeling to study infectious disease dynamics has rapidly increased over the last two decades (Fig. [1.3](#Fig3){ref-type="fig"}), the great majority of those studies are still associated with HIV/AIDS, although this trend has declined somewhat during the last decade, followed by tuberculosis. In addition, the number of studies associated with emerging infectious diseases such as Ebola, dengue, chikungunya, and Zika has been increasing during the last 5--10 years as a result of recent regional and global epidemics (Fig. [1.3](#Fig3){ref-type="fig"}).Fig. 1.3Number of publications on mathematical modeling and infectious disease (left panel) and the fraction of those publications related to different infectious diseases (right panel) by publication year The Motivation {#Sec1} ============== Mathematical modeling plays an important role in ordering our thoughts and sharpening vague intuitive notions. Initial models are verbal descriptions that tend to become insufficient as soon as the scenarios become complicated. Mathematics provides a powerful language that forces us to be logically consistent and explicit about assumptions. Over the years, we have encountered very interesting, inspiring, and challenging discussions at the end of workshops on infectious disease modeling with the following recurrent themes: While most disease transmission models predict an expected exponential growth at the beginning of the epidemic, empirical data often exhibit sub-exponential growth patterns (Viboud et al. [@CR29]). How do we best characterize these non-unique sub-exponential growth functions in the context of infectious disease modeling?Are there many, even infinitely many, mechanisms that lead to the same or very similar sub-exponential growth functions?Does a slower than expected initial growth at the beginning of the epidemic imply a smaller value of the basic reproduction number *R* ~0~, a key quantity in the field of infectious disease epidemiology (Anderson and May [@CR3]; Diekmann and Heesterbeek [@CR15]; Brauer [@CR7]), as suggested by many transmission models?What exactly does it mean when we say "deterministic models approximate their stochastic counterparts by the law-of-large numbers"? Are we referring to a population that is infinitely large or something else?Which features of the population-based models, in which the exponential distribution is assumed at the individual levels, can be generalized with non-exponential distributions?Regarding effectiveness of control measures against the spread of diseases, even if imperfect implementation in terms of coverage or compliance has been explicitly taken into account in the models, empirical observations often leave us with impressions that the control measure that "looks good" in theory "do not work at all" in practice. Are there more theories that could capture this phenomenon?How do we reconcile the quantities as predicted by disease transmission models with observed data from outbreak investigations and public health surveillance?The need for precise definitions of verbal descriptions in quantitative analyses. For instance, What do we mean by "a case" when data from outbreak investigations and surveillance are presented as time-series of "number of cases"?Are "generation intervals" consistently defined across literature in epidemiology and infectious disease models?How do we characterize and compare "variability" among random variables, such as the infectious periods or the numbers of secondary infections transmitted by a primary infector?What do we mean by "non-identifiability" when fitting models to data? Of models formulated in mathematical languages, there are different types that are designed for different purposes. Broadly speaking, there are mathematical models aimed at facilitating our understanding of the medical, biological, ecological, and social interactions that manifest the outbreaks and epidemics of infectious diseases in order to gain insight into specific questions or to generate theories about what must or might happen; and there are statistical models aimed at capturing the data generation process, for detecting general patterns, predicting epidemic trajectories, managing control strategies, or simply describing epidemic trends. Within both mathematical and statistical models, there are models designed at the population level in a phenomenological way versus models that are individual-based with which researchers aim to capture relevant mechanistic processes. Individual-based models start from descriptions or assumptions about the evolution of the infectiousness and the natural history of the disease progression within an infected host. These include models for the latent periods, the infectious periods, the incubation periods, recovery, mortality, and so on. Some of the individual-based models also combine social contacts with the evolution of the infectiousness in terms of infectious contacts (Dietz [@CR16]). Phenomenological models can be deterministic or stochastic and include transmission dynamics models formulated using differential equations or stochastic processes as well as empirical growth functions, such as the generalized logistic growth models. Transmission dynamics models depend on tacit assumptions at the individual level. The developments of many new statistical models and methods in the study of infectious diseases were driven by the HIV epidemic (Brookmeyer and Gail [@CR9]). Data arising from infectious disease investigations pose unique challenges in classic statistical theory and practice because disease outbreak data do not arise from designed experiments. Each outbreak cannot be repeated naturally under identical conditions, whereas the large amount and multiple sources of clinical data, outbreak investigation data from non-conventional surveys, public health surveillance, and observational data from prevalence and incidence cohorts are collected addressing the same outbreak event. Before statistical methods are used to understand and control the epidemic, statistical models are needed to address the data generation processes, which not only include the epidemiologic and biologic processes that give rise to the disease outbreaks, but also the processes that dictate how data are observed and how a "case" is documented and reported. When talking about "fitting the model to data," we tend to think of one type of model designed for a specific purpose. However, fitting a dynamic mathematical model to observed outbreak data (e.g., for the purpose of estimating important transmission parameters) involves all three levels of models: the population phenomenological model which depends on tacit assumptions of the individual-based model nested within it, and the statistical model that links the disease transmission process to the data generation process. Very often in practice, these different types of models are considered simultaneously even without the investigators' consciousness. Driven by the HIV epidemic that started in the late 1970s, the outbreaks of the severe acute respiratory syndrome (SARS) in 2003, pandemic influenza preparedness, and preparedness for other emerging and re-emerging epidemics, the literature on infectious disease modeling has flourished during the past 40 years. However, most articles are confined within subdisciplines according to model characteristics and research focus. While the field of mathematical epidemiology has a long history (e.g., Ross [@CR23], [@CR24]; Anderson and May [@CR3]; Diekmann and Heesterbeek [@CR15]; Keeling and Rohani [@CR18]; Sattenspiel [@CR25]; Allen [@CR1]; Vynnycky and White [@CR31]; Becker [@CR5]; Andersson and Britton [@CR4]; Manfredi and D'Onofrio [@CR21]; Kermack and McKendrick [@CR19]; Brauer [@CR7]; Brauer and Castillo-Chávez [@CR8]), formal efforts at connecting mathematical models with epidemiological data with the goal of calibrating models for predictive/forecasting purposes have only started to take hold during the last decade (Chretien et al. [@CR13]; Biggerstaff et al. [@CR6]; Chowell [@CR10]; Viboud et al. [@CR30]). Structure of the Book with Brief Summary {#Sec2} ======================================== Chapter 10.1007/978-3-030-21923-9_2 provides a review of basic concepts of probability and statistical models for the distributions of continuous lifetime data, closely related to individual-based models that describe the evolution of infectiousness and the natural history of the disease progression. We re-tell the story from a different angle with emphases on the shapes of hazard functions and tail properties of the lifetime distributions instead of repeating the subject commonly found in a typical textbook on survival analysis. These characteristics have profound impacts on outcomes of the transmission dynamic models at the population level. We will discuss and compare two lifetime random variables, both in terms of magnitude and variability, together with the Laplace transform of lifetime distributions. These concepts will provide the foundations for most of the remaining chapters. Chapter 10.1007/978-3-030-21923-9_3 addresses the distributions of random counts and counting processes, which are closely related to population-based phenomenological models. Section 10.1007/978-3-030-21923-9_3\#Sec6 provides a framework that links the continuous lifetime distributions at the individual level to the distributions of random counts at the population level. It also provides a historical account. Contemporary discussions on "super-spreading events" as seen in outbreak investigation data in SARS-like diseases are typically associated with transmissions along highly heterogeneous networks characterized by long tailed degree distributions (Lloyd-Smith et al. [@CR20]). Similarly, in the context of incurring accidents, publications in actuarial science journals can be traced back to debates on proneness, contagion, or spells in the first half of the twentieth century that gave rise to important models such as the mixed-Poisson process and the Yule process. Section 10.1007/978-3-030-21923-9_3\#Sec11 lays the foundation for measuring the evolution of random counts over time, which are key measurements in all population-based models. Chapter 10.1007/978-3-030-21923-9_4 focuses on behaviors of a disease outbreak during the initial phase, immediately after a single (or very few) infected individual are "seeded" into a very large susceptible population. The first part discusses extinction versus growth and relationships among three key parameters: the basic reproduction number *R* ~0~, the initial (exponential) growth rate *r*, and the probability of extinction *δ* are made and established. With the notion of the "prevalence cohort" (Fig. 10.1007/978-3-030-21923-9_4\#Fig8), we re-write the classic Lotka equation (10.1007/978-3-030-21923-9_4\#Equ36) as (10.1007/978-3-030-21923-9_4\#Equ40) under the assumptions about homogeneous mixing. It reveals that: *R* ~0~ only depends on the average value of the infectious periods regardless of the variance or the exact distribution. In models without natural births and deaths in the population, the value of *R* ~0~ is not affected by the presence or absence of latent periods.The probability of extinction *δ* depends on the specific distribution of the infectious periods but is not affected by the presence or absence of latent periods.If the infectious disease does not become extinct during the first few generations, the initial (exponential) growth rate *r* depends on specific distributions for both the latent periods and the infectious periods.Each of the mathematical relationships between *R* ~0~ and *δ*, and between *R* ~0~ and *r*, as found in the literature, is under a set of strict assumptions on the social contact process and the progression of infectiousness within infected individuals. Therefore, Given the fixed value *R* ~0~ \> 1 and the infectious periods distribution, the model with latent periods has a smaller initial growth rate *r* than the one without.Given the fixed value *R* ~0~ \> 1 and the latent periods distribution, the more variable the infectious periods, the smaller the value of *r*.Without specifying the distributions of the latent periods and the infectious periods, there is no order between the values of *r* and of *R* ~0~.If *R* ~0~ \> 1, without specifying the distribution of the number of secondary infections generated by the primary infectious individual (through the distribution of the infectious periods), there is no order between the values of *δ* and of *R* ~0~.There is a direct relationship between *r* and *δ*, rarely mentioned in the literature, that *r* = *β*(1 − *δ*), provided that there is no latent period and that the number of infections produced by a typical infectious individual during a time interval of length *x* is Poisson distributed with mean value *βx*. This relationship does not depend on the distribution of the infectious period. The second part of Chap. 10.1007/978-3-030-21923-9_4 emphasizes that the three parameters *R* ~0~, *δ*, and *r* are intrinsic in the sense that they represent the state of the system at (disease-free) equilibrium when the initially infected individuals are seeded. Section 10.1007/978-3-030-21923-9_4\#Sec39 presents growth patterns that are most likely to happen when the system moves away from the equilibrium condition. Many discussions are on empirically observed slower growth patterns that largely deviate from the exponential growth assumption (Chowell et al. [@CR11]; Chowell [@CR10]). We attempt to precisely define the sub-exponential growth functions in the context of infectious disease transmission and enlist several assumptions about the transmission dynamics that all lead to such early growth pattern, from the depletion of the susceptible population to scaling of epidemic growth shaped by various factors and their combination including the level of contact clustering and reactive behavior changes (Chowell et al. [@CR12]) and to unobservable individual-level heterogeneity. A special sub-exponential growth function of the form, (1 + *rvt*)^1∕*v*^, *r*, *t* \> 0, 0 \< *v* ≤ 1, is introduced in Chap. 10.1007/978-3-030-21923-9_4 which frequently appears in later chapters (10.1007/978-3-030-21923-9_6, 10.1007/978-3-030-21923-9_8 and 10.1007/978-3-030-21923-9_9) in examples and discussions. Chapters 10.1007/978-3-030-21923-9_5 and 10.1007/978-3-030-21923-9_6 discuss compartment models when the outbreak moves beyond the initial phase. Much of Chap. 10.1007/978-3-030-21923-9_5 is the synthesis of previously published literature on both stochastic and deterministic transmission dynamic models, with our added perspectives. Our interest is to generalize some of the features of these models beyond the assumptions based on the exponential distribution on durations of various stages, and beyond the simple generalizations such as the Erlang distribution (which is a subset of the gamma distribution characterized by smaller variances compared to the exponential distribution with equal mean values). These discussions start in Sect. 10.1007/978-3-030-21923-9_5\#Sec32 and continue in Sect. 10.1007/978-3-030-21923-9_6\#Sec3. In these discussions, Laplace transforms of probability distributions are extensively used as tools to calculate transition probabilities among compartments and average durations within compartments. They are valid for arbitrary distributions without specific assumptions of these distributions. When these distributions are exponential, general results in Sects. 10.1007/978-3-030-21923-9_5\#Sec32 and 10.1007/978-3-030-21923-9_6\#Sec3 return to those published in the literature, such as the expression of the reproduction number as the non-negative eigenvalue of the next generation matrix (van den Driessche and Watmough [@CR28]) as well as in examples in these sections. We also point out a transcendental relationship among (10.1007/978-3-030-21923-9_4\#Equ43), (10.1007/978-3-030-21923-9_5\#Equ66), and (10.1007/978-3-030-21923-9_6\#Equ24). In these expressions, the Laplace transforms are tools to compare distributions ranked by variability which lead to Propositions 10.1007/978-3-030-21923-9_6\#FPar2 and 10.1007/978-3-030-21923-9_6\#FPar3 along with discussions in subsequent paragraphs. Other distinct topics in Chap. 10.1007/978-3-030-21923-9_5 are empirical models to describe population-based phenomena without "mechanically" modeling the transmission dynamics at the level of individuals and interactions among individuals. These models are useful for curve fitting, as used in examples later in Chap. 10.1007/978-3-030-21923-9_8. Models in Chap. 10.1007/978-3-030-21923-9_6 are more complex and involve intervention measures during the epidemic. Section 10.1007/978-3-030-21923-9_6\#Sec4 demonstrates a potential application of these models in the context of preparedness for an influenza-like acute respiratory infectious disease with numerical illustrations in hypothetical race-to-treat scenarios and with limited treatment supply. Section 10.1007/978-3-030-21923-9_6\#Sec16 discusses the impact of unobservable heterogeneity in treatment rates on effectiveness. This section addresses Question 6 in Sect. [1.1](#Sec1){ref-type="sec"}. We also draw the attention of the expression $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document} $$\left ( 1+\phi xv\right ) ^{-1/v}$$ \end{document}$ in (10.1007/978-3-030-21923-9_6\#Equ31) which echoes the sub-exponential growth function (1 + *rvt*)^1∕*v*^ in Chap. 10.1007/978-3-030-21923-9_4. This is because in both cases, a frailty model from survival analysis is used to model the unobservable heterogeneity among individuals. Chapter 10.1007/978-3-030-21923-9_7 addresses Question 7, 8, and 9 in Sect. [1.1](#Sec1){ref-type="sec"} and serves as a transition between the theoretical topics in previous chapters and Chaps. 10.1007/978-3-030-21923-9_8 and 10.1007/978-3-030-21923-9_9. The focus is on the data generating processes and statistical issues of fitting models to data. As repeatedly emphasized in Chaps. 10.1007/978-3-030-21923-9_4--10.1007/978-3-030-21923-9_6, population-based models involve tacit assumptions at the level of individuals, such as the exponential, gamma, or other distributions of the infectious periods. These are conceptual models to address general issues and general patterns, such as the prediction of "incidence" according to time at infection (which is usually unobservable). On the other hand, statistical models address the data generating processes, which include the epidemiology aspects but also the observational schemes, including "case definition," surveillance and reporting, and adjustments for observational biases. In each model, choices are made with respect to which aspects of "the real world" should be included in the description of the model and which should be ignored. These choices not only depend on the perceived importance of various factors, but also on the purpose of each of these models. Frequently, fitting a mathematical model, such as a transmission model, to data collected from surveillance and outbreak investigations involves three types of models (assumptions) that take place at the same time. This requires "nesting" one type of model within another. For example, the statistical model that describes data may involve assumptions of the mean and variance, and in some instances, the assumptions of specific distributions such as Poisson or negative binomial. In addition, the model also handles observational biases such as adjustment for reporting delays (Sect. 10.1007/978-3-030-21923-9_7\#Sec20). The mean of the statistical model may be a function of time with unknown parameters. This function may involve convolution structures, such as back-calculation (Sect. 10.1007/978-3-030-21923-9_7\#Sec26), to connect predictions from a conceptual model to expected values of observable outcomes. The conceptual model is thus embedded inside a statistical model. However, this will inevitably result in statistical issues such as non-identifiability (Sect. 10.1007/978-3-030-21923-9_7\#Sec9). This section mainly discusses concepts, with a few examples as well as some simple methods where applicable. This is an important field that needs more research and development. Chapters 10.1007/978-3-030-21923-9_8 and 10.1007/978-3-030-21923-9_9 focus more heavily on applications, although some models not covered in Chaps. 10.1007/978-3-030-21923-9_5 and 10.1007/978-3-030-21923-9_6 are presented such as metapopulation spatial models and individual-based network models (Chap. 10.1007/978-3-030-21923-9_9). Examples presented are based on a case study for the 2016 Zika epidemic in Antioquia, Colombia (Sect. 10.1007/978-3-030-21923-9_8\#Sec8), a case study of the 2016 epidemic of yellow fever in two areas of Angola: Luanda (the capital) and Huambo (Sect. 10.1007/978-3-030-21923-9_8\#Sec22), and a case study of the 2014 Ebola outbreak in Mali (Sect. 10.1007/978-3-030-21923-9_9\#Sec5).

DARWIN'S "SPECIAL PROBLEM" ========================== Eusocial (Greek *eu*: "good/real" + "social") insects include the Hymenoptera (ants, bees, and wasps) and the Isoptera (termites). In honeybees, which is the focus of this perspective, a hive has caste differences; the diploid queen and the haploid drones are the sole reproducers, while the nurses, soldiers, guards, and foragers are "sub-castes" or "task groups" of the worker caste of sterile females that work together to benefit the group as a whole ([@B27]). However, as pointed out by [@B35] in a previous perspective in this journal, having sterile females in a colony is a potentially fatal flaw in Darwin's theory of natural selection, which states that the fittest individuals pass their traits (i.e., genes) to the next generation. [@B17] referred to sterile workers in insect communities as, "... one special difficulty, which at first appeared insuperable, and actually fatal to the whole theory." [@B18] later proposed a way around this "special difficulty" by proposing a "group selection" model for evolution of altruistic behaviors in eusocial insects. Darwin argued that "group selection" can occur when the benefits of altruism between castes are greater than the individual benefits of selfishness (egotism) within a subpopulation. [@B32], the great population geneticist Hamilton formalized the idea of group selection in a mathematical model, rb \> c, where b represents the benefit to the recipient of altruism, c the cost to the altruist, and r their degree of relatedness. Kin selection takes into account the genetic relatedness of individuals in a group, was a further refinement of the group selection theory. In kin selection, rb~k~ + b~e~ \> c, in which b~k~ is the altruistic benefit to kin and b~e~ is the altruistic benefit accruing to the group as a whole ([@B94]). However, group- and kin- selection models are mathematically complex and remain controversial amongst many evolutionary theorists, such [@B19] and [@B57], who argue that group selection is unlikely because, among many reasons, selfishness (i.e., "selfish genes" -- a phrase Dawkin's coined) would always predominate over altruism. Here, we further elaborate on [@B35] thesis that epigenetics might be a way around Darwin's "special difficulty." We argue that epigenetic inheritance systems (EISs) can allow rapid evolution of traits specific for sterile workers and fertile queens. Epigenetics does not involve changes in the DNA sequence, but rather covalent, yet reversible, changes to the DNA in the form of 5-methylcytosine (5mC). EISs should work fine for short-term evolutionary changes. However, natural selection of DNA sequence variants would still be needed for long-term evolutionary changes. Histone modifications, long non-coding RNAs, prions, and other types of EISs will not be discussed in detail, but rather we will focus on 5mC, since 5mC represents a reversible change to the genome that can be modified by the environment ([@B13]). Heritable changes in 5mC, such as occurs in imprinted genes in mammals, are also called metastable epialleles ([@B63]; [@B21]). The most important aspect of DNA methylation in the hypothesis presented in this paper is that, unlike histone modifications, 5mC is mutagenic and can lead to permanent changes to the DNA. Specifically, 5mC can undergo spontaneous deamination, which converts 5mC to T ([@B15]; [@B22]). A hypothesis for how natural selection of metastable epialleles can lead to DNA mutations that permanently stabilize the epialleles into real alleles, the epigenetic directed genetic error (EDGE) hypothesis, is presented in the last section of this perspective. The inspiration for many of the ideas in this perspective is a chapter in [@B39] excellent book "Evolution in Four Dimensions" on EISs. They created an imaginary planet named Jaynus where the variety of organisms all had exactly the same genome sequences, yet had many different phenotypes. They wrote: > Jaynus organisms have a genetic system that is based on DNA, and replication transcription, and translation are much the same as on Earth. However, there is one very extraordinary thing about the DNA of Jaynus creatures -- every organism has exactly the same DNA sequences. From the simplest organism, a tiny unicellular creature, to the enormous fanlike colonial worms, the DNA is identical. Their genomes are large and complex, but no organism deviates from the universal standard sequences because there are cellular systems that check DNA and destroy any cell suspected of carrying a mutation. In this perspective, we describe how intra-caste evolution is, in many respects, similar to how evolution proceeds on the mythical planet of Jaynus. MECHANISMS OF CASTE DETERMINATION AND EPIGENETIC MODIFICATION IN HONEYBEES ========================================================================== > *At the extreme superorganismic phase, the level of selection becomes the genome of the queen and the sperm she stores, and the workers can be viewed as robotic extensions of her phenotype ([@B95]*). As beautifully described in the above quote, eusocial insects are even more extreme in some respects than the mythical organisms on Jaynus because they have evolved to a "superorganismic" stage in which the queen is the reproductive organ(ism) and the workers are the "robotic extensions" or the somatic cells of the superorganism. [@B45] called the formation of eusocial insect colonies and super colonies "a new major transition in evolution." We propose that the intra-caste evolutionary process in honeybees might share epigenetic mechanisms with those proposed on Jaynus. [@B17] evolution utilizes the concepts of "survival of the fittest" in a population and "natural selection" of genetic variation to eventually form new species. Darwin, of course, did not know about either genetic or epigenetic variation, for his work pre-dated Gregor Mendel's discoveries (or, more accurately, rediscovery in the 20th century ([@B56]; [@B25]), but the modern interpretation of "natural selection" is selection of genetic variation. We propose that "intra-caste evolution" is a type of micro-evolution, which is small-scale evolution within a population, and refers to survival of the fittest members of a caste. We propose here that "intra-caste evolution" is initially based on natural selection of metastable epialleles of the most-fit caste (i.e., queen and worker) and sub-caste members (i.e., nurse, soldier, guard, and forager). As in the mythical Jaynus example, genetic selection probably cannot be the primary mechanism for selecting the fittest worker bee, since most worker bees cannot breed (however, see below). For example, the most efficient forager sub-caste of workers cannot be selected for by direct genetic selection, since workers, in most situations, are sterile females. However, a hive with more efficient foragers can be produced by group selection of metastabile epialleles that produce an increased foraging efficiency. One phenotype that worker bees have evolved to increase foraging efficiency are the pollen baskets on the hind legs, which are present on workers but not on queens. A possible mechanism for the evolution of pollen baskets is presented later in this perspective. If the "intra-caste evolution" hypothesis of honeybee castes is not mediated primarily by genetic means, then how are the desirable phenotypes, such as efficiency in foragers, transmitted to the next generation? We propose that, first, queens undergo a great deal of stress (i.e., malnutrition) when there is not an adequate amount of foraging being performed by the workers. The stress, in a mechanism that we present in a later section, leads to an activation of random stress-induced metastable epialleles, some of which increase the food-carrying capacity of pollen baskets. Second, the metastable epialleles which improve the fitness of the colony, such as those that serendipitously alter pollen baskets in workers in a manner that increases storage capacity, are selected over several generations by group selection. Third, random mutations can potentially be directed to the selected metastable epialleles by the EDGE mechanism, described in the final section of this perspective. The main reason for the need for the EDGE hypothesis is, we believe, because the "normal" background mutation and group selection processes are not adequate when the effective population size of a species is too low, as it arguably is in eusocial insects (i.e., only the queen breeds). The EDGE hypothesis provides an additional mechanism to increase the mutation rate of specific genes required to ensure the survival of the colony. CHEMICAL MEANS OF EPIGENETIC MODIFICATIONS IN HONEYBEE CASTES ============================================================= In addition to the selection of the most-fit caste and sub-caste members in each generation by group selection, honeybees have evolved to produce royal jelly to alter the epigenetic and developmental machinery of their offspring. The active ingredients of royal jelly include a fatty acid, (E)-10-hydroxy-2-decenoic acid (10HDA), which accounts for up to 5% of royal jelly. The fatty acid 10HDA, interestingly, is an epigenetic modifier molecule with a histone deacetylase inhibitor (HDACi) activity ([@B87]). HDACs remove acetyl groups from histones, which are present in actively transcribed genes to open up the chromatin, presumably by repellent ionic charges pushing the nucleosomes apart ([@B40]). HDACi's inhibit the deacetylation of histones, which would lead to the acetyl groups remaining on histones, and therefore transcriptional activity would be high in the "queen-specific genes" of larvae fed royal jelly. Another component of royal jelly is the protein royalactin, which increases body size and ovary development in queens ([@B44]). The mechanism of action of royalactin is thought to be multifold: activation of mitogen-activated protein kinase (MAPK), which decreases developmental time, activation of p70 S6 kinase, which increases body size, and increasing juvenile hormone production, which is an essential hormone for ovary development ([@B44]). Interestingly, the same paper also showed that royalactin dramatically increases body size and ovary development when fed to the fruit fly, *Drosophila melanogaster* ([@B44]). Based on the fact that royalactin has similar effects on the solitary fruit fly as on eusocial bees, we propose a theoretical epigenetic mechanism for how the queen's dependence on royal jelly for ovary development evolved. In our model, bees originally were solitary, like fruit flies, and every female fended for herself in terms of feeding and reproduction. However, when the food is in short supply, the absence of nutrients would lead to a reduction of reproductive fitness and a diminution in ovary development. Consequently, the population reaches a bottleneck when the food runs low, and only those few individuals that have sufficient nutrition survive. If the few survivors evolved the capacity to feed some of their offspring, which would be one of the first steps in eusocial evolution, then when the food runs too low, they can feed adequately only some of their offspring and leave the other offspring malnourished. The female offspring that are fed would develop ovaries, whereas the female offspring that were not sufficiently fed would develop atrophied ovaries and would be sterile. A decrease in reproductive fitness is a universal character of most animals during starvation ([@B10]). However, and this is key, both fertile and sterile offspring are produced by the same mother, in a manner that is dependent on how much food or the quality of food they were fed. If the sterile offspring provided a selective-advantage to the group as a whole, then those mothers that produced both fertile and sterile offspring would have a selective advantage over those mothers that produced only fertile offspring. After millions of years of fine-tuning this process, the honeybee sterile-fertile dimorphism could have theoretically evolved by group selection. As pointed out by an anonymous reviewer, there are at least three potential problems with our hypothesis on how honeybees evolved to produce royal jelly. First, the feeding behavior would need to be developed when food was scarce. Second, the sterile-fertile dimorphism would have to be maintained even though food became abundant again. Third, altruism would have to be developed when the sterile-fertile dimorphism emerged. It is hard to argue around these criticisms for a solitary insect such as *Drosophila*, and that might be why *Drosophila* and other solitary insects never evolved a sterile-fertile dimorphism. However, as suggested by [@B95], perhaps a way to circumvent all of these problems is the fact that the first step in eusocial evolution is probably the ability to form nests or colonies. This would allow the development of the dichotomy in bees, wasps, and ants of being a forager or staying in the nest to lay eggs. Since foraging is dangerous and taxing, if the workers are bringing the proto-queen pollen and nectar, then she is less inclined to forage for it. Once the proto-queen evolved the ability to produce royal jelly, then she would become the only fertile member of the colony -- all of the workers could be chemically sterilized by withholding royal jelly. The development of altruism, in this case, could be an emergent property of the sterile-fertile dimorphism. As discussed further in a later section, there are many examples of emergent behavior in eusocial insects ([@B41]), and we argue that altruism of sterile workers could be one of them. In addition to royal jelly, honeybees have evolved an arsenal of other chemical weapons that subvert developmental and behavioral processes in the workers. For example, after about 8 days post-emergence, the nurse bees who take care of the eggs will transition into foragers, and foraging is more metabolically taxing because it requires the filling of baskets on the hind legs to transport the pollen ([@B27]). However, the behavioral transition from nurse to forager depends on the needs of the hive, and individuals in the hive transmit these needs by both direct contact and by pheromones. The transition among sub-caste members in honeybees, but not some ant species that have physically different worker castes ([@B95]), is purely behavioral because all honeybee workers have pollen baskets despite the fact that only foragers use them -- pollen baskets do not develop *de novo* in the nurse when she transforms into a worker. Queen mandibular pheromone (QMP) is emitted by the queen to recruit nurses to her and to suppress ovary growth ([@B27]). The larvae emit brood pheromone (BP) to stimulate nurse bees to feed and care for them (i.e., the brood). BP affects the nurses and foragers in different manners: it stimulates nurses to care for the brood and to delay their transition into foragers, while it stimulates foragers to collect nutrient-rich pollen to feed the brood ([@B84]). Foragers, in turn, emit ethyl oleate (EO) to suppress nurse honeybees from foraging ([@B47]). Isoamyl acetate, which has a similar odor to the banana and pear, was found in [@B3] to be an active component in the sting pheromone of the honeybee which is presumably released by honeybee guards when a hive is disturbed. It is through such chemical (i.e., environmental) signals that the honeybees are able to epigenetically maintain the caste structure in a manner that circumvents, in most aspects, the need for the selection of genetic variation. THE POTENTIAL ROLE OF NATURAL SELECTION OF GENETIC AND EPIGENETIC VARIATION IN DRIVING THE EVOLUTION OF CASTE SYSTEMS ===================================================================================================================== As mentioned earlier, an important consideration regarding genetic-stabilization of the sterile-fertile dimorphism, is that the sterile workers can, in rare cases, develop ovaries. Removal of a queen can cause some workers to develop ovaries, in part because they are no longer exposed regularly to QMP ([@B36], [@B37]). Also, nurses and foragers can revert back-and-forth rapidly in either direction in a manner that is dependent on the needs of the hive ([@B2]). We believe that the occasional reversion of a sterile worker to a reproductive female is a critical mechanism for transmitting both metastable epialleles and genetic variation that is required for the worker caste. According to the EDGE hypothesis, genetic variation that is induced by the metastable epialleles, can be selected to increase worker specialization in the next generation. The purpose of the metastable epialleles, in the EDGE hypothesis, are not to circumvent the need for genetic variation, but rather to increase genetic variation in precisely the genes that need to be adapted for the organism, or superorganism in the case of eusocial insects, to survive the novel environment. The selection of metastable epialleles in *Drosophila* is well-established in our laboratory ([@B69], [@B73], [@B68]; [@B85]; [@B74]) and in other laboratories ([@B11]; [@B74]; [@B89]; [@B28]; [@B91]; [@B6]; [@B48]; [@B58]; [@B86]; [@B88]; [@B93]). We showed, for instance, that stress, or the inactivation of the chaperone protein Hsp90, can activate a metastable epiallele of the *Kruepple*^Incomplete\ facets-1^ (*Kr*^If-1^) allele, which causes ectopic large bristle outgrowths (ELBOs) to protrude from the eyes ([@B85]). We indicate the metastable epiallele with the nomenclature \[*Kr*^If-1^\]^∗^ and showed that the metastable epiallele can be transmitted through both the male and female germlines for tens or even hundreds of generations ([@B69], [@B73]). What makes a metastable epiallele an example of an epigenetic variant rather than a genetic variant is the fact that a metastable epiallele, such as \[*Kr*^If-1^\]^∗^, can be reverted back to the original allele, in this case *Kr*^If-1^, in just one or two generations by negative selection ([@B85]). Since *Drosophila* has very little DNA methylation, the metastable epialleles in *Drosophila* are probably not the result of differential DNA methylation. However, [@B28] presented evidence that the \[*Kr*^If-1^\]^∗^ metastable epiallele requires Piwi and Pi RNAs, which are small non-coding RNAs in the germline and function similarly to siRNAs and mi-RNA ([@B67]; [@B31]). We are still actively trying to determine the exact nature of the \[*Kr*^If-1^\]^∗^ metastable epiallele and how it is transmitted through both the male and female germlines. As discussed later, we believe that *Drosophila*, and more generally most or all Dipterans (flies) and Coleopterans (beetles), lost DNA methylation because the presence of 5mC would slow down the syncytial blastoderm mitotic cycles, which at ∼8 min are the fastest in the animal kingdom ([@B72]). There is no direct laboratory evidence that selection of metastable epialleles occurs in eusocial insects, such as honeybees. However, there are at least three indirect indications that metastable epialleles that utilize differential DNA methylation occur in eusocial insects. First, [@B36], [@B37]) showed that reverting foragers back to nurses reestablished the nurse-pattern of DNA methylation. This was the first evidence of reversible epigenetic changes associated with behavior. Second, [@B38] found that worker-biased proteins exhibited slower evolutionary rates than queen biased proteins or non-biased proteins. This is consistent with the idea that metastable epialleles must be transmitted through the germline, and the queen and fertile workers are the only females that produce eggs. Finally, as described in the next section, the bimodal distributions of CG content and/or DNA methylation in most insect genes suggests a role for differential DNA methylation and the existence of metastable epialleles in most insects. MECHANISMS OF EPIGENETIC MODIFICATION IN HONEYBEES ================================================== How might an EIS in honeybees and other organisms evolve? In order to understand this, it is necessary to describe the patterns of DNA methylation in mammals and honeybees (**Figures [1A,B](#F1){ref-type="fig"}**). In mammals, ∼60% of genes have so-called CpG islands in the promoter regions and 5′ regions, which are defined as regions of higher than average CG content. DNA methylation of CpG islands in mammals occurs primarily at CpG sites in somatic cells but often at CHH (where H = C, A or T) sites in stem cells (reviewed in [@B62]). The degree of CpG island methylation is inversely proportional to gene expression for most genes; i.e., highly expressed genes have little CpG island DNA methylation, whereas, low-expressed genes have large amounts of CpG island DNA methylation (**Figure [1A](#F1){ref-type="fig"}**). Two mechanisms that CpG island DNA methylation in mammals are thought to function to reduce gene expression are by inhibiting binding of some transcriptional activation factors, such as AP1, which binds to GC-rich consensus sequences, and by increasing the binding of transcriptional inhibitory factors, such as MeCP2, which recruits HDACs to inhibit transcription (reviewed in [@B42]). {#F1} The CpG island DNA methylation story is the most-well-known aspect of epigenetic regulation of transcription in mammals. However, several studies have shown that gene-body DNA methylation also occurs in a manner that is mostly proportional to gene expression in both mammals and insects ([@B46]; reviewed in [@B42]). In other words, highly expressed genes have the most gene-body DNA methylation, and this DNA methylation is mostly restricted to exon sequences (**Figure [1B](#F1){ref-type="fig"}**), but this is partly because exons, since they encode proteins, are CG rich compared to intronic and intergenic regions, which do not encode proteins. DNA methylation in mammals also occurs at repeat sequences, such as ALUs, SINES, LINES, and retroviruses, and this has been shown to prevent expression and, thereby, retro-transposition of the retroviruses to new genomic regions ([@B42]). Interestingly, DNA methylation in honeybees occurs primarily in gene bodies, particularly in exons (**Figure [1B](#F1){ref-type="fig"}**). However, in contrast to mammals, CpG islands are not apparent in the promoters of honeybee genes (i.e., there are very few genes with enriched CG-content in the promoter regions). Additionally, in the honeybee, little or no DNA methylation occurs in repeat or intergenic sequences ([@B52]; [@B98]; [@B12]). Therefore, in honeybees, DNA methylation is not thought to epigenetically regulate expression of genes by controlling transcription factor binding to promoter regions, but rather is a consequence of gene expression. Gene body methylation in honeybees likely improves the fidelity of gene expression by allowing transcription to initiate only at the promoter and not at intergenic regions. Gene body DNA methylation in plants, for instance, has been shown to suppress intragenic transcriptional start sites and anti-sense transcription, presumably by preventing transcriptional activation proteins from binding to the gene body and inappropriately activating transcription from cryptic promoters ([@B99]). Originally, it was reported that most DNA methylation occurs primarily in CpG sequences in honeybees ([@B52]; [@B98]; [@B12]). However, we have shown, by analyzing our own data, and by reanalyzing the data from [@B52], that there is actually more CHH DNA methylation in honeybees than CpG DNA methylation ([@B14]). The other laboratories that analyzed DNA methylation in the honeybee used software that removed most of the CHH DNA methylation, presumably because this type of DNA methylation occurs in less complex regions of the genome (i.e., CG poor) and are therefore harder to align to the reference genome. Also, multiple CHH methylation events in a single next-generation DNA sequencing (NGS) read are often, sometimes improperly, interpreted as poorly converted by bisulfite and thrown out. However, we validated that most of the CHH methylation events are real by alternative methods, such as sequencing honeybee genomic DNA after immunoprecipitation with anti-5mC antibodies, and enzymatic digestion of DNA at 5-hydroxymethylcytosine (5hmC) sites ([@B14]). We did confirm, however, like the other groups, that CpG DNA methylation is primarily in exons. Interestingly, we also found that CHH DNA methylation is primarily in introns, partly because introns are larger and have a lower CG content ([@B14]). We were also the first group to find significant amounts of 5hmC in bees ([@B14]). 5hmC is an oxidized form of cytosine, and is presumably produced by the honeybee ortholog to the ten-eleven-translocation (TET) protein, a dioxygenase that converts 5mC to 5hmC, and is involved in epigenetic reprogramming in mammals (reviewed in [@B13]). [@B96] recently confirmed the presence of 5hmC in honeybees and characterized the enzymatic function of the TET enzyme. Because of the uncertainty of whether 5hmC is a stable epigenetic mark, as some investigators believe (including us), or a transient DNA modification in the de-methylation pathway, as most investigators believe, we will not discuss 5hmC further in this review but will await future clarification on this topic. Genome sequencing the honeybee showed that it has an unusual genome structure that we believe facilitates the generation of metastable epialleles ([@B23]). In honeybees, there are two types of genes based on CG content in exons (**Figure [1C](#F1){ref-type="fig"}**). Highly expressed, so-called housekeeping genes, which are expressed in all cells, have a lower CG content than low-expressed genes. This bimodal distribution of CG content in genes, which are called isobars, was first observed by a bioinformatics analysis of the newly sequenced honeybee genome ([@B43]). The discovery of isobars in the honeybee genome was made prior to the mapping of the 5mC sites by whole-genome shotgun bisulfite sequencing by our laboratory and several other laboratories ([@B52]; [@B98]; [@B12]; [@B14]). Sodium bisulfite converts C to uracil (U) unless it is methylated (5mC), and whole genome shotgun bisulfite sequencing is used to map all of the 5mC sites in the genome ([@B97]). Interestingly, all of the groups that performed whole-genome shotgun bisulfite sequencing to map the 5mC distribution in honeybees found, at first impression paradoxically, that the low-CG content genes have much more DNA methylation than the high-CG content genes ([@B52]; [@B98]; [@B12]; **Figure [1C](#F1){ref-type="fig"}**). We additionally found that CHH DNA methylation is also greater in the low-CG content genes than in the high-CG content genes. We proposed, since there is [not]{.ul} a bimodal distribution of CHH sequences, that the same DNA methyltransferases (i.e., DNMT1 and DNMT3) methylate both CG and CHH sequences in a manner that is directly proportional to the level of gene expression ([@B14]). **Figure [2](#F2){ref-type="fig"}** shows an example of a high-CG content, low-5mC gene (Ubx, **Figure [2A](#F2){ref-type="fig"}**) and a low-CG content, high-5mC gene (Actin, **Figure [2B](#F2){ref-type="fig"}**). Both Ubx and Actin will be discussed as examples throughout this perspective. {#F2} To reiterate, low-CG content genes have more 5mC than high-CG content genes. This is counter-intuitive because it indicates that the greater the CG content, the less the DNA methylation, despite there being more cytosines (specifically, CpG sites) to methylate. However, high-CG content genes having low DNA methylation makes biological sense for the same reason that CpG islands (by definition, with high CG content) have low DNA methylation. The biological sense is based on the fact that 5mC has a much higher (up to 10-fold) mutation rate to thymidine (T) than non-methylated cytosine ([@B64]). Therefore, the more highly expressed genes would have more 5mC (**Figure [1B](#F1){ref-type="fig"}**), and, consequently, more of the cytosines would become thymidine. Consequently, in highly expressed genes, the CG-content would be expected to become lower-and-lower as more-and-more CGs are converted to TGs. The reason for the higher mutation rate of 5mC-to-T compared with C-to-T is that 5mC spontaneously deaminates at the 6-position to form T, which is a natural DNA base. However, unmethylated C deaminates to U, which is normally not present in DNA, and there are enzymes \[specifically uracil *N*-glycosylase (UNG)\] to remove the U bases in DNA ([@B64]). The diagrams in **Figures [1A,B](#F1){ref-type="fig"}** are a simplification for clarity purposes because the most highly expressed genes, which we will call "ultra-high," usually have less DNA methylation than the medium and highly expressed genes in the gene bodies in both insects and mammals. This might be because the ultra-high expressed genes may have lost so many of their CpGs that there are not enough remaining to allow them to enter the most highly methylated class -- in other words, the amount of DNA methylation that can occur in genes is saturated and peaks before it reaches equilibrium. The genetic code for certain amino acids and intra-exon RNA-splicing enhancers requiring CGs in their consensus sequencings are likely two additional reasons for retaining a few CGs in housekeeping genes. As discussed in the next section, differential gene-body methylation might be a contributing factor to the emergence of eusociality. However, the bimodal distribution of CG content seems to be less of a contributor to eusociality than the bimodal distribution of DNA methylation. Bees, wasps, and ants all have bimodal distributions in DNA methylation in genes, but only bees and wasps have a bimodal distribution in CG content. In all three eusocial insects -- bees, wasps and ants -- the highly expressed genes are generally more methylated than the low-expressed genes ([@B80]). [@B80] studied the evolution of gene-body DNA methylation in invertebrates and showed that silkworm (*Bombyx mori*), which has DNA methylation at appreciable levels in the genome, nevertheless does not have a bimodal peak of CG content in genes. This is similar to our finding of a unimodal peak of CHH sites but a bimodal peak of DNA methylation based on CG content, discussed earlier ([@B14]). Interestingly, the silkworm has a bimodal peak in DNA methylation levels similar to the honeybee, in which highly expressed genes have higher levels of DNA methylation in the gene body. The unimodal peak in CG content but bimodal peak in DNA methylation levels seen in the silkworm genome also occurs in all ant species studied so far ([@B29]; [@B5]; [@B4]). The bimodal peak in CG content in genes is not unique for the honeybee, however, because it is also seen in other invertebrates such as the sea anemone (*Nematostella vectensis*) and the sea squirt (*Ciona intestinalis*; [@B80]). We conclude that while bimodal peaks in CG content and DNA methylation might facilitate the formation of metastable epialleles, they are not essential for the generation of metastabile epialleles. In the next section, we explore the possibility that metastable epiallele hyper-mutability, a key component of the EDGE hypothesis, is an emergent property of bimodal levels of DNA methylation in eusocial insects. METASTABLE EPIALLELE HYPER-MUTABILITY MIGHT BE AN EMERGENT PROPERTY OF BIMODAL LEVELS OF DNA METHYLATION ======================================================================================================== > The movement from low-level rules to higher level sophistication is what we call emergence ([@B41]). The above quote is from [@B41] best-selling 2001 book, "Emergence: the connected lives of ants, brains, cities, and software." In the book, [@B41] describes how a simple behavior, such as an increasing number of ants following a weak-and-winding scent trail laid down by one ant to a food supply, can lead to a complex behavior, such as all of the ants following a direct path to the food. Eusocial insects show many other examples of bottom-up behavior where workers follow simple rules that emerge into complex hive behaviors ([@B41]). However, in contrast to human societies, there is little if any top--down behaviors in eusocial insects. For example, as mentioned above, the queen is best characterized as the "reproductive organ" in the hive and does little to influence the behaviors of the worker sub-castes ([@B41]), who themselves follow simple rules that are programmed into their genomes and epigenomes. We believe that the differential methylation of genes based on the level of gene expression is just such a simple rule that can lead to complex emergent phenomena, such as metastable epiallele hypermutability and, ultimately, eusociality. We hypothesize that an emergent property of low-expressed genes having low levels of DNA methylation is that they become more susceptible to epigenetic control, for the simple fact that they have more unmethylated cytosines. Highly expressed genes with high levels of DNA methylation can also potentially become metastable epialleles, but this would require differential de-methylation, such as by TET enzymes, in the germline cells after a stress response. In another review we presented a model for how oxidative stress can alter the function of the TET enzyme ([@B13]). However, what is the normal function(s) of gene body DNA methylation? In addition to preventing intragenic and antisense transcription within genes, mentioned above, one process that we and others have shown evidence to be regulated by gene body DNA methylation is alternative mRNA processing. For example, DNA methylation of cassette exons, at both CpG and CHH sites, correlates with their preferential exclusion in the mature mRNA ([@B52]; [@B14]; **Figure [1D](#F1){ref-type="fig"}**). Furthermore, [@B50] have shown that RNA interference (RNAi) knockdown of DNMT3a, the *de novo* DNA methlyltransferase, alters RNA splicing and causes intron retention in hundreds of genes in the honeybee fat bodies. How DNA methylation affects alternative mRNA splicing is not known in bees, but in mammals, DNA methylation inhibits the binding of the transcription factor CCCTC binding factor (CTCF), which affects alternative splicing ([@B83]). We speculate that there might be some biophysical processes involved too, since methylated DNA has a higher melting temperature (T~m~) than unmethylated DNA ([@B82]). Therefore, the increased T~m~ of methylated DNA might alter RNA polymerase translocation rates, cause pausing, and thereby affect the alternative mRNA splicing pattern. One interesting observation is that most insects, such as honeybees, have relatively large amounts of DNA methylation (but much less than mammals), but *Drosophila* has very little DNA methylation ([@B53]; [@B51]). The reason for the scarcity in DNA methylation in *Drosophila* is that *Drosophila* appears to have lost Dnmt1, the maintenance DNA methyltransferase, which methylates hemizygous DNA after replication, and Dnmt3, the *de novo* DNA methyltransferase, which methylates unmethylated DNA. The existence of DNA methylation in *Drosophila* is controversial because the only cytosine methyltransferase orthologs in *Drosophila* is a homolog to DNA methyltransferase 2 (MT2), but this enzyme was shown to methylate transfer-RNA-Asp (tRNA~Asp~) and presumably not DNA ([@B30]). However, the controversy appears to be resolved (at least to some in the field) by a recent paper that shows CHH methylation, albeit at very low levels, in *Drosophila* in a manner that is independent of MT2 ([@B9]). The authors were able to detect low levels of 5mC in *Drosophila* embryos in a two-step protocol of first immunoprecipitation of DNA with anti-5mC antibodies, followed by bisulfite sequencing of the immunoprecipitated DNA fragments ([@B9]). Our laboratory has similar evidence for low levels of 5mC in *Drosophila* and we speculate that it is generated non-enzymatically by spontaneous methylation of cytosines by intrinsic alkylation of DNA. We speculate that Dipterans (flies) and Coleopterans (beetles) lost DNA methyltransferases 1 and 3 because DNA methylation is redundant with histone modifications, such as H3K9me3 and H3K27me3, in repressing gene expression. Furthermore, we speculate that methylated DNA slows down DNA replication because of the higher melting temperature (T~m~) of methylated DNA compared with unmethylated DNA ([@B82]), which we mentioned earlier in the discussion of mRNA splicing. The predicted slowing down of DNA replication by DNA methylation is important in *Drosophila* because the first 10 syncytial nuclear divisions in the blastoderm embryo are in "hyper-drive" and are less than 8--10 min in duration (a world record, to our knowledge). Therefore, any process that slows down these rapid divisions would presumably be selected against because the faster-developing siblings would breed sooner ([@B72]). EPIGENETIC DIRECTED GENETIC ERRORS AND THE EVOLUTION OF CASTS IN HONEYBEES ========================================================================== Macroevolution requires selection of existing genetic variation to generate new species with greater fitness, but how does the sub-caste worker specialization increase when the effective population size of eusocial insects is so low (i.e., only one reproductive female per hive)? We mentioned group selection and kin-selection models at the beginning of this review, but they remain controversial in light of [@B19] "selfish gene" hypothesis. [@B19] argued that "selfish genes" that benefit the immediate survival and propagation of the "vessel" (the organism) would have much greater (and more immediate) selective advantage than altruistic genes that benefited the group. We speculate again, as we did in several other reviews, that one possible mechanism to facilitate genetic variation in the evolution of species is what we call the EDGE hypothesis ([@B66]; [@B70], [@B73]; [@B74]). In the simplest version of the EDGE hypothesis, the first step is the intra-caste selection of metastable epialleles that increase the specialization of a worker. The metastable epialleles could initially be generated by a stressful (i.e., non-optimal) environment, which would lead to a functional inactivation of Hsp90 ([@B77]), which is a chaperone for many chromatin remodeling proteins ([@B74]), including the Trithorax (Trx) protein ([@B89]; **Figures [3A,B](#F3){ref-type="fig"}**). Hsp90 has been called a "capacitor for morphological evolution" because many previously cryptic phenotypes are revealed when stress inactivates Hsp90 protein and this alters multiple signaling pathways ([@B79]; [@B54]; [@B76]; [@B77],[@B78]). The Trx protein, since it is a client for Hsp90, is an environmentally sensitive component of the Trx Group (TrxG) complex of proteins that is involved in maintaining transcriptional memory (i.e., activation) of the Hox genes, such as the Ultrabithorax (Ubx) gene during early embryogenesis in insects ([@B60]). One of the enzymatic functions of the TrxG complex is trimethylation of histone 3 at lysine 4 (H3K4me3), which is an activating mark for transcription ([@B40]). {ref-type="fig"}**. In the epigenetic directed error hypothesis (EDGE), the repair of base substitutions caused by methylated cytosines increases the mutation frequency of not only of the methylated cytosine but also neighboring bases. This could lead to an increase in the mutation frequency of genes with metastable epialleles, such as in the Ubx gene. This figure is modified from a previous review from our laboratory and we retain the copyright ([@B67]). **(B)** In stressed conditions, Hsp90 is inactive and cannot activate Trx, and transcription of Ubx is low. **(C)** Diagram of an empty and full pollen basket in forager bees. This diagram is used with permission from the Encyclopedia of Science, Copyright © The Worlds of David Darling (<http://www.daviddarling.info/>). **(D)** In queens, the pollen basket does not form because Ubx is low in T3. This causes an anterior transformation of the third thorax (T3) leg to look like the T2 leg. **(E)** In workers, the pollen basket forms because Ubx expression is high in T3. This figure represents a simplified representation of the homotic transformation that occurs when Ubx levels are reduced and are not meant to be accurate illustrations. This photograph is used with permission from Spike Walker, Wellcome Images, London (<http://wellcomelibrary.org/>).](fgene-06-00060-g003){#F3} When Hsp90 is inactivated in a stressful environment, Hox genes such as Ubx would have lower expression, presumably because there would be less H4K4me3 histone marks at the promoters. Since stress inactivates the Trx protein (**Figure [3B](#F3){ref-type="fig"}**), then stress would be expected to cause an increase in the DNA methylation status of the Ubx gene. The reason for this is that, in the absence of the Trx protein, the gene would no longer be in an activated state but switch to a repressed state by the Polycomb Group (PcG) repressor proteins ([@B61]). It is not known whether this occurs in bees, but in mammals genes that are initially repressed by PcG proteins are often further repressed by intragenic DNA methylation during cellular differentiation ([@B20]). We speculate that Ubx would have originally become a metastable epiallele in the proto-queen, who still has pollen baskets, because full pollen-baskets could immobilize her, and hence stress her, in the confines of the hive. In the EDGE hypothesis, the repair of base substitutions caused by methylated cytosines increases the mutation frequency of not only of the methylated cytosine, as mentioned above ([@B64]), but also neighboring bases because of error-prone DNA repair mechanisms ([@B66]; [@B70],[@B71]). This error-prone DNA repair could lead to an increase in the mutation frequency of genes with metastable epialleles, such as in the Ubx gene (**Figure [4D](#F4){ref-type="fig"}**). Through this "mutation-spreading" effect, the metastable epialleles could cause not only an increase in the mutation frequency of the exons, but also regulatory sequences in the adjacent promoters and introns. In other words, simply by becoming a metastable epiallele, the EDGE hypothesis predicts that the mutation frequency of a gene would increase. Fortuitously, genes with increased mutation frequencies are precisely those that need to be mutated to stabilize the metastable epialleles in a genetic manner. {ref-type="fig"}**. **(B)** Left, few 5mCs in the few CG s in high-expressed genes become mutated to TGs (red circles). Right, few 5mCs in the many CG s in poor-expressed stressed genes become mutated to TGs. **(C)** Left, mutations in high expressed genes get removed by purifying selection. Right, mutations in metastable epialleles remain because of low levels of purifying selection. **(D)** Mutations in bases near CG s (\*n) can be generated by error prone DNA repair of 5mC \> TG mutations.](fgene-06-00060-g004){#F4} How EDGE mutations generated in sterile workers are transmitted to the next generation in honeybees is a major issue that warrants discussion. One possible mechanism for transmitting the EGDE mutations to the next generation could be through honeybee workers who develop ovaries and become fertile after queen removal, as mentioned above ([@B26]). They could then directly transfer the mutations (as well as the metastable epialleles) to their offspring. Those mutations that are beneficial to the hive by stabilizing the metastable epialleles would have a selective advantage for the whole hive and would thereby be selected by group selection. An important consideration is that fertile-workers only have drone progeny (i.e., haploid males) and queens have both drone and worker progeny. This would necessitate that the metastable epialleles be transmitted through the male germline in the offspring of fertile workers. However, it is not clear whether worker-to-fertile-female conversions are frequent enough to explain the evolution of sterile-worker specializations. Another possible mechanism for the transfer of EGDE mutations to the next generation is that that the queen can transmit EDGE mutations to her offspring directly, without having to go through a worker-to-fertile-female conversion process. Metastable epialleles have to be in either the queen or the worker (by definition), but affect them in different manners. Therefore, the genes that become metastable epialleles would be predicted to have a higher mutation rate by the EDGE process. Some of these mutations would not affect the queen (and therefore reproduction), but might stabilize the metastable epialleles that affect the workers. Support for the EDGE hypothesis is the fact that queen-specific genes mutate faster than worker-specific genes ([@B38]; [@B34]) This makes sense since queens breed much more frequently than fertile workers, which, as mentioned above, only occur when the queen is removed from the colony ([@B26]). Queens would therefore have a greater opportunity to transmit both metastable epialleles, and mutations in the metastable epialleles, to the offspring, than the fertile workers. EPIGENETIC DIRECTED GENETIC ERRORS AND THE EVOLUTION OF POLLEN BASKETS IN HONEYBEE WORKERS ========================================================================================== A recent paper that we believe supports the EDGE hypothesis for intra-caste evolution in honeybees discusses the dimorphism in pollen basket formation in genetically similar queens and workers. This fascinating paper shows that the Hox gene Ubx, mentioned throughout this perspective, promotes pollen basket formation on the tibia of the hind legs in bees in the third thoracic segment (T3) ([@B55]). The pollen basket is a hollow indentation on the large and mostly bristle-free tibia segment that the forager bees use to store and transport impressive amounts of pollen (**Figure [3C](#F3){ref-type="fig"}**). In the queen, who does not collect pollen, the tibia is covered with hairs that would otherwise inhibit pollen collection. The investigators showed that reduction of Ubx levels in the workers by injecting inhibitory RNA (RNAi) into the worker embryos caused the hind legs to resemble that of the queens and become reduced in bristles ([@B55]). In *Drosophila*, mutations in Ubx, combined with other mutations in the bithorax complex (BXC), produced the famous four-winged fly that won Edward Lewis the 1995 Nobel Prize in Physiology and Medicine ([@B16]). Normally, *Drosophila* one pair of wings on the second thoracic segment (T2) and one pair of halteres (balancer organs that counteract the wing movement) on the third thoracic segment (T3). [@B49] explained the Ubx phenotype as causing an anterior homeotic transformation of T3 to T2, hence, the famous four-winged fly. Ubx has a conserved 60 amino acid homeobox (Hox) domain, which is nearly identical from *Drosophila* to humans, and a highly variable transcriptional regulatory domain. Hox genes, such as Ubx, not only regulate segmentation during embryogenesis, but they also affect subtle changes in limb, brain, and other organ development. [@B55] found that mutations in the Hox gene, Ubx, causes complex fate decisions in each segment of the honeybee T3 legs. A simplification of the results of the Ubx-RNAi experiments in honeybees is that the third leg has a partial homeotic transformation to the second leg by a similar T3-to-T2 homeotic transformation as seen in *Ubx*-mutant flies (**Figures [3D,E](#F3){ref-type="fig"}**). As mentioned earlier, the way the nurse honeybee controls pollen basket development in workers is by withholding royal jelly. In honeybees, the targets of Ubx are not known, but the authors speculated on what might be occurring in honeybees, based on what is known in the much better characterized *D. melanogaster* genetic system. In honeybee queens, when they are fed royal jelly as larvae, the HDACi activity in the royal jelly could possibly help in the activation of expression of the likely Ubx-target genes, such as grunge (gug) and Ataxin-2 (Atx2), which play a role in the formation of bristles in *Drosophila* ([@B24]; [@B1]). Consequently, the authors speculate, this might be one reason why the T3 tibia segments in queens have bristles in the area of the pollen basket, while workers do not ([@B55]). EPIGENETIC DIRECTED GENETIC ERRORS IN NON-CG DINCULEOTIDES IN METASTABLE EPIALLELES =================================================================================== In the EDGE hypothesis, we propose that methylated cytosines are mutagenic not only in the 5mC sites but also in the surrounding bases. The reason we propose this broader-range of mutagenicity is because error-prone DNA repair mechanisms can increase the mutation frequency of surrounding bases while repairing 5mC \> T base substitution mutations. Metastable epialleles, which have variable levels of 5mC, can occur in both somatic cells and germline cells, but they are generally referred to as simply "differentiated cells" when they occur in somatic cells. When metastable epialleles occur in somatic cells, they cannot be transmitted to the progeny. However, when a metastable epiallele occurs in a germline cell, then it can be transmitted to the progeny, as we and others have demonstrated in *Drosophila* ([@B85]; [@B89]). It is not yet known whether there is a bimodal distribution of 5mC in bee germline cells, but for the sake of argument, let's assume for this perspective that it is similar to what occurs in somatic cells -- i.e., housekeeping genes have low CG-content and high levels of 5mC and low-expressed genes have high CG-content and low levels of 5mC (**Figure [4A](#F4){ref-type="fig"}**). Therefore, housekeeping genes, such as Actin (**Figure [4A](#F4){ref-type="fig"}**, left) would never be metastable epialleles because their few CG s are always heavily methylated -- i.e., there can be no differential 5mC if it is always high. In contrast, low-expressed genes, such as Ubx, which is presumably not expressed at all in germline cells, would have high CG content but very little 5mC (**Figure [4A](#F4){ref-type="fig"}**, middle). We hypothesize that maternal stress can increase the DNA methylation in low-expressed genes, such as Ubx, and turn them into metastable epialleles (**Figure [4A](#F4){ref-type="fig"}**, right). This has not yet been demonstrated in any organism, but it should be possible to test this hypotheses in the laboratory once single-cell epigenomics techniques are further optimized ([@B90]). In housekeeping genes, such as Actin, there would still be expected to be an increase in mutations near and surrounding the 5mC sites. However, since there is a great deal of purifying selection in housekeeping genes, that would make any deleterious mutations in such important structural genes selected against (**Figures [4B,C](#F4){ref-type="fig"}**, left). Also, the 5mC rate in housekeeping genes is so high that there has probably been a maximum change in CG-to-TG sequences so that no further such mutations can occur without having deleterious structural or regulatory changes to the gene. In contrast, in low-expressed genes, such as Ubx, there would be mutations in CG-sites that do not undergo as much purifying selection (**Figures [4B,C](#F4){ref-type="fig"}**, right). As mentioned above, while the Ubx Hox domain is a 60 amino acid sequence that is almost absolutely conserved from *Drosophila* to humans ([@B81]), the remaining amino acids, such as in the transcriptional regulatory domains, are amongst the most variable sequences in proteins ([@B75]; [@B65]). In the Ubx-mutagenesis hypothesis for basket formation in honeybees, several questions arose during review of this manuscript. First, "How did Ubx changed its biological role without affecting fitness?" Second, "Is it possible that Ubx regulates both body plan and caste differentiation in honeybee but not in solitary insects?" Third, "Could some other gene(s) compensate for the supposed "functional loss" of Ubx in honeybee?" To answer these questions, we do not believe that the mutations in Ubx would necessarily affect fitness by causing a "functional loss." Rather, we believe that the mutations in Ubx were most likely regulatory mutations in the promoter and introns and they represent a functional gain rather than a functional loss. Developmental genes have large and complex regulatory regions, such as individual enhancers for each of the eight stripes in segmentation genes such as fushi tarazu ([@B59]). The Hox genes in *Drosophila*, such as Ubx and Antennapedia (Antp), have enhancer regions 10s or even 100s of kilobases from the promoter regions ([@B8]; [@B7]). The Ubx gene in *Drosophila* has a complex array of alternative spliced products and an unusual mechanism for splicing the 74 kb intron that involves multiple steps of re-splicing the intron ([@B33]). This re-splicing mechanism avoids competition between distant splice sites and allows removal of the 74 kb intron as a series of smaller RNA fragments ([@B33]). The diverse array of transcriptional and RNA splicing regulatory sequences should allow Ubx to evolve multiple additional roles in caste formation without the need for other genes to compensate for it proposed "functional loss." CONCLUSION ========== We propose an intra-caste model of evolution that is based on selection of metastable epialleles in worker bees that runs parallel to the macro-evolution and group selection of DNA mutations. Like the mythical world of Jaynus, the evolution of the most-fit sub-caste members occurs through the selection of metastable epialleles by group selection. However, our EDGE hypothesis expands upon the limited world of Jaynus, in which all of the organisms have exactly the same sequence, by proposing a mechanism to direct mutations to the metastable epialleles that were selected. These directed mutations can, in turn, stabilize and increase the penetrance of the metastable epialleles in future generations of superorganism colonies. Waddington, who is often considered the father of epigenetics, proposed a mechanism similar to the EDGE hypothesis in [@B92] for the inheritance of acquired characteristics that were induced by stress. In follow-up experiments, in response to Waddington, we provide a possible epigenetic mechanism for how stress can reveal previously cryptic phenotypic information by the inactivation of Hsp90 ([@B69]; [@B85]). Finally, our EDGE hypothesis presented here provides a possible mechanism for the stabilization of metastable epialleles, thereby allowing the evolution of castes and sub-castes in eusocial insects. Conflict of Interest Statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. This research was supported by R01 ES012933 and R21 ES021893 and the WSU-NIEHS Center (P30 ES020957). We thank Greg Hunt for excellent editorial assistance and much needed expertise on eusocial insects. DMR worked as an undergraduate in Dr. Edward B. Lewis' laboratory in 1980--1981 dedicates this review to his memory. [^1]: Edited by: *Greg J. Hunt, Purdue University, USA* [^2]: Reviewed by: *Aaron Arthur Comeault, University of Sheffield, UK; Feng-Chi Chen, National Health Research Institutes, Taiwan* [^3]: This article was submitted to Evolutionary and Population Genetics, a section of the journal Frontiers in Genetics.

Background {#Sec1} ========== Myocardial infarction (MI) is the leading cause of non-cancer mortality worldwide \[[@CR1]\]. Timely reperfusion approaches, including coronary artery bypass grafting (CABG) and percutaneous coronary intervention (PCI), are standard post-MI treatments, but the limited ability of cardiomyocytes to regenerate damaged myocardial tissue results in the development of cardiac dysfunction \[[@CR2]\]. Mechanistically, post-infarction heart remodeling and/or chronic cardiac damage result from cardiomyocyte death and subsequent cardiac fibrosis \[[@CR3], [@CR4]\]. First, MI causes excessive cardiomyocyte death via apoptosis or necrosis, leading to a decline in the number of functional cells over a short period \[[@CR5], [@CR6]\]. MI also initiates the chronic inflammatory response that increases the rate of cardiomyocyte death \[[@CR7], [@CR8]\]. In addition, MI activates cardiac fibroblasts, which produce excessive collagen and promote extracellular matrix accumulation \[[@CR9]\]. The decreased number of functional cardiomyocytes and the increased amount of non-contractile tissue components collectively contribute to the progression of cardiac decompensation \[[@CR10]\]. Therefore, a means to reduce cardiomyocyte death and alleviate cardiac fibrosis is essential to retard or delay the onset of post-infarction cardiac injury. Mitochondria are the energy center of cardiomyocytes \[[@CR11]\]. An increasing number of studies have identified mitochondrial damage as the pathogenesis for cardiomyocyte death and heart dysfunction. If the cardiomyocytes are deprived of energy due to mitochondrial injury, well-coordinated contraction cannot be guaranteed. Mitochondrial damage is accompanied by cellular oxidative stress or calcium overload, which lead to cardiomyocyte dysfunction \[[@CR12]--[@CR14]\]. Extensive mitochondrial stress is closely associated with cardiomyocyte apoptosis due to the liberation of pro-apoptotic factors (such as cyt-c) from the mitochondria into the cytoplasm \[[@CR15]\]. For example, cyt-c interacts with and activates the apoptotic executor, caspase-3 \[[@CR16], [@CR17]\]. With its regulatory role in cardiac dysfunction and cell death, mitochondrial damage is considered the primary target for controlling the development of post-infarction cardiac injury. Interestingly, recent studies have found that mitochondrial damage in cardiomyocytes is mainly triggered by mitochondrial fission in response to MI \[[@CR15]\]. Due to the coronary obstruction and nutrition shortage, mitochondria divide into several sister mitochondria to meet the energy requirement \[[@CR15], [@CR18]\]. However, excessive mitochondrial fission causes an uneven distribution of mitochondrial DNA in the sister mitochondria, most of which cannot then produce ATP but aggravate the cellular damage via multiple mechanisms \[[@CR19]\]. This shows that inhibiting mitochondrial fission is necessary to sustain mitochondrial homeostasis and promote cardiomyocyte survival in the context of post-infarction cardiac injury. Mammalian STE20-like kinase 1, a key component of the Hippo pathways \[[@CR20]\], has received considerable attention as a regulator of acute and chronic cardiac injuries, such as ischemia reperfusion (IR) injury, diabetic cardiomyopathy and myocardial hypertrophy \[[@CR21]--[@CR23]\]. Convincing experimental data show the harmful effect of Mst1 on cardiovascular disease. Mst1 has also been reported to promote cardiac fibrosis and cardiomyocyte death in cases of post-infarction cardiac injury \[[@CR22]\]. However, the mechanism underlying these effects remains unclear. More importantly, there is an incomplete understanding of whether Mst1 acts via regulation of mitochondrial homeostasis in post-infarction cardiac injury. The aim of our study is to explore the role of Mst1 in repairing the infarcted heart, with a focus on mitochondrial fission. Methods {#Sec2} ======= Myocardial infarction model {#Sec3} --------------------------- Wild-type (WT) and Mst1-knockout mice (Mst1^KO^) mice with a C57BL/6 background were purchased from K&D Gene Technology (WuHan) based on information from a previous study \[[@CR22]\]. The mice were 12 weeks old and were housed under standard laboratory conditions (27 °C, 40--60% humidity, a 12-h light and dark cycle) with fresh drinking water and a commercial pellet diet. The MI mouse model was created by passing a 7--0 silk suture underneath the left anterior descending coronary artery with a knot, as described in a previous study \[[@CR4]\]. After 28 days, the hearts were isolated. Chronic hypoxia model in vitro {#Sec4} ------------------------------ Chronic hypoxia was applied to cardiomyocytes in vitro to mimic chronic post-infarction cardiac injury. This treatment was previously reported to be effective in imitating post-infarction injury in vitro \[[@CR3]\]. Cardiomyocytes were isolated from the WT and Mst1^KO^ mice with trypsin and collagenase as described previously \[[@CR11]\]. The cardiomyocytes were placed into a hypoxic incubator (95% N~2~ and 5% CO~2~) at 37 °C for approximately 48 h. Sample preparation and histological analysis {#Sec5} -------------------------------------------- The hearts were excised and rapidly frozen in Optimal Cutting Temperature medium at room temperature (Agar Scientific Ltd.) for the preparation of 4-μm thick frozen sections. Masson trichrome staining was performed at room temperature and the sections were observed with an inverted microscope (magnification, 40×; BX51; Olympus Corp.) \[[@CR24]\]. The levels of lactate dehydrogenase (LDH), troponin T, creatine kinase-MB (CK-MB), laminin and precollagen III in the blood were measured with ELISA assays as previously described \[[@CR25]\]. Echocardiography {#Sec6} ---------------- Cardiac function was evaluated with an echocardiograph as previously described \[[@CR26]\]. The mice were anesthetized using 3% isoflurane inhalation and studied using a Sequoia Acuson 15-MHz linear transducer echocardiography system (Siemens). The left ventricular ejection fraction (LVEF), left ventricular fraction shortening (LVFS), E/A ratio and left ventricular volume in systole (LV vol-s) were calculated using computer algorithms. Electron microscopy {#Sec7} ------------------- Tissues were fixed at 4 °C with 2% glutaraldehyde in 0.1 mol/l sodium cacodylate buffer and post-fixed for 1 h on ice with 1% osmium tetroxide. The 60-nm sections were rinsed with distilled water and dehydrated using acetonitrile and graded methanol (50%, 20 min; 70%, 20 min; 95%, 20 min; and 100%, 20 min), and then embedded in epoxy resin (EMbed-812; Electron Microscopy Sciences) and polymerized at 70 °C overnight \[[@CR27]\]. The sections were 60 nm thick and stained with lead citrate and uranyl acetate. The samples were imaged using a Hitachi H600 Electron Microscope. At least 30 cells in 5 randomly selected fields were observed \[[@CR28]\]. Immunofluorescence staining {#Sec8} --------------------------- The samples were first washed with cold PBS and then permeabilized in 0.1% Triton X-100 for 10 min at 4 °C. Then, 10% goat serum albumin (Invitrogen) was used to block the samples for 1 h at room temperature. The samples were incubated with primary antibodies overnight at 4 °C \[[@CR29]\]. After three rinses in PBS, secondary antibodies were added to the samples for 1 h at room temperature \[[@CR30]\]. The primary antibodies were: mitochondrial import receptor subunit TOM20 homolog (Abcam; cat. no. ab78547), F4/80 (1:1000, Abcam, \#ab111101), troponin T (1:1000, Abcam, \#ab8295), ICAM1 (1:1000, Abcam, \#ab119871) and cyt-c (1:1000, Abcam, \#ab ab133504). Images were observed with an inverted microscope (magnification, 40×; BX51; Olympus Corp.). Western blotting {#Sec9} ---------------- Total protein (40--60 μg) was loaded onto a 12--15% SDS-PAGE gel. After electrophoresis, the proteins were transferred to a PVDF membrane (Roche Applied Science) \[[@CR31]\]. Bands were detected using an enhanced chemiluminescence substrate (Applygen Technologies, Inc.). Band intensities were normalized to the respective internal standard signal intensity (β-actin, 1:2000; Abcam; cat. no. ab8224) \[[@CR25]\]. The primary antibodies were: Mst1 (1:1000, Cell Signaling Technology, \#3682), Bax (1:1000; Abcam; \#ab32503), Bcl2 (1:1000, Cell Signaling Technology, \#3498), Bad (1:1000; Abcam; \#ab32455), caspase-9 (1:1000, Cell Signaling Technology, \#9504), survivin (1:1000, Cell Signaling Technology, \#2808), Mst1 (1:1000, Abcam, \#ab184154), JNK (1:1000; Cell Signaling Technology, \#4672), p-JNK (1:1000; Cell Signaling Technology, \#9251), pro-caspase-3 (1:1000; Cell Signaling Technology, \#9662), cleaved caspase-3 (1:1000; Cell Signaling Technology, \#9664), Drp1 (1:1000, Abcam, \#ab56788), TGFβ (1:1000, Abcam, \#ab92486), MMP9 (1:1000; Cell Signaling Technology, \#3852). RNA isolation and qPCR {#Sec10} ---------------------- Trizol reagent (Invitrogen) was used to isolate total RNA. The Eurogentec Reverse Transcription Kit was applied to transcribe RNA (one μg in each group) into cDNA. Quantitative PCR was performed with primers and matched probes from the Roche Universal Fluorescence-labeled Probe Library. The primers were: TNFα (forward, 5'-AGATGGAGCAACCTAAGGTC-3′; reverse, 5'-GCAGACCTCGCTGTTCTAGC-3′), IL6 (forward, 5'-CAGACTCGCGCCTCTAAGGAGT-3′; reverse, 5'-GATAGCCGATCCGTCGAA-3′), MCP1 (forward, 5'-GGATGGATTGCACAGCCATT-3′; reverse, 5'-GCGCCGACTCAGAGGTGT-3′) \[[@CR2]\]. Cellular ROS {#Sec11} ------------ To observe the cellular ROS levels, the ROS probe (DHE, Molecular Probes) was incubated with the cells for 30 min at 37 °C in the dark \[[@CR32]\]. The cells were then washed with PBS to remove the ROS probe and immediately analyzed under a fluorescence microscope \[[@CR33]\]. mPTP opening assay, JC-1 staining and ATP production {#Sec12} ---------------------------------------------------- mPTP opening is an early event in mitochondrial apoptosis. In our study, mPTP opening was measured via tetramethylrhodamine ethyl ester fluorescence. Samples were washed with PBS three times and then were loaded with tetramethylrhodamine ethyl ester (TEE). The fluorescence of TEE was recorded at the start and after 30 min. As detailed in a previous study \[[@CR34]\], the mPTP opening rate was determined based on the time taken for the fluorescence intensity to decrease to half of the baseline. Mitochondrial potential was assessed using a JC-1 probe, which is a sensitive fluorescent dye used to detect alterations in mitochondrial potential \[[@CR11]\]. Following treatment, cells were incubated with 10 mg/ml JC-1 for 10 min at 37 °C in the dark and monitored with a fluorescence microscope (magnification, 100×; BX51; Olympus Corp.) \[[@CR33]\]. Red-orange fluorescence was attributable to potential-dependent dye aggregation in the mitochondria. Green fluorescence, reflecting the monomeric form of JC-1, appeared in the cytosol following mitochondrial membrane depolarization \[[@CR35]\]. ATP production was detected to reflect mitochondrial function. The samples were washed with cold PBS three times. Then the samples were lysed and a luciferase-based ATP assay kit (Beyotime Institute of Biotechnology) was used. ATP production was measured using a microplate reader \[[@CR13]\]. LDH assay and caspase-3 and -9 activity detection {#Sec13} ------------------------------------------------- LDH is released into the medium when cellular membranes rupture. To evaluate the LDH level in the medium, an LDH Release Detection kit (Beyotime Institute of Biotechnology) was used. To analyze changes in caspase-3 and caspase-9, caspase-3/− 9 activity kits (Beyotime Institute of Biotechnology) were used according to the manufacturer's protocols \[[@CR36]\]. To analyze caspase-3 activity, 5 μl of DEVD-p-NA substrate (4 mM, 200 μM final concentration) was added to the samples for 2 h at 37 °C. To measure caspase-9 activity, 5 μl of LEHD-p-NA substrate (4 mM, 200 μM final concentration) was added to the samples for 1 h at 37 °C. The wavelength at 400 nm was recorded using a microplate reader to reflect the caspase-3 and caspase-9 activities \[[@CR37]\]. MTT and TUNEL assays {#Sec14} -------------------- MTT experiments were performed in 96-well plates. Samples were washed 3 times with PBS and 50 μl of MTT reagent was added to each well. The samples were subsequently incubated for 4 h at 37 °C in a humid atmosphere containing 5% CO~2~. The MTT solution was removed, 200 μl of dimethyl sulfoxide was added to each sample, and the samples were incubated for 10 min \[[@CR38], [@CR39]\]. Following the addition of Sorensen's buffer, the absorbance at the wavelength of 570 nm was determined. To detect DNA fragmentation in the cell nuclei (a marker of apoptosis in testicular tissue), a TUNEL assay was performed using an In Situ Cell Death Detection kit (Roche Diagnostics GmbH) according to the manufacturer's protocol. DAPI was used to label the nuclei (at room temperature for approximately 30 min) \[[@CR40]\]. Statistical analysis {#Sec15} -------------------- All data are expressed as the means ± standard deviation. Statistical analyses were performed with SPSS software (version 17.0; SPSS, Inc., Chicago, IL, USA). The results from more than two groups were evaluated via one-way analysis of variance with the least significant difference test. *p* \< 0.05 was considered statistically significant. Results {#Sec16} ======= Loss of Mst1 in post-infarcted hearts reduces cardiac fibrosis {#Sec17} -------------------------------------------------------------- Western blotting was used to observe the change in Mst1 expression after myocardial infarction (MI). As shown in Fig. [1a and b](#Fig1){ref-type="fig"}, Mst1 expression was significantly upregulated 28 days after MI when compared to the results for the sham group myocardium.Fig. 1Mst1 levels were higher in the myocardium after a myocardial infarction (MI) and this contributed to the chronic cardiac damage. **a** and **b** The expression of Mst1 in the post-MI myocardium (Post-MI) in wild-type (WT) and Mst1-knockout (Mst1^KO^) cells relative to the control (Sham). **c**--**f** The cardiac function was detected as LVEF, LVFS, E/A ratio and LV vol-s via echocardiography in WT and Mst1^KO^ cells. **g** Masson staining was used to observe cardiac fibrosis. **h**--**j** The signaling pathways related to cardiac fibrosis (TGFβ and MMP9) were assessed via western blotting in WT and Mst1^KO^ cells. **k** and **l** The serum laminin and precollagen III levels in WT and Mst1^KO^ cells were measured via ELISA. **m** Electron microscope observations of the ultra-structural alterations in WT and Mst1^KO^ mice after MI. Red arrows indicated Z-line disappearance, cardiac muscle dissolution and cardiomyocyte disorganization. \**p* \< 0.05 vs. sham group, ^\#^*p* \< 0.05 vs. WT mice in post-MI group We then asked whether an increased Mst1 level plays a causal role in cardiac dysfunction after MI. Mst1^KO^ mice were used for this set of experiments. Cardiac function was analyzed via echocardiography. We demonstrated that cardiac contractile function (LVEF and LVFS) was better in Mst1^KO^ mice than in the WT mice (Fig. [1c and d](#Fig1){ref-type="fig"}). Moreover, cardiac diastolic function (E/A ratio and LV volume) was also better in the Mst1^KO^ mice than in the WT mice (Fig. [1e and f](#Fig1){ref-type="fig"}). This information shows that Mst1 was activated by myocardial infarction and contributed to cardiac dysfunction. Cardiac fibrosis is a key feature of infarcted hearts. Interestingly, Mst1 knockout reduced the level of cardiac fibrosis, as determined via Masson staining (Fig. [1g](#Fig1){ref-type="fig"}), showing that Mst1 is the cause of cardiac fibrosis. TGFβ and MMP9 expressions were higher in the heart after MI, but these phenotypic alterations were prevented by Mst1 knockout/were alleviated when Mst1 levels were reduced (Fig. [1h--j](#Fig1){ref-type="fig"}). The serum laminin and precollagen III concentrations (Fig. [1k--l](#Fig1){ref-type="fig"}) were notably increased in the samples from post-MI mice. Mst1 knockout reduced fibrosis marker levels. To provide more solid evidence for Mst1-mediated cardiac damage, electron microscopy (EM) was used to observe the ultrastructural changes in the myocardium. The post-infarcted hearts showed Z-line disappearance, cardiac muscle dissolution and cardiomyocyte disorganization (Fig. [1m](#Fig1){ref-type="fig"}). These conformational alterations were recused by Mst1 deletion. Ablation of Mst1 expression alleviates cardiac inflammatory injury {#Sec18} ------------------------------------------------------------------ The cardiac inflammatory response is also involved in the development of cardiac dysfunction in post-MI hearts. We demonstrated that the post-infarcted myocardium expressed more ICAM1 (Fig. [2a and b](#Fig2){ref-type="fig"}), which a kind of adherence factor that could capture the inflammation cells. By contrast, deletion of Mst1 reduced ICAM1 expression.Fig. 2Mst1 regulated the inflammation response in the myocardium after a myocardial infarction (MI). **a** and **b** The expression of ICAM1 in myocardial tissue. **c** and **d** F4/80-positive macrophage migration into myocardial tissue was observed in wild-type (WT) cells post-MI using immunofluorescence assay. **e**--**g** The transcriptional alterations of inflammation factors (TNFα, IL6 and MCP1) in the myocardium post-MI. **h**--**j** The cardiac damage markers LDH, CK-MB and troponin-T were assessed in WT and Mst1-knockout (Mst1^KO^) cells via ELISA. \*p \< 0.05 vs. sham group, ^\#^p \< 0.05 vs. WT mice in post-MI group Through immunofluorescence assays of F4/80 macrophages, we demonstrated that more macrophages permeated the myocardial tissue of mice with post-MI hearts (Fig. [2c and d](#Fig2){ref-type="fig"}), indicating an excessive inflammatory response in the damaged heart. Mst1 knockout repressed the macrophage infiltration. As a result of cardiac inflammation, the IL6, TNFα and MCP1 transcription levels were also higher in post-infarcted myocardium (Fig. [2e--g](#Fig2){ref-type="fig"}), and this effect was inhibited by Mst1 knockout. This information indicates that Mst1 dysregulation is involved in the increased inflammatory response in post-MI hearts. Excessive inflammatory injury induces cardiomyocyte injury, so we evaluated cardiac damage markers. Lactate dehydrogenase (LDH), troponin T and creatine kinase-MB (CK-MB) levels were significantly higher in the post-MI hearts than in the sham group hearts (Fig. [2h--j](#Fig2){ref-type="fig"}). However, Mst1 knockout reduced LDH, troponin T and CK-MB levels. These data indicate that Mst1 deletion is associated with a reduced inflammatory injury in post-MI hearts. Loss of Mst1 inhibits cardiomyocyte death {#Sec19} ----------------------------------------- Cardiomyocyte death is the primary risk factor in the progression of post-infarction cardiac injury. Through TUNEL assays, we found that post-MI hearts contained an higher number of TUNEL-positive cells (Fig. [3a and b](#Fig3){ref-type="fig"}). However, Mst1 knockout reduced the cardiomyocyte death ratio. Caspase-3, caspase-9, Bad and Bax expression levels were also higher in the post-MI heart (Fig. [3c--g](#Fig3){ref-type="fig"}). The levels of anti-apoptotic proteins such as Bcl2 and Survivin were lower (Fig. [3h and i](#Fig3){ref-type="fig"}). Interestingly, Mst1 knockout upregulated the level of anti-apoptotic factors and downregulated the expression of pro-apoptotic proteins in the post-MI heart (Fig. [3c--i](#Fig3){ref-type="fig"}).Fig. 3Mst1 promoted cardiomyocyte death in the heart after a myocardial infarction (MI). **a** and **b** The TUNEL assay was used to observe cellular apoptosis in vivo in wild-type (WT) and Mst1-knockout (Mst1^KO^) cells. **c**--**i** The mitochondrial apoptotic proteins pro-caspase-3, cleaved caspase-3 (Cle.caspase-3), caspase-9, Bax, Bad, Bcl2, survivin and β-actin were measured via western blots. Mst1 elevated mitochondrial apoptotic protein levels in the post-infarcted heart. **j** and **k** Cardiomyocytes were isolated from WT and Mst1^KO^ mice and a hypoxia model was used to induce cardiomyocyte damage in vitro. The MTT and caspase-3 activity assays were used to assess cellular viability \**p* \< 0.05 vs. sham group, ^\#^*p* \< 0.05 vs. WT mice in post-MI group To provide more solid evidence for the anti-apoptotic role of Mst1 deletion, cardiomyocytes were isolated from WT (WT cells) and Mst1^KO^ mice (Mst1^KO^ cells) and subjected to chronic hypoxia stress in vitro. Cell viability was detected via the MTT and caspase-3 activity assays. Compared to the control group, the chronic hypoxia group presented with reduced cell viability (Fig. [3j and k](#Fig3){ref-type="fig"}). This effect was reversed by Mst1 knockout. These data indicate that Mst1 deficiency attenuates cardiomyocyte death in the post-MI heart. Mst1 deletion sustains mitochondrial homeostasis {#Sec20} ------------------------------------------------ Mitochondrial damage is tightly linked to the cardiomyocyte death that occurs, at least in part, through cellular ROS release, mitochondrial potential collapse, mPTP opening and mitochondrial pro-apoptotic factor leakage. Through cellular ROS staining, we confirmed that chronic hypoxia stimulation increased ROS production in WT cells but not in Mst1^KO^ cells (Fig. [4a and b](#Fig4){ref-type="fig"}). The released ROS attack the mitochondrial membrane, leading to a reduction in mitochondrial potential.Fig. 4Mst1 impacted mitochondrial function and structure. **a** and **b** ROS production was higher in response to hypoxia treatment in wild-type cells (WT-cell) but was lower with Mst1 knockout (Mst1^KO^-cell). **c** and **d** The mitochondrial potential was measured via JC-1 staining. **e** The mPTP opening rate was evaluated and found to be highest in WT cells under conditions of hypoxia. Mst1 knockout meant mPTP opening rates closer to the normal level. **f** and **g** The immunofluorescence assay of cyt-c translocation from the cytoplasm (green staining) into the nucleus (blue staining). **h** Caspase-9 activity was measured to reflect mitochondrial apoptosis. \**p* \< 0.05 vs. ctrl group, ^\#^*p* \< 0.05 vs. WT-cell+hypoxia Through JC-1 staining, we demonstrated that WT cells exhibited a dissipated mitochondrial potential in response to chronic hypoxia condition, as evidenced by a higher green fluorescence and lower red fluorescence (Fig. [4c and d](#Fig4){ref-type="fig"}). By contrast, Mst1^KO^ cells sustained their mitochondrial potential under chronic hypoxia conditions. We also observed an increase in the mPTP opening rate in hypoxia-treated cells compared to that for the control group, and this tendency was reversed in Mst1^KO^ cells (Fig. [4e](#Fig4){ref-type="fig"}). Notably, mPTP opening provides a channel to facilitate the leakage of mitochondrial pro-apoptotic factors, such as cyt-c, into the cytoplasm. Through immunofluorescence assays, we illustrated that cyt-c diffusion into the nucleus increased after the hypoxia treatment, and that Mst1 knockout limited cyt-c migration (Fig. [4f and g](#Fig4){ref-type="fig"}). As a consequence, cytoplasmic cyt-c interacted with caspase-9 and increased caspase-9 activity, leading to caspase family activation and mitochondrial apoptosis. Through the analysis of caspase-9 activity, we uncovered that hypoxia stimulation enhanced caspase-9 activity (Fig. [4h](#Fig4){ref-type="fig"}). This effect was nullified by Mst1 knockout. Our data reveal that Mst1 deficiency protected cardiomyocytes against chronic hypoxia injury by maintaining mitochondrial homeostasis. Mst1 knockout abates excessive Drp1-mediated mitochondrial fission {#Sec21} ------------------------------------------------------------------ Recent studies have reported that mitochondrial fission occurs during the early stage of mitochondrial apoptosis. Based on this, we asked whether mitochondrial fission is involved in cardiomyocyte mitochondrial apoptosis under chronic hypoxic stress. First, mitochondrial morphology was observed via confocal microscopy. As shown in Fig. [5a](#Fig5){ref-type="fig"}, compared to the control cardiomyocytes, the hypoxia-treated cardiomyocytes contained more punctate mitochondria that were smaller (significantly shorter long axes). However, most of the mitochondria in the Mst1^KO^ cells exhibited a long filamentous morphology, indicating the probable ability that Mst1 deletion suppresses hypoxia-induced mitochondrial fission. The average mitochondrial length was recorded to quantify the mitochondrial fission (Fig. [5b](#Fig5){ref-type="fig"}) and found to be 8.2 ± 1.3 μm in the control group, 1.9 ± 0.6 μm in WT cells after hypoxia treatment, and 7.8 ± 0.9 μm in Mst1^KO^ cells after hypoxia treatment. This suggests that mitochondrial fission was activated by hypoxia via Mst1 in cardiomyocytes.Fig. 5Mst1 controlled cellular apoptosis via mitochondrial fission. **a** The mitochondria of wild-type cells (WT-cell) and Mst1-knockout cells (Mst1^KO^-cell) were stained with Tom20 under control (Ctrl) and hypoxia conditions and mitochondrial fission was measured. **b** The average length of mitochondria was quantified. **c** and **e** The TUNEL assay was used to label the apoptotic cells. **d** The LDH release assay was carried out to measure the cardiomyocyte damage with mitochondrial fission activation and inhibition. \**p* \< 0.05 vs. ctrl group, ^\#^*p* \< 0.05 vs. WT-cell+hypoxia, ^†^p \< 0.05 vs. Mst1^KO^-cell+hypoxia To investigate whether fission is responsible for cardiomyocyte death, LDH release and TUNEL assays were performed. In hypoxia-treated cells, Mdivi1, a fission inhibitor, was used to inhibit hypoxia-activated fission. By comparison, FCCP, a fission activator, was administered to Mst1^KO^ cells to promote mitochondrial fission. Interestingly, blocking fission reduced LDH release and the number of TUNEL-positive cells, which is similar to the results obtained with Mst1 knockout (Fig. [5c--e](#Fig5){ref-type="fig"}). However, fission activation re-increased the LDH content and the ratio of TUNEL-positive cells despite Mst1 ablation (Fig. [5c--e](#Fig5){ref-type="fig"}). These data indicated that Mst1-mediated mitochondrial fission contributes to cardiomyocyte death under chronic hypoxic stress. Mst1 governs mitochondrial fission via the JNK-Drp1 pathway {#Sec22} ----------------------------------------------------------- Mitochondrial fission is finely regulated by Drp1, which migrates to and locates on the surface of mitochondria. It can form a ring-structure around the mitochondria and divide the mitochondria into several fragments via GPT-dependent contraction. We monitored the subcellular Drp1 localization via western blotting. Compared to the control group, hypoxic stress induced more Drp1 migration from the cytoplasm to the mitochondria (Fig. [6a--d](#Fig6){ref-type="fig"}), i.e., we observed an increase in mito-Drp1 and a decrease in cyto-Drp1. Mst1 knockout restored the cyto- and mito-Drp1 balance.Fig. 6Mst1 regulated mitochondrial fission via the JNK-Drp1 pathway. **a**--**d** Western blotting was performed to analyze the signaling pathways. SP600125 (SP) was used to inhibit JNK activation (WT + SP). Anisomycin (Ani), an activator of JNK, was applied in Mst1-knockout cells to reactivate the JNK (Mst1^KO^ + Ani). **e** and **f** Mitochondrial fission and mitochondrial length were measured with JNK activation and inhibition. \**p* \< 0.05 vs. ctrl group, ^\#^*p* \< 0.05 vs. WT-cell+hypoxia, ^†^*p* \< 0.05 vs. Mst1^KO^-cell+hypoxia Previous studies have suggested that the Drp1 migration to the surface of mitochondria is regulated via JNK pathways. Interestingly, hypoxic stress activated JNK, as evidenced by an increase in phosphorylated JNK expression (Fig. [6a--d](#Fig6){ref-type="fig"}). This conformational alteration was inhibited by Mst1 deletion. To illustrate whether JNK activation accounts for Mst1-mediated Drp1 mitochondrial translocation, a JNK activator and inhibitor were used. In hypoxia-treated cells, SP600125 (SP), an inhibitor of JNK, strongly alleviated JNK phosphorylation, reducing mito-Drp1 expression, which is similar to the results for the Mst1^KO^ cells (Fig. [6a--d](#Fig6){ref-type="fig"}). By contrast, anisomycin-mediated JNK reactivation in Mst1^KO^ cells via increased JNK phosphorylation/activation, and this was accompanied by increased mito-Drp1 expression (Fig. [6a--d](#Fig6){ref-type="fig"}). This illustrates that Drp1 mitochondrial localization is regulated via the JNK pathway in cardiomyocytes in the context of chronic cardiac damage. To provide more direct support for the regulatory role of JNK in mitochondrial fission, mitochondrial morphology was observed again. JNK inhibition prevented mitochondrial fragmentations and maintained "normal" mitochondrial length, which is similar to the results for Mst1-deleted cells (Fig. [6e and f](#Fig6){ref-type="fig"}). By contrast, anisomycin-mediated JNK reactivation enhanced mitochondrial fission despite the Mst1 deficiency (Fig. [6e and f](#Fig6){ref-type="fig"}). Our data confirmed that excessive mitochondrial fission is controlled via the Mst1-JNK-Drp1 signaling pathway. Discussion {#Sec23} ========== A growing body of evidence suggests the involvement of the Mst1-Hippo pathway in cancer proliferation, cellular migration, cardiac ischemia reperfusion and diabetic cardiomyopathy \[[@CR41]--[@CR43]\]. However, little is known about the role of Mst1 in chronic cardiac injury after myocardial infarction (MI). In this study, we confirmed that:Mst1 is significantly upregulated in the myocardium after MIMst1 knockout alleviated cardiac fibrosis, the excessive inflammatory response and cardiomyocyte deathAt the molecular level, Mst1 knockout favored cardiomyocyte survival and sustained mitochondrial homeostasis by inhibiting mitochondrial fissionMst1 knockout reduced JNK activation and Drp1 mitochondrial translocation, effectively inhibiting fatal mitochondrial fission To the best of our knowledge, this is the first paper to describe the role of Mst1 in post-infarction cardiac injury and to show the involvement of the JNK-Drp1-mitochondrial fission pathway. Mst1, a downstream factor of the Hippo pathway, has been implicated in acute and chronic cardiovascular disorders. Excessive Mst1 activation aggravates acute ischemia reperfusion injury by augmenting cardiomyocyte oxidative stress \[[@CR44]\], promotes cardiac hypertrophy by enhancing cardiomyocyte necrosis \[[@CR45]\], and induces diabetic cardiomyopathy \[[@CR21]\] by inhibiting protective autophagy. In addition, Mst1 is associated with the survival, development and metastasis of colorectal cancer \[[@CR20]\], non-small cell lung cancer \[[@CR46]\] and hepatocellular carcinoma \[[@CR47]\]. Our study also found that Mst1 expression was much higher in the myocardium post-MI and that this contributed to post-infarction cardiac injury through the promotion of cardiomyocyte mitochondrial apoptosis. These data validate Mst1 as a vital regulator of cellular survival, highlighting that Mst1 may be a target for sustaining cardiomyocyte viability in response to acute and/or chronic stress damage \[[@CR48], [@CR49]\]. We found that excessive mitochondrial fission was involved in cardiomyocyte death in the post-MI heart. Moderate mitochondrial fission into sister mitochondria is activated to meet the increased metabolic requirements of cardiomyocytes \[[@CR50]--[@CR52]\]. However, based on our data, excessive mitochondrial fission induced mitochondrial dysfunction, as evidenced by uncontrolled oxidative stress, reduced mitochondrial potential, increased mPTP opening and extensive cyt-c leakage. This concurs with the findings of an earlier study that identified mitochondrial fission as the pathogenesis for cardiac acute ischemia reperfusion \[[@CR53], [@CR54]\]. Notably, earlier studies \[[@CR6], [@CR15]\] argued that mitochondrial fission caused mitochondrial DNA damage, cardiolipin oxidation and hexokinase 2 liberation, finally activating the mitochondria-dependent apoptotic pathway. Interestingly, another study suggested that excessive mitochondrial fission impaired mitophagy, ultimately resulting in the accumulation of damaged mitochondria \[[@CR5], [@CR55]\]. Based on this information, we concluded that mitochondrial fission lies upstream of mitochondrial damage, including mitophagy, mitochondrial DNA integrity and mitochondrial apoptosis. This makes mitochondrial fission an early hallmark of cardiomyocyte damage. However, more clinical evidence is needed to support our conclusion \[[@CR56]\]. We also demonstrated that the JNK-Drp1 pathway is responsible for mitochondrial fission. Drp1 migration from the cytoplasm onto the mitochondrial surface is a prerequisite for successful mitochondrial fission \[[@CR15], [@CR57]\]. Several researchers have explored the molecular basis for Drp1 migration. The AMPK, JNK, DUSP1 and p53 pathways have been verified as the molecular machinery for Drp1 mitochondrial translocation \[[@CR6], [@CR16], [@CR18], [@CR19], [@CR58]\]. In cardiac reperfusion and liver cancer, JNK activation is associated with excessive Drp1-mediated mitochondrial fission \[[@CR59]--[@CR61]\]. Our results provide the first evidence for the direct role of JNK in Drp1-mediated mitochondrial fission in post-infarction cardiac injury. Notably, whether JNK could indirectly interact with Drp1 and influence Drp1 activity remains unclear. Further research is needed. Conclusion {#Sec24} ========== We explored the role and mechanism of Mst1 in post-infarcted myocardial injur. Upregulation of Mst1 activated the JNK-Drp1 mitochondrial fission pathway, sending mitochondrial apoptotic signals to the cardiomyocytes. Based on this conclusion, novel therapeutic approaches that regulate the balance between the Mst1 level and mitochondrial homeostasis may improve the prognosis of patients after a myocardial infarction. Cyt-c : Cytochrome c Drp1 : Dynamin-related protein 1 MI : Myocardial infarction mPTP : Mitochondrial permeability transition pore Mst1 : Mammalian STE20-like kinase 1 The data and materials involved in this report are available to readers. Both authors conceived the research, performed the experiments, and participated in discussion and revision of the manuscript. Both authors read and approved the final manuscript. Ethics approval {#FPar1} =============== All animal procedures described herein were in accordance with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication No. 8523, revised 1996). All experimental protocols were approved (Approval ID: 2015053) by the Ethics Committee of Department of Critical Care Medicine, the Chinese PLA General Hospital, Beijing, China. Competing interests {#FPar2} =================== The authors declare that they have no competing interests. Publisher's Note {#FPar3} ================ Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Background {#Sec1} ========== The in-hospital mortality rate among intensive care patients is 20--30%, and these patients account for 20--50% of all hospital deaths \[[@CR1], [@CR2]\]. Thus, clinicians, hospital administrators, and policy makers are challenged to reduce hospital mortality among critically ill patients. The greatest risk of death is related to the intensive care unit (ICU) admission, triage, and discharge, with up to 10.8% of patients dying after being discharged from the ICU \[[@CR3], [@CR4]\]. In this context, death after ICU discharge is predicted by a higher acute physiology score, organ or system failure, older age, prolonged hospitalization, discharge destination, and a do-not-resuscitate order \[[@CR3], [@CR4]\]. It is only in recent years that the possible relationship between time of discharge from ICU and hospital mortality has been recognized \[[@CR5]\]. Nighttime ICU discharge refers to discharge from ICU at night and during out-of-office hours, and is also known as an "out-of-hours discharge", "after-hours discharge", "nighttime transfer", or "night shift transfer". Retrospective and prospective studies from the UK \[[@CR6]\], Australia \[[@CR7]--[@CR10]\], Canada \[[@CR11], [@CR12]\], and the USA \[[@CR13]\] have highlighted the risks of adverse outcomes that may be associated with nighttime ICU discharge. These unfavorable outcomes may include greater in-hospital mortality \[[@CR6], [@CR10]\], a higher unplanned ICU readmission rate \[[@CR9], [@CR13]\], and prolonged hospitalization \[[@CR13]\]. However, several other studies, including the most recent large-scale prospective study, failed to draw similar conclusions \[[@CR14], [@CR15]\], and similar inconsistencies have been observed in studies of whether weekend discharge is harmful \[[@CR8], [@CR10], [@CR12], [@CR14], [@CR16]\]. These discrepancies are likely influenced by the local healthcare systems, patient populations, definitions of nighttime or weekend discharge, disease severity at admission or discharge, therapy limitations, sample size, and study design. Nevertheless, no study has comprehensively examined the discrepancies and similarities among these research results. Evidence-based practical guidelines have cited the existing research data to support their suggestions on ICU discharge time \[[@CR17], [@CR18]\]. Unfortunately, the strength of these recommendations is relatively weak. For example, the UK Faculty of Intensive Care Medicine and the Intensive Care Society suggest avoiding nighttime discharge (between 22:00 and 06:59) to reduce mortality and patient discomfort \[[@CR17]\], although this suggestion cites only two retrospective studies \[[@CR11], [@CR12]\]. Based on a more comprehensive literature search, the newly revised American Society of Critical Care Medicine ICU practice guidelines recommend avoiding nighttime ICU discharge but not weekend discharge \[[@CR18]\]. However, these two practical recommendations were graded as evidence level 2C (the highest evidence level is 1A), because they were formulated using a consensus review of contradictory research evidence \[[@CR18]\]. Thus, there is a need for stronger evidence that combines all relevant data in a quantitative manner. We performed this systematic review and meta-analysis to identify whether nighttime or weekend ICU discharge is associated with hospital mortality. Methods {#Sec2} ======= Data sources and search strategy {#Sec3} -------------------------------- Two independent investigators (SY and JW) performed a systematic search, without language or publication type restrictions, of the PubMed, Embase, and Scopus databases from their inception to 1 August 2016. The searches used a combination of the following search terms with the appropriate wildcards and spelling variations: "intensive care unit", "night-shift", "night", "nighttime", "out-of-hours", "evening", "off-hour", "after-hours", "time", "discharge", "transfer", "mortality", and "death". The search was limited to studies of human adults (Additional file [1](#MOESM1){ref-type="media"}: Table S1). Publications in non-English languages (e.g., French, Japanese, or German) were translated by an independent translation service. Additional searches were performed using two clinical trial registries (<http://clinicaltrials.gov/> and <http://www.isrctn.com/>), and abstracts from major international conferences were manually searched at their official journal websites (Society of Critical Care Medicine: Critical Care Medicine (1998--2015); American Thoracic Society: American Journal of Respiratory and Critical Care Medicine (2009--2016); European Society of Intensive Care Medicine: Intensive Care Medicine (1988--2014); International Symposium on Intensive Care and Emergency Medicine: Critical Care (1997--2016); American College of Chest Physicians: Chest (2003--2015); Australian and New Zealand Intensive Care Society Annual Meeting: Anaesthesia and Intensive Care (1990--2015)). Articles that were published online ahead of print in major intensive care journals were searched manually. The two investigators also reviewed the reference lists (Additional file [1](#MOESM1){ref-type="media"}: Table S1) of the retrieved studies and relevant reviews to identify additional articles \[[@CR19], [@CR20]\]. In instances where further clarification was required, a third investigator (ZW) emailed the corresponding author of the relevant article. The meta-analysis was pre-specified and performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria (Additional file [2](#MOESM2){ref-type="media"}) \[[@CR21]\]. The meta-analysis and systematic review protocol has not been published, and is not registered with the PROSPERO database or the Cochrane Library. Study selection {#Sec4} --------------- Studies were considered eligible if they fulfilled the following criteria: (1) a cohort study design; (2) a study population of mainly adult patients who were discharged alive from an ICU (general surgical, medical, or mixed) and were grouped into nighttime/daytime discharges and/or weekday/weekend discharges, and the study assessed outcomes for nighttime versus daytime discharges or outcomes for weekend versus weekdays discharges; (3) the primary outcome was hospital mortality among the patients who were discharged from the ICU (according to their nighttime, daytime, weekday, and/or weekend grouping); and (4) the study reported the effect size and 95% confident interval (CI) with adjustment for disease severity (or data to calculate these results). Studies were excluded if they fulfilled any of the following criteria: (1) the study population comprised mainly pediatric patients; (2) the study population comprised patients discharged from a high-dependency or step-down unit; (3) there was no control population; (4) the study was not original research; or (5) the study did not provide sufficient information for data extraction and quality assessment (even after contacting the relevant authors). In cases of duplicate publication, we only included the most informative and complete study (typically the most recent publication). Two investigators (SY and JW) independently screened the titles and abstracts of all citations. The full-text articles were retrieved for full-text review if either investigator thought that the citation might fulfill our eligibility criteria. The same two investigators independently evaluated the eligibility of all full-text articles that were selected during the screening process, and the κ value (i.e., chance-independent agreement) was found to be 0.82. Disagreements were resolved through a consensus process in which investigators discussed the reasoning behind their decisions. In all disagreements, one of the investigators realized that they had made an error. Data extraction and quality assessment {#Sec5} -------------------------------------- Data extraction was independently performed by two investigators (SY and ZL); discrepancies were resolved using discussion and consensus. A predefined standardized data extraction form was used to collect data. The following data were collected from each study: the study name, the first author's name, publication year, the study design, the study location, the patients' ages, the patients' sex distribution, the definition of night or weekend, disease severity, adjustments, outcomes, odds ratios (ORs) and 95% CIs, numbers of patients discharged during the nighttime and the daytime, crude hospital mortality among patients discharged during nighttime and daytime, numbers of patients discharged during the weekend and weekdays, crude hospital mortality among patients discharged during the weekend and weekdays, and total number of patients discharged. We also checked the supplementary files and contacted the study authors in cases where more detailed information was needed. Because all of the included studies were cohort studies, the Newcastle-Ottawa Scale (NOS) was used to assess study quality (available at: <http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp>) \[[@CR22]\]. This scale uses a star system to evaluate study quality in three domains: cohort selection (maximum of four stars), comparability (maximum of two stars), and outcome (maximum of three stars). A score of nine stars indicates the highest possible quality. For the present study, we defined high-quality studies as having \>5 stars (a low risk of bias) and low-quality studies as having ≤5 stars (a high risk of bias) \[[@CR23], [@CR24]\]. Two investigators (SY and ZL) independently performed the quality assessment; discrepancies were resolved using discussion and consensus. Statistical analysis {#Sec6} -------------------- The primary meta-analysis evaluated the association between nighttime ICU discharge and hospital mortality. The secondary meta-analysis evaluated the association between weekend ICU discharge and hospital mortality. The overall estimates were presented as OR and 95% CI values, which were determined using a random-effects model that accounted for any differences between the studies, even if there was no statistically significant heterogeneity \[[@CR25]\]. The individual estimates were used in the sub-analyses for one study \[[@CR12]\] that separately reported ORs for weekday night discharge, weekend daytime discharge, weekend night discharge, and weekday daytime discharge. However, to compare nighttime and daytime discharge, we combined the weekend night discharges and weekday night discharges into a single group. An overall estimate for this group was calculated from the available ORs for weekend night discharge and weekday night discharge using a fixed-effects model and the inverse-variance method. To compare weekend and weekday discharge, we combined weekend night discharges and weekend day discharges into another group. An overall estimate for this group was calculated from the available ORs for weekend night discharge and weekend day discharge using a fixed-effects model and the inverse-variance method \[[@CR26]\]. Heterogeneity was evaluated using the Cochran *Q* statistic and *I* ^2^ statistic, which are quantitative measures of inconsistency across studies \[[@CR27]\], and heterogeneity was considered statistically significant at *P* values \<0.1 or *I* ^2^ values \>50%. Subgroup analyses were performed to examine the potential sources of heterogeneity using pre-specified subgroups that included geographical region, study design, and study characteristics (definition of nighttime and the total number of patients discharged). We also performed post-hoc subgroup analyses according to adjustments for certain confounding factors (adjustment for illness severity at the ICU discharge, adjustment for treatment limitation orders, and adjustment for premature discharge), as these confounding factors might affect the results of our analyses. Sensitivity analysis was performed to explore the possible causes of any heterogeneity and to estimate the influence of missing studies on the overall estimates by changing the pooling model (from a random-effects model to a fixed-effects model) and using the one-study-out method. Egger linear regression testing was performed to test for publication bias \[[@CR28]\]. All statistical analyses were performed using STATA software (version 12.1; StataCorp, College Station, TX, USA), and differences were considered statistically significant at a two-tailed *P* value \<0.05. Results {#Sec7} ======= Study selection and study characteristics {#Sec8} ----------------------------------------- The initial search identified 5259 potentially relevant publications, although 2113 reports were excluded because of duplicate publication. We also excluded 3106 studies based on reviews of the titles and abstracts. Full-text reviews were performed for the remaining 40 studies, and we ultimately identified 14 cohort studies for inclusion in the meta-analysis \[[@CR5]--[@CR15], [@CR29]--[@CR31]\]. The justifications for the study exclusions are shown in Additional file [1](#MOESM1){ref-type="media"}: Table S2. The strategies for study identification and study selection are shown in Fig. [1](#Fig1){ref-type="fig"}.Fig. 1Flow chart of the article selection process The main characteristics of the 14 included studies are shown in Tables [1](#Tab1){ref-type="table"}, [2](#Tab2){ref-type="table"} and [3](#Tab3){ref-type="table"}. The 14 studies were all published in English between 2000 and 2015. Six studies were performed in Oceania \[[@CR7]--[@CR10], [@CR15], [@CR30]\], four studies were performed in Europe \[[@CR5], [@CR6], [@CR14], [@CR29]\], and the other four studies were performed in North America \[[@CR11]--[@CR13], [@CR31]\]. The studies included six single-center studies and eight multicenter studies, and used a retrospective design (n = 12) or a prospective design (n = 2).Table 1The main characteristics of the cohort studies included in this meta-analysisFirst author/publication yearStudy locationStudy designAge (years)Male (%)Definition of night or weekendDisease severityPopulationAdjustmentsOutcomeOR and 95% CIsNOS scoreReferenceSantamaria et al./201540 ICUs in Australia and New ZealandProspective multicenter cohortMedian (IQR) 63 (49--74)61Night (18:00--06:00)APACHE III-J risk of death median (IQR) 0.09 (0.03--0.25)AdultMarkers of illness severity at the time of ICU discharge: age, cardiac surgery, treatment limitation order, tracheostomy, ongoing dialysis, parenteral nutrition, and altered conscious stateIn-hospital mortality1.16 (0.89, 1.53)8\[[@CR15]\]Azevedo et al./20155 ICUs in CanadaRetrospective multicenter cohort57.5 (18.0)57.9Night (19:00--07:59), weekend (Fri 19:00--Mon 07:59)APACHE II score 19.4 (7.5)AdultDemographics, co-morbidity, APACHE II score at ICU admission, use of mechanical ventilation, ICU length-of-stay, surgical status, admission source, primary diagnostic category, study year, type of hospitalIn-hospital mortalityNight 1.29 (1.14, 1.46), weekend 0.95 (0.84, 1.07)6\[[@CR31]\]Gantner et al./2014103 ICUs contributing to ANZICS APD from 2005 to 2012Retrospective multicenter cohortAfter-hours 59.4 (19.8), in-hours 60.3 (19.3)NANight (18:00--06:00)APACHE III score After-hours 50.0 (25.3), in-hours 46.5 (22.9)AdultAPACHE III risk of death at ICU admission, presence of treatment limitation orders at ICU admission, diagnostic category, hospital siteIn-hospital mortality1.34 (1.30, 1.38)7\[[@CR30]\]Laupland et al./2011French ICUs The Outcomerea databaseRetrospective multicenter cohortMedian (IQR) 62 (49--75)61Night 18:00-07:59)SAPS II score median (IQR) 40 (28--56)AdultAdmission SAPSII, medical/surgical classification, presence of septic shock, admission decision to forego life-sustaining therapy(DFLST) order, discharge SOFA scoreIn-hospital mortality1.54 (1.12, 2.11)7\[[@CR29]\]Singh et al./20101 ICU in a tertiary care teaching hospital in AustraliaRetrospective single-center cohortMedian 6061.3Night (18:00--07:59), weekend (Sat and Sun)APACHE II score median 18 (range, 1--44)AdultAge, APACHE II score at ICU admission, discharge destinationIn-hospital mortalityNight 1.38 (1.01, 1.88), weekend 1.04 (0.73, 1.46)6\[[@CR10]\]Hanane et al./20083 ICUs of Mayo Medical Center in USARetrospective single-center cohortNight 61.6 (18.0), day 62.7 (17.8)Night 56.0, day 53.1Night (19:00--06:59)APACHE III score night 47.6 (21.1), day 44.9 (19.0)AdultDNR order by the time of transfer, the last ICU day APACHE III predicted mortalityIn-hospital mortality1.05 (0.64, 1.70)7\[[@CR13]\]Laupland et al./20084 ICUs in the Calgary Health Region, Alberta, CanadaRetrospective single-center cohortMedian (IQR) 63.7 (49.9--73.8)64Night (18:00--07:59), weekend (Sat and Sun)APACHE II score 25.1 (8.48)AdultNoncardiac surgery, cardiac surgery, age, APACHE II at ICU admission, weekend admission, night admission, regional residentIn-hospital mortalityWeekday night discharge 1.20 (1.01, 1.41), weekend day discharge 0.81 (0.67, 0.98), weekend night discharge 1.35 (1.05, 1.73)6\[[@CR12]\]Pilcher et al./200740 ICUs in Australia and New ZealandRetrospective multicenter cohortNight 58.6 (0.08), day: 59.1 (0.17)NANight (18:00--06:00)APACHE III score night 47.7 (0.2), day: 46.0 (0.1)AdultAPACHE III risk of death at admission, emergency admission to ICUIn-hospital mortality1.42 (1.32, 1.53)6\[[@CR9]\]Tobin et al./20061 ICU in AustraliaRetrospective single-center cohort64 (13--98)65Night (22:00--06:59), weekend (Fri 18:00--Mon 07:59)APACHE II score median 13 (range 0--53)AdultAge, APACHE II score at admission, origin of admission, treatment categoryIn-hospital mortalityNight 1.63 (1.03, 2.57), weekend 1.46 (1.18, 1.81)6\[[@CR8]\]Priestap et al./200631 Critical care units across CanadaRetrospective multicenter cohortNight 61.6 (17.7), day 61.7 (17.5)Night 58, day 57.4Night (21:00--06:59)APACHE II score night 15.7 (7.7), day 15.0 (7.4)AdultDifferences in illness severity at admission, gender, age, admission source, admission diagnosis, siteIn-hospital mortality1.22 (1.10, 1.36)6\[[@CR11]\]Duke et al./20041 ICU in the Northern Hospital in AustraliaProspective single-center cohortMedian (IQR) 62 (42--73)NANight (22:00--07:30)APACHE II score 15 (10--21)Adultage, APACHE II predicted mortality at admission, premature discharge, delayed discharge, limitation of medical treatment decision, emergency admission, mechanical ventilation, APACHE II diagnosis, chronic health status categoriesIn-hospital mortality1.7 (1.03, 2.9)7\[[@CR7]\]Uusaro et al./200318 ICUs in university and central hospitals in FinlandRetrospective multicenter cohortNANANight (16:00--08:00), weekend (Fri 16:00--Sun 24:00)SAPS II score 34 (17)AdultDisease severity at ICU admission, intensity of care, and whether restrictions for future care were setIn-hospital mortalityNight 1.11 (0.93, 1.31), weekend 0.88 (0.73, 1.07)7\[[@CR14]\]Beck et al./20029 ICUs in a district general hospital in United KingdomRetrospective single-center cohort57 (19)61.7Night (20:00--07:59)APACHE II probabilities 18.3 (18.7)AdultDisease severity at ICU admissionIn-hospital mortality1.70 (1.28, 2.25)6\[[@CR6]\]Goldfrad et al./200088 ICUs in the United KingdomRetrospective multicenter cohortMean (95% CI) night 57.5 (56.4--58.7), day 58.2 (57.9--58.5)NANight (22:00--06:59)APACHE II score mean (95% CI), night 15.5 (15.1--16.0), day 14.6 (14.5--14.7)AdultCase-mix (age, medical history, acute severity), premature dischargeIn-hospital mortality1.17 (0.92, 1.49)6\[[@CR5]\]*Abbreviations*: *OR* odds ratio, *CI* confidence interval, *ICU* intensive care unit, *IQR* interquartile range, *APACHE* Acute Physiology and Chronic Health Evaluation, *SAPS* Simplified Acute Physiology Score, *SOFA* sequential organ failure assessment, *DNR* do-not-resuscitate assessment, *ANZICS APD* Australian and New Zealand Intensive Care Society Adult Patient Database, *NOS* Newcastle-Ottawa Scale, *NA* information not available Continuous data given as mean (sd or 95% CI) or median (interquartile range) if provided by the study authors Table 2Number of patients and crude hospital mortality in studies in which outcomes were assessed for nighttime versus daytime dischargeFirst author/publication yearTotal number of patients dischargedDischarged during daytime, *n* (%)Discharged during nighttime, *n* (%)Crude hospital mortality among patients discharged during daytime, %Crude hospital mortality among patients discharged during nighttime, %Santamaria et al./2015 \[[@CR15]\]10,2118539 (83.6)1672 (16.4)4.87.4Azevedo et al./2015 \[[@CR31]\]19,62216,117 (82.1)3505 (17.9)8.811.8Gantner et al./2014 \[[@CR30]\]710,535601,151 (84.6)109,384 (15.4)3.66.4Laupland et al./2011 \[[@CR29]\]59925333 (89.0)659 (11.0)59Singh et al./2010 \[[@CR10]\]18711221 (65.3)650 (34.7)10.113.7Hanane et al./2008 \[[@CR13]\]11,65911,241 (96.4)418 (3.6)4.55.3Laupland et al./2008 \[[@CR12]\]17,86414,151 (79.2)3713 (20.8)512Pilcher et al./2007 \[[@CR9]\]76,69062,704 (81.8)13,986 (18.2)5.38Tobin et al./2006 \[[@CR8]\]10,903NANANANAPriestap et al./2006 \[[@CR11]\]47,06242,290 (89.9)4772 (10.1)911.8Duke et al./2004 \[[@CR7]\]18701578 (84.0)292 (16.0)4.38.2Uusaro et al./2003 \[[@CR14]\]20,62316,952 (82.2)3671 (17.8)9.811.5Beck et al./2002 \[[@CR6]\]16541351 (81.7)303 (18.3)11.218.8Goldfrad et al./2000 \[[@CR5]\]16,75615,747 (94.0)1009 (6.0)1318.1*NA* information not available Table 3Number of patients and crude hospital mortality in studies in which outcomes were assessed for weekend versus weekday dischargeFirst author/publication yearTotal number of patients dischargedDischarged during weekend, *n* (%)Discharged during weekdays, *n* (%)Crude hospital mortality among patients discharged during weekend, %Crude hospital mortality among patients discharged during weekdays, %Azevedo et al./2015 \[[@CR31]\]19,6224676 (23.8)14,946 (76.2)NANASingh et al./2010 \[[@CR10]\]1871567 (30.3)1304 (69.7)NANALaupland et al./2008 \[[@CR12]\]17,8644661 (26.1)13,203 (73.9)67Tobin et al./2006 \[[@CR8]\]10,903NANANANAUusaro et al./2003 \[[@CR14]\]20,6232932 (14.2)17,691 (85.8)9.210.2*NA* information not available Disease severity was reported based on the Acute Physiology and Chronic Health Evaluation (APACHE) II score (n = 8 studies), the APACHE III score (n =4), or the Simplified Acute Physiology Score (SAPS) II (n = 2). In all 14 studies there was adjustment for a wide range of potential confounders, such as age, treatment limitation orders, premature discharge, diagnostic category, and the use of mechanical ventilation. In 11 studies there was adjustment for disease severity on ICU admission \[[@CR5]--[@CR12], [@CR14], [@CR30], [@CR31]\], and in 3 studies adjustment for disease severity on ICU discharge \[[@CR13], [@CR15], [@CR29]\]; in 6 studies there was adjustment for treatment limitation orders \[[@CR7], [@CR13]--[@CR15], [@CR29], [@CR30]\], but no such adjustment in the other 8 studies \[[@CR5], [@CR6], [@CR8]--[@CR12], [@CR31]\]. In two studies there was adjustment for premature discharge \[[@CR5], [@CR7]\]. A total of 953,312 patients were included in the meta-analysis, with the study samples ranging from 1654 patients to 710,535 patients. Four studies included ≤10,000 patients \[[@CR6], [@CR7], [@CR10], [@CR29]\] and 10 studies included \>10,000 patients \[[@CR5], [@CR8], [@CR9], [@CR11]--[@CR15], [@CR30], [@CR31]\]. The mean proportion of nighttime discharges in 13 studies was 15.3% (range 3.6--34.7%), and one study did not report this information \[[@CR8]\]. The evaluated discharge times included nighttime in all 14 studies \[[@CR5]--[@CR15], [@CR29]--[@CR31]\] and weekends plus nighttime in 5 studies \[[@CR8], [@CR10], [@CR12], [@CR14], [@CR31]\]. None of the 14 studies used consistent definitions of nighttime or weekend. The average NOS score of the included studies was 6.5 (range 6--8) (see Additional file [1](#MOESM1){ref-type="media"}: Table S3). Nighttime discharge and hospital mortality {#Sec9} ------------------------------------------ The 14 studies had 953,312 patients who were evaluated for daytime/nighttime discharge (Table [2](#Tab2){ref-type="table"}). The adjusted OR for hospital mortality was significantly higher among patients discharged during the nighttime, compared to patients discharged during the daytime (OR 1.31, 95% CI 1.25--1.38, *P* \< 0.0001) (Fig. [2](#Fig2){ref-type="fig"}), and the individual studies had low heterogeneity (*I* ^2^ = 33.8%, *P* = 0.105). We also performed subgroup analyses and sensitivity analyses (Table [4](#Tab4){ref-type="table"}), which revealed that the significant association between nighttime discharge and hospital mortality was not substantially modified by geographical region, the total number of discharges, or study design (Additional file [3](#MOESM3){ref-type="media"}: Figures S1--S3).Fig. 2Forest plots of the association between nighttime discharge from the ICU and hospital mortality. The size of each *square* is proportional to the study weight. *Open diamond* represents the overall pooled OR. *D + L* random effects, *I-V* fixed effects Table 4Subgroup and sensitivity analyses for hospital mortalityAnalysisStudies, *n*Odds ratio (95% CI)*P* heterogeneity*I* ^2^Study referenceSubgroup analysis The definition of night  18:00--06:0031.36 (1.29, 1.43)0.19838.30%\[[@CR9], [@CR15], [@CR30]\]  18:00--07:5931.30 (1.15, 1.46)0.4320.00%\[[@CR10], [@CR12], [@CR29]\]  19:00--07:5911.29 (1.14, 1.46)\[[@CR31]\]  19:00--06:5911.05 (0.64, 1.70)\[[@CR13]\]  22:00--07:3011.7(1.03, 2.9)\[[@CR7]\]  21:00--06:5911.22 (1.10, 1.36)\[[@CR11]\]  16:00--08:0011.11(0.93, 1.31)\[[@CR14]\]  20:00--07:5911.70 (1.28, 2.25)\[[@CR6]\]  22:00--06:5921.30 (0.96, 1.76)0.20936.80%\[[@CR5], [@CR8]\] Geographic region  Oceania61.35 (1.31, 1.39)0.4570.00%\[[@CR7]--[@CR10], [@CR15], [@CR30]\]  Europe41.33 (1.08, 1.63)0.03964.00%\[[@CR5], [@CR6], [@CR14], [@CR29]\]  North America41.24 (1.16, 1.33)0.820.00%\[[@CR11]--[@CR13], [@CR31]\] Total discharge number   ≤ 1000041.56 (1.32, 1.84)0.7850.00%\[[@CR6], [@CR7], [@CR10], [@CR29]\]   \> 10000101.29 (1.23, 1.36)0.09139.90%\[[@CR5], [@CR8], [@CR9], [@CR11]--[@CR15], [@CR30], [@CR31]\] Study design  Multicenter studies81.30 (1.23, 1.38)0.06747.00%\[[@CR5], [@CR9], [@CR11], [@CR14], [@CR15], [@CR29]--[@CR31]\]  Single-center studies61.38 (1.20, 1.59)0.26822.10%\[[@CR6]--[@CR8], [@CR10], [@CR12], [@CR13]\] Whether or not adjusted for severity of illness at the time of ICU discharge  YES31.26 (1.02, 1.57)0.29617.80%\[[@CR13], [@CR15], [@CR29]\]  NO111.31 (1.25, 1.38)0.07640.90%\[[@CR5]--[@CR12], [@CR14], [@CR30], [@CR31]\] Whether or not adjusted for treatment limitation orders  Yes61.28 (1.15, 1.43)0.15238.10%\[[@CR7], [@CR13]--[@CR15], [@CR29], [@CR30]\]  No81.32 (1.23, 1.42)0.11639.40%\[[@CR5], [@CR6], [@CR8]--[@CR12], [@CR31]\] whether or not adjusted for premature discharge  Yes21.31 (0.94, 1.84)0.239.20%\[[@CR5], [@CR7]\]  No121.31 (1.25, 1.38)0.0937.70%\[[@CR6], [@CR8]--[@CR15], [@CR29]--[@CR31]\] Sensitivity analysis  Fixed-effects model141.33 (1.30, 1.37)0.10533.80%\[[@CR5]--[@CR15], [@CR29]--[@CR31]\]  Random-effects model141.31 (1.25, 1.38)0.10533.80%\[[@CR5]--[@CR15], [@CR29]--[@CR31]\] One-study-out method  Santamaria et al./201511.31 (1.24, 1.39)\[[@CR15]\]  Azevedo et al./201511.31 (1.25, 1.37)\[[@CR31]\]  Gantner et al./201411.31 (1.24, 1.38)\[[@CR30]\]  Laupland et al./201111.30 (1.21, 1.39)\[[@CR29]\]  Singh et al./201011.32 (1.25, 1.39)\[[@CR10]\]  Hanane et al./200811.31 (1.25, 1.38)\[[@CR13]\]  Laupland et al./200811.32 (1.25, 1.39)\[[@CR12]\]  Pilcher et al./200711.31 (1.24, 1.37)\[[@CR9]\]  Tobin et al./200611.29 (1.22, 1.36)\[[@CR8]\]  Priestap et al./200611.33 (1.26, 1.40)\[[@CR11]\]  Duke et al./200411.32 (1.25, 1.38)\[[@CR7]\]  Uusaro et al./200311.31 (1.24, 1.38)\[[@CR14]\]  Beck et al./200211.31 (1.24, 1.38)\[[@CR6]\]  Goldfrad et al./200011.33 (1.27, 1.39)\[[@CR5]\] Among the 11 studies in which there was adjustment for disease severity on ICU admission, the summary OR for hospital mortality was 1.31 (95% CI 1.25--1.38). A summary OR was provided for hospital mortality with a broader range (OR 1.26, 95% CI 1.02--1.57) in the three studies in which there was adjustment for disease severity at ICU discharge (Fig. [3](#Fig3){ref-type="fig"}). In our meta-analysis, the risk of hospital mortality did not depend on whether or not the study analyses were adjusted for disease severity at ICU discharge.Fig. 3Forest plots of the association between nighttime discharge from the ICU and hospital mortality stratified by whether or not the data were adjusted for severity of illness at the time of ICU discharge. The size of each *square* is proportional to the study weight. *Open diamonds* represent the pooled OR. *D + L* random effects, *I-V* fixed effects Among the six studies in which there was adjustment for treatment limitation orders, the summary OR for hospital mortality was 1.28 (95% CI 1.15--1.43); among the remaining eight studies without this adjustment, the summary OR for hospital mortality was 1.32 (95% CI 1.23--1.42) (Fig. [4](#Fig4){ref-type="fig"}). In our meta-analysis, the risk of hospital mortality did not depend on whether or not there was adjustment for treatment limitation orders. We did not observe a significant relationship between nighttime discharge and hospital mortality in the subgroup analysis that was adjusted for premature discharge (Fig. [5](#Fig5){ref-type="fig"}). Our sensitivity analyses suggested that the overall estimates were not materially altered by changing the pooling models (random-effects model, OR 1.31, 95% CI 1.25--1.38; fixed-effects model, OR 1.33, 95% CI 1.30--1.37) and were not materially altered when an individual study was omitted from the sequence, with a range of 1.29 (95% CI 1.22--1.36) to 1.33 (95% CI 1.27--1.39) (Table [4](#Tab4){ref-type="table"}).Fig. 4Forest plots of the association between nighttime discharge from the ICU and hospital mortality stratified by whether or not the data were adjusted for treatment limitation orders. The size of each *square* is proportional to the study weight. *Open diamonds* represent the pooled OR. *D + L* random effects, *I-V* fixed effects Fig. 5Forest plots of the association between nighttime discharge from the ICU and hospital mortality stratified by whether or not the data were adjusted for premature discharge. The size of each *square* is proportional to the study weight. *Open diamonds* represent the pooled OR. *D + L* random effects, *I-V* fixed effects Weekend discharge and hospital mortality {#Sec10} ---------------------------------------- Five studies with 70,883 patients evaluated weekday/weekend discharges (Table [3](#Tab3){ref-type="table"}). There was no difference in the adjusted ORs for hospital mortality when we compared patients who were discharged during the weekend or on weekdays (OR 1.03, 95% CI 0.88--1.21, *P* = 0.68) (Fig. [6](#Fig6){ref-type="fig"}). However, the studies exhibited significant heterogeneity (*I* ^2^ = 72.5%, *P* = 0.006), and we performed sensitivity analyses to explore the possible explanations for the heterogeneity. Our sensitivity analyses suggested that the overall estimates were not materially altered by changing the pooling models (random-effects model, OR 1.03, 95% CI 0.88--1.21; fixed-effects model, OR 1.00, 95% CI 0.93--1.08) and were not materially altered when an individual study was omitted from the sequence, with a range of 0.95 (95% CI 0.87--1.03) to 1.08 (95% CI 0.89--1.31). We were unable to identify the specific study that caused the heterogeneity, which might have been explained by the different patient populations, weekend definitions (including a different number of weekend days and weekend nights), and adjustments for confounding factors. However, when the Tobin study \[[@CR8]\] was omitted, the heterogeneity among the remaining four studies was markedly reduced (*I* ^2^ = 0.0%, *P* = 0.792), and we did not observe an association between weekend discharge and hospital mortality.Fig. 6Forest plots of the association between weekend discharge from the ICU and hospital mortality. The size of each square is proportional to the study weight. *Open diamond* represents the overall pooled OR. *D + L* random effects, *I-V* fixed effects Publication bias {#Sec11} ---------------- Egger's test did not reveal any significant evidence of publication bias (*P* = 0.662 for nighttime studies, *P* = 0.507 for weekend studies) (Additional file [3](#MOESM3){ref-type="media"}: Figures S4 and S5). Discussion {#Sec12} ========== The present study revealed that nighttime ICU discharge is associated with an increased risk of hospital mortality compared to daytime ICU discharge. Furthermore, it seems that nighttime ICU discharge was also associated with an increased risk of hospital mortality in the subgroup analyses that were stratified according to geographical region, the total number of discharges, or study design (multicenter or single-center). Patient-related factors, such as disease severity and limitation of medical treatment, are considered crucial determinants of mortality after nighttime ICU discharge \[[@CR7], [@CR15]\], and disease severity at the ICU admission or discharge is useful for predicting post-ICU mortality. Some of the studies in our meta-analysis found that patients discharged at nighttime were typically older and had more severe injuries, co-morbidities, or multi-organ dysfunction, compared to patients with a daytime discharge, which may explain the higher mortality with nighttime ICU discharge \[[@CR10], [@CR29], [@CR30]\]. However, our methodology was initially based on precluding the confounding effect of disease severity. Three studies included adjustment for disease severity at ICU discharge, and the remaining 11 studies included adjustment for disease severity at ICU admission, and both sets had acceptable heterogeneity. After adjusting for disease severity at ICU discharge or admission, there was still a significant relationship between nighttime discharge and hospital mortality (Fig. [3](#Fig3){ref-type="fig"}). Limitation of medical treatment orders refers to limiting or withholding one or more life-support therapies, such as cardiopulmonary resuscitation (do-not-resuscitate orders) or the use of palliative care measures \[[@CR7]\]. Patients discharged at nighttime were more likely to have a medical treatment limitation order, which was found to be an independent predictor of mortality in previous studies \[[@CR3], [@CR7], [@CR13], [@CR15]\]. For example, Santamaria et al. found that treatment limitation orders were an independent predictor of hospital survival (OR 35.4, 95% CI 27.5--45.6). However, Ouanes et al. observed a relationship between nighttime discharge and 7-day mortality or readmission after excluding patients with treatment limitation orders \[[@CR32]\]. In our meta-analysis, adjustment for treatment limitation orders was documented in six studies, and nighttime discharge remained associated with an increased risk of hospital mortality after the adjustment. Poor-quality medical care (i.e., lower staffing levels, lower nurse-to-patient ratios, and less surveillance) may partly explain the association between nighttime discharge and increased mortality. Staffing levels and nurse-to-patient ratios in the ICU and general wards are invariably lower at night, and previous studies have reported that increased mortality is associated with decreased staffing levels in the ICU and general wards \[[@CR33], [@CR34]\]. In addition, monitoring devices and life-sustaining devices are considered less immediately available in the general wards \[[@CR35]\], and 64% of major adverse events that occur within 72 h after ICU discharge may be predicted and prevented by monitoring the patient's vital signs (e.g., an abnormal respiratory rate or tachycardia) \[[@CR36], [@CR37]\]. Moreover, transfer to a ward is associated with increased post-ICU mortality, while transfer to a high-dependency unit is not, which suggests that patients' outcomes may be associated with the intensity of post-ICU medical care \[[@CR6]\]. There are also systemic factors that may contribute to the post-ICU mortality rate. Practical guidelines recommend that ICU patients should be constantly evaluated for discharge based on their physiological status and the necessity of ICU monitoring \[[@CR38]\]. However, in practice, patients are rarely evaluated for discharge at night, unless the ICU is facing pressure from additional admissions. This dilemma may lead to premature discharge, which refers to an unplanned transfer of a patient to make an ICU bed available for a more acutely or seriously sick patient \[[@CR31]\]. The proportions of premature discharges vary among different studies based on their defining criteria. Premature ICU discharge was an independent risk factor for mortality in some studies \[[@CR5], [@CR39]\], and an additional 48 h of ICU stay may provide a 39% reduction in mortality among high-risk patients \[[@CR40], [@CR41]\]. Two of the included studies included adjustment for premature discharge, and we did not observe a significant relationship between nighttime discharge and hospital mortality in this subgroup \[[@CR5], [@CR7]\]. However, in these two studies (by Goldfrad et al. \[[@CR5]\] and Duke et al. \[[@CR7]\]) there are conflicting results, and mild heterogeneity further obscures the results of this subgroup analysis (*I* ^2^ = 39.2%). Another study by Santamaria et al. also did not identify an association between premature discharge and subsequent mortality \[[@CR15]\]. Thus, the current studies have inconsistent results on whether premature discharge is an independent risk factor for post-ICU mortality, and whether premature discharge affects the association between nighttime ICU discharge and hospital mortality and the magnitude of any related effects remains unclear. Delayed discharge refers to a planned or prepared discharge that is delayed for some reason, which is commonly a lack of ward beds \[[@CR42]\]. Similar to premature discharge, delayed discharge is more likely to occur at night, and each 1-h delay is estimated to be associated with an adjusted 3% increase in the risk of mortality \[[@CR43], [@CR44]\]. Delayed discharge may increase the risk of ICU-acquired infections, which independently influence post-ICU mortality and lead to additional delayed discharges \[[@CR45]\]. Both delayed discharge and premature discharge are thought to reflect a limited bed capacity, and some authors even consider nighttime discharge as a marker of a bed shortage \[[@CR11], [@CR32]\]. This may partially reflect the increasing trend in the ratio of nighttime-t-daytime discharge during recent years. Although the average bed occupancy rate in American ICUs has remained fairly constant (1985--2000, 65%; 2000--2010, 68%), there has been an increase in the demand for intensive care practitioners to provide critical care services \[[@CR46]--[@CR48]\]. Interestingly, improving the number of beds does not always reduce delayed discharge, as Williams et al. found that delayed discharges increased by 4%, despite a significant increase in bed capacity (2000--2001 vs. 2008), which reduced the proportion of "no-bed delays" from 74% to 36% \[[@CR49]\]. This finding suggests that the problem does not lie in the number of beds, but rather in the inability of the ward to accept patients who are discharged from the ICU in a timely manner. In this scenario, nighttime discharge may be considered an indicator of inefficient patient delivery \[[@CR50]\], with nighttime discharge serving to optimize the use of existing ICU beds, instead of simply enlarging the ICU \[[@CR15], [@CR50]\]. We did not find an association between weekend discharge and hospital mortality, although heterogeneity among the studies may be a question. The different patient populations, weekend definitions (including a different number of weekend days and weekend nights), statistical analyses, and adjustments for confounding factors may also explain the significant heterogeneity. For example, the only positive result (from Tobin et al. \[[@CR8]\]) was based on a univariate analysis, while the other studies used a multivariate or logistic regression model. In addition, Tobin et al. defined weekends as being from 18:00 on Friday to 07:59 on Monday (3 days and 3 nights), while the other studies defined weekends as only Saturday and Sunday \[[@CR10], [@CR12]\]. Moreover, the fact that weekend staffing levels and medical care resources are similar to weekdays might explain the negative association between weekend discharge and hospital mortality, although this conclusion remains speculative \[[@CR12]\]. Strengths and limitations {#Sec13} ------------------------- This study has several strengths. First, we identified studies using a comprehensive systematic literature search. Second, we evaluated 953,312 patients with daytime/nighttime discharge, and the large sample size significantly increased the statistical power of the analysis. Third, the pooled estimates were stable after the comprehensive sensitivity analyses. Fourth, we did not detect publication bias, which indicates that the pooled estimates may be unbiased. This study also has several limitations. First, the cohort studies were observational and descriptive, and we cannot comment on the causality of the relationships between hospital mortality and the factors that we evaluated. Second, the studies used different definitions for nighttime and weekend, which inevitably introduces heterogeneity. Third, although both disease severity at discharge and admission can predict post-ICU mortality, only three of the included studies measured disease severity at ICU discharge. Premature ICU discharge was also more common at night, although only two of the included studies adjusted for premature discharge. Thus, the limited numbers of studies that adjusted for premature discharge or disease severity at ICU discharge might cause residual confounding. Fourth, none of the studies reported the exact demand of care and actual level of care in the ward, which obscures whether or how treatment and nursing care insufficiencies might influence post-ICU mortality. Last, the included studies did not report long-term prognosis or its possible relationship with nighttime discharge, which may also be an important issue. Moreover, the studies failed to address novel trends in critical care, including tele-ICU, liaison nurses, and rapid response teams, which may profoundly alter intensive care and post-ICU care and further alter the rate and effect of nighttime discharge. Therefore, the generalizability of our results in the real world must be tested in future studies \[[@CR51]\]. Conclusions {#Sec14} =========== The present meta-analysis revealed that nighttime ICU discharge was associated with an increased risk of hospital mortality, compared to daytime discharge. Although we did not detect a significant association between hospital mortality and weekend ICU discharge, there was significant heterogeneity among the included studies. Thus, these conclusions should be interpreted with caution, and further large-scale, well-designed, multicenter prospective studies are needed to improve our understanding of the association between ICU discharge times and hospital mortality. Additional files {#Sec15} ================ Additional file 1: Table S1.Database search strategies. **Table S2.** The justifications for the study exclusions (n = 26). **Table S3.** Newcastle-Ottawa quality assessment of included studies. **Table S4**. The main characteristics of the excluded meeting abstracts. (DOCX 98 kb) Additional file 2:PRISMA 2009 checklist. (DOC 62 kb) Additional file 3: Figure S1.Forest plots of the association between nighttime discharge from the ICU and hospital mortality stratified by geographic region. The size of each *square* is proportional to the study weight. *Open diamonds* represent the pooled OR. *D + L* refers to random effects and *I-V* to fixed effects. **Figure S2.** Forest plots of the association between nighttime discharge from the ICU and hospital mortality stratified by study design. The size of each *square* is proportional to the study weight. *Open diamonds* represent the pooled OR. *D + L* refers to random effects and *I-V* to fixed effects. **Figure S3.** Forest plots of the association between nighttime discharge from the ICU and hospital mortality stratified by the total discharge number. The size of each *square* is proportional to the study weight. *Open diamonds* represent the pooled OR. *D + L* refers to random effects and *I-V* to fixed effects. **Figure S4.** Funnel plots showing the association of nighttime discharge from the ICU with hospital mortality. *s.e.* refers to standard error, or refers to odds ratio. **Figure S5.** Funnel plots showing the association of weekend discharge from the ICU with hospital mortality. *s.e.* refers to standard error, or refers to odds ratio. (ZIP 395 kb) APACHE : Acute Physiology and Chronic Health Evaluation CI : confidence interval DNR : do-not-resuscitate assessment ICU : intensive care unit IQR : interquartile range NA : information not available OR : odds ratio SAPS : Simplified Acute Physiology Score SOFA : sequential organ failure assessment Not applicable. Funding {#FPar1} ======= This study was supported by the National Natural Science Foundation of China (Grant No.81600047) and the Natural Science Foundation of Henan Province (Grant No.142300410381). Availability of data and materials {#FPar2} ================================== All data generated or analyzed during this study are included in this published article and its additional files. Authors' contributions {#FPar3} ====================== SY conceived of and designed the study, participated in the study search, study selection, data collection, and drafted the manuscript. ZW participated in the statistical analysis and interpretation of data and revision of the manuscript. ZDL participated in the data collection and statistical analysis and interpretation of data. JLW participated in the study search, study selection, and statistical analysis and interpretation of data. LJM conceived of and designed the study and revised the manuscript. All authors read and approved the final manuscript. Competing interests {#FPar4} =================== The authors declare that they have no competing interests. Consent for publication {#FPar5} ======================= Not applicable. Ethics approval and consent to participate {#FPar6} ========================================== Not applicable.

[^1]: Dr. Hagopian\'s present address is the Department of Pathology, College of Physicians and Surgeons, New York

Introduction ============ In plants, many different species have a direct social impact through such problems as food production, biofuels and ecological services. In the past 20 years, genetic and genomic research have largely focused on a few strategically chosen models but, for many economically and environmentally important plant species, few genetic and genomic resources existed. The Plant Genomes and Biotechnology meeting at Cold Spring Harbor featured many examples of the use of next-generation sequencing (NGS) and other technologies to advance the development of emerging models that address specific problems, ranging from food security in Africa to a rare glimpse at the biochemical warfare between plants and the pests that eat them. In addition, the meeting provided an early look at an impending explosion of new data, emerging from NGS and other SNP technologies, that associate genetic (and epigenetic) variation to phenotypic variation in *Arabidopsis*, maize and rice. The ability to create inbred lines and easily share seed stocks makes plants a potential exemplar for large-scale genome annotation by association, with the same genotypes easily scored for multiple phenotypes. Other developments at the meeting included new efforts in \'industrial-grade\' phenotyping (employing robotics, satellites and even military-style drones) and advances in techniques that have allowed the molecular dissection of the specialized cells that produce useful plant products. Fast-track development of new models ==================================== How can new technologies rapidly accelerate plant research to help control a world-class pest? One such example is work on the parasitic genus *Striga*, which senses its hosts\' strigolactones, hormones that are exuded into the soil. *Striga hermonthica*, for example, sends specialized roots known as haustorium to raid its hosts\' nutrient supplies via vascular fusion, devastating subsistence crops such as sorghum in sub-Saharan Africa, and affecting the lives of nearly 100 million people. Ken Shirasu (RIKEN, Japan) presented his work on whole genome and RNA sequencing in *Striga*using next-generation technology. Using both facultative and obligate parasites in the genus, he described new research to identify genes expressed during the formation of haustorium by *Striga*in an effort to find an Achille\'s heel of sorts of the pest. The sequencing work has shown that the dicot *Striga*has also incorporated into its genome up to a dozen genes from its monocot hosts via horizontal gene transfer, according to Shirasu, leading to new insights into host-pest evolution. Pest control: lessons from basic science ======================================== Why can\'t Sorghum and other crops be engineered to reduce or eliminate strigolactone production to avoid parasite detection? Keynote speaker Ottoline Leyser (Cambridge University, UK) presented her work in the model plant *Arabidopsis*showing that strigolactones play essential roles in development by modulating the abundance of auxin efflux carriers (PINs) on plasma membranes. In Leyser\'s model, the balance of auxin sink capacity in the primary stem relative to the source capacity of the axillary bud regulates bud outgrowth. High auxin flux from the primary shoot prevents canalization of auxin in axillary buds, inhibiting vascular organization and bud outgrowth. Strigolactones reduce PIN accumulation to prevent vascular organization from buds and modulate axillary bud-mediated branching. Leyser showed computational models incorporating strigolactone\'s effect on polar auxin transport that could explain complex phenotypes. The models also confirm the central role of strigolactones in modulating auxin. The net result is that knocking out strigolactones leads to high levels of branching. This role of strigolactones in branching is also conserved in sorghum and other monocots, as Shirasu confirmed. Thus, basic research in *Arabidopsis*shows that the strigolactone pathway is too fundamental to development to be the target for manipulation, a poignant example of how basic research in powerful model systems informs applied research. These realities necessitated Shirasu\'s alternative approach to combat *Striga*using genomics. Leyser also presented recent work using natural variants to investigate the role of plasticity by examining variation in shoot branching under different nitrogen treatments. She described two strategies: low-plasticity, low-branching genotypes with rapid life cycles (so called \'James Dean\' strategies) versus a high-plasticity set of genotypes with longer life cycles (so-called \'wait-and-see\' strategies). However, after vernalization, some \'wait-and-see\' varieties switched to the \'James Dean\' strategy, apparently conditioning their strategy based on a memory of winter and the likely environmental conditions to follow. This was a keen example of a growing number of cases that document the ability of a plant to integrate many types of information to shape its body plan, a phenomenon that Leyser called \'the plasticity of plasticity\'. von Humbolt goes proteomic ========================== The sensitivity of plant development to environment underscores the importance of examining plants in their native surroundings. Here again, new technologies have opened insights into the role of secondary metabolites in an ecological context. Keynote speaker Ian Baldwin (Max Planck Institute for Chemical Ecology, Germany) presented a body of work that began with metabolic profiling of wild tobacco (*Nicotiana attenuta*) subjected to herbivory. The ensuing research was a tale of partnership and betrayal in which pollinators turn pests, whose predators, in turn, are chemically summoned to devour them. Baldwin\'s group used gene silencing technology to knock down specific secondary metabolite synthesis or responses in *N. attenuta*and put plants back into their native habitats in Utah. The next step in the protocol, according to Baldwin, was perhaps the most essential: the low-tech art of field observation. In one example, a caterpillar herbivore, *Manduca sexta*, was shown to secrete fatty acid-amino conjugates (FACs) in its saliva that trigger plant defenses. In addition to toxins and digestibility factors, FACs induce the plant to release a set of 146 volatile compounds. Using genetic manipulation (and observation), the group discovered that one of these compounds attracts carnivorous insects that prey on *M. sexta*, which is guided to the precise location of its prey by another set of compounds released from plant-FAC interactions. Baldwin also showed a new system in which RNA interference in *M. sexta*is induced by insect feeding on plants expressing hairpin constructs, a technology that could have important implications for crop protection. He invoked the approach of the 19th Century Biogeographer Alexander von Humbolt, who attempted to unify different scientific disciplines in explanation of natural phenomena. Baldwin described his vision of a new breed of field scientists with an interdisciplinary perspective \'like von Humbolt with spectrophotometer in one hand and a microarray in the other\'. In another view of the plant-pest-symbiont relationship in *Medicago*, Blake Meyers (University of Delaware, USA) showed that small-RNA sequencing revealed a group of *trans*-acting small interfering RNAs (siRNAs) that appear to regulate a large family of defense-related proteins, NB-LRRs. Such a large-scale targeting of genes represents a new potential mode of regulation in small RNAs and provides potential insights into the control of symbiosis in *Medicago*. Popgen takes on functional genomics =================================== Another group of researchers focused on using NGS to explore plant diversity at a finer scale. Magnus Nordborg (Gregor Mendel Institute, Austria) provided an overview of progress in the genetic characterization of natural variation in *Arabidopsis*. With the combined output of the international 1001 Genomes sequencing project and other SNP data from hybridization technologies, Nordborg estimated that the *Arabidopsis*community will have high-density SNP and other polymorphisms for about 2,000 natural inbred lines within the next year. His group showed some early results in which associations were mapped to combined phenotypes, where multiple traits borrow mapping strength from each other and reduce noise. Alluding to published work, Nordborg provided some words of caution about instances in which genome-wide association mapping can be positively misleading. Nevertheless, Nordborg said such cases are relatively infrequent; he envisioned that genome-wide association studies will soon provide the plant community with a general resource in which each variable position in the *Arabidopsis*genome is annotated with associations for many phenotypes. Edward Buckler (Cornell University, USA) described a similar scale project in maize, looking at 100 lines for 50 million segregating variable regions. Buckler said early analysis of the domesticated maize genome compared with that of its wild relative teosinte showed that about 500 to 600 genes were the targets of domestication. Surprisingly, there appear to be an equal number of genes targeted between tropical and temperate maize varieties - underscoring the importance of local adaptation. Buckler said the costs of genotyping are declining rapidly, with current costs at about \$9 a sample, which presents both an opportunity and a massive data problem. He envisioned that 30,000 samples will soon be genotyped at 200 million variable positions, with a looming \'12 trillion data point\' storage and access challenge. Nonetheless, like Nordborg, Buckler foresees the pending genotype-to-phenotype resources as a game-changing development for genetics. In the case of maize, such a dense set of markers and phenotype associations should provide predictive power to cut down on one of the biggest bottlenecks in breeding programs: waiting for crops to mature to score their phenotype. In another potential advance for plant breeding, Simon Chan (UC Davis, USA) showed his work on creating haploid plants using perturbation of the gene encoding the centromere-specific histone (*CENH3*). He showed his work on creating haploid plants using perturbation of the gene encoding the centromere-specific histone (*CENH3*). Chan showed that haploid plants can be rapidly generated, with aberrant-parent chromosomes efficiently lost after fertilization and diploids recovered in the next generation. He showed a remarkable set of pilot experiments in which he and collaborators created completely homozygous recombinant inbred mapping populations within two generations, a potential fast track for quantitative trait locus analysis in crops and other plants. The epigenetic contribution to variation ======================================== Such technological advances in genotyping are leading to a more thorough accounting of the genetic causes of phenotypic variation. Indeed, Nordborg showed instances in which almost all the heritable variation for a given trait could be mapped to specific genome locations. Still, one question that looms in genotype-to-phenotype analyses is the role of epigenetic effects. Bob Schmitz (Salk Institute, USA) from the Ecker Lab showed recent results of high-throughput whole-genome bisulfite DNA sequencing in 182 different accessions. The group found 14,000 differentially methylated regions among the accessions, with the majority in transposable elements, but a significant proportion in coding regions. Approximately 50% of the epialleles in coding regions tended to be rare or often exclusive to one background. Moreover, using an association mapping approach, methylQTLs identified epialleles that are present at high allele frequencies in the population. These methylQTLs will help reveal the genetic architecture of some epialleles. For epialleles that did not have genetic basis, Schmitz posited that they come from incomplete methylation during reproduction, as evidenced by the overlap of loci affected in natural variants and *met1*or *rdd*mutants. Clearly, some proportion of epigenetic variation is caused by underlying genetic variation, such as transposable elements. However, it will be important to know how much phenotypic variation could be caused by purely epigenetic effects, which are potentially invisible to the genotype map. Both Schmitz and Detlef Weigel (Max Planck Institute for Developmental Biology, Germany) described recently published work on the spontaneous accumulation of CG methylation variation after 30 generations of single-seed propagation. The take-home message was similar in both presentations: the accumulation of DNA methylation variation is perhaps 1,000-fold faster than the rate of appearance of spontaneous genetic variation. In addition, newly arising epigenetic variation is far from random. For example, Weigel reported that although only 1% of all methylated sites distinguished a random pair of lines 30 generations apart, many sites changed in more than one line. Although phenotypic effects of some methylation variants are well documented, the next question is how much of the natural variation and spontaneous DNA methylation observed in genome-wide surveys has an impact on phenotype, answers to which should be coming soon. Another important question that arises from genotype-to-phenotype mapping when \'phenotypes\' are transcriptomic readouts, as noted by Weigel, is what tissue and cell types will be assayed to associate with genetic variation. On the flip side of the genotype-to-phenotype efforts, the characterization of phenotype is also benefiting from an infusion of new technology. Ulrich Schurr (Forschungszentrum Jülich, Germany) described a battery of automated phenotyping tools for root and shoot being developed at Jülich\'s Institute of Plant Sciences. The pipeline included magnetic resonance imaging and positron emission tomography of roots grown in soil to characterize root architecture and function, with Schurr emphasizing the collection of high-throughput physiological measurements. They have also tracked soil nutrient distribution in fields, showing dynamic changes in local nitrogen profiles on a day-to-day basis. Schurr explained later that efforts to coordinate different types of measurements taken from unmanned drones and satellites will go toward gathering more accurate measurements of photosynthetic rates, which are notoriously inaccurate. He said that such revised measures of photosynthesis could, for example, improve the accuracy of climate models. Cellular genomics ================= Another set of well-developed genomic techniques in plants has led to analysis of specialized cells and stages of maturation. Several speakers presented recent work using high-resolution transcriptomic profiles as a diagnostic tool to interpret phenotypes. Zachary Lippman (Cold Spring Harbor Laboratory, USA) used a library of developmental stage-specific profiles to show that mutants or natural variants in tomato with highly branched inflorescences appeared to arise from perturbations of developmental staging, a heterochronic effect of sorts; sympodial inflorescence meristems in backgrounds that gave rise to more reproductive structures were \'born\' at younger stages of development than less branched backgrounds. The apparent juvenilization of secondary meristems correlated with the potential to bear an inflorescence with more fruiting structures, a potentially important agronomic trait. I used the diagnostic features of cell type profiles to quantitatively analyze single cell profiles, showing that a single cell could be fine mapped using its RNA-seq profile to a small population of cells in the root. The goal of the work is to examine how the cell buffers transcriptional noise and how plant cells retain their remarkable plasticity. Siobhan Brady (UC Davis, USA) used a compendium of tissue- specific expression profiles to define a candidate set of transcription factors that together may regulate xylem fate. She then used yeast one hybrid analyses to construct a transcriptional network for xylem cells. Since many useful products or nutritive parts of the plant arise from specialized cells, knowledge of their circuitry in these powerful cellular dissection systems in *Arabidopsis*is likely to be useful in improving plant production, much as basic research in the strigolactone pathway has informed applied research in crop protection. Abbreviations ============= FAC: fatty acid-amino conjugate; NGS: next-generation sequencing; SNP: single-nucleotide polymorphism. Acknowledgements ================ KDB is supported by the National Institutes of Health and the National Science Foundation.

1. Introduction {#sec1} =============== How organisms respond appropriately to the wide variety of pathogens and antigens they encounter, and how pathogens can subvert the host immune response, has not been fully analyzed. The immune response to infectious agents involves a complex interaction of different cell types, and two defense arms have evolved to protect the host from microbial attack: a rapidly responding innate immune response to sequester and eliminate pathogens followed by a highly specific adaptive immune response. Dendritic cells (DCs) represent the bridge between the innate and adaptive immune response \[[@B1]\], and several studies support the hypothesis that DCs specifically strengthen the cellular immune response against mycobacterial infections \[[@B2], [@B3]\]. Even though the critical role of DCs in the initiation of immune response has been established \[[@B4]\], their involvement in *Mycobacterium tuberculosis* (MTB) infection is not completely characterized. Following pulmonary infection with MTB, DCs are activated rapidly to produce a specific pattern of chemokines and cytokines, key participants in the early immune response, and to express maturation markers that allow them to migrate to the draining lymph nodes. DCs become fully competent antigen presenting cells (APCs) and participate to the development of T helper 1 (Th-1) cells, required for the elimination of intracellular pathogens \[[@B4]--[@B6]\]. In particular, interleukin 12 (IL-12) produced by activated DCs induces Th-1 cells that, in turn, release interferon *γ* (IFN-*γ*) and tumour necrosis factor *α* (TNF-*α*). These cytokines activate intracellular microbicidal mechanism and initiate a protective granulomatous response \[[@B7]\]. The magnitude of host immune response depends, to a large extent, on the presence of costimulatory molecules and signaling lymphocytic activation molecules on DC surface, as well as local production of cytokines \[[@B8]\]. *Mycobacterium bovis* bacillus Calmette-Guérin (BCG) is a widely used vaccine against tuberculosis (TB) but comparative genetic analyses of BCG around the globe have revealed that each vaccine currently in use has different traits \[[@B9]\]. For decades, a number of factors have been considered responsible for the variable efficacy of BCG, including the type of strains used. In general, different kinds of stimuli result in differently activated DCs that induce qualitatively different T cell responses. Recently, it has been described that DCs are able to discriminate between phylogenetically diverse pathogens. In fact, the analysis of the DCs responses to *E. coli* and *C. albicans* showed that a unique number of genes, were regulated by each pathogen \[[@B10], [@B11]\]. However, the downstream target genes induced in DCs by the different BCG strains have not yet been fully determined. The importance of DCs in initiating an immune response against mycobacterial infections led us to investigate the activation induced on these APCs following stimulation with two widely employed and different BCG strains. The goal of the present study was to determine whether the strains divergences may influence their relative immunogenicity \[[@B12], [@B13]\], virulence \[[@B14], [@B15]\], and viability \[[@B16]\], factors that must be considered for the design and improvement of a vaccine against TB. We also analyzed the DCs\' response to the commonly used MTB virulent laboratory strain (H37Rv) and to an MTB clinical isolate (CMT97), previously reported to behave differently from H37Rv in human macrophages \[[@B17]\], in order to understand if the laboratory strain could be considered a real model of DCs-MTB interaction. We aimed to understand whether the maturation reprogramming occurring on DCs, following infection with MTB H37Rv, MTB CMT97, BCG Aventis and BCG Japan, could be different as a consequence of the ability of DCs to discriminate between these mycobacterial strains. We used oligonucleotide macroarrays to characterize DCs gene expression profile and we found that although all infecting mycobacteria induced a core response, a strain-specific program emerged. The data obtained showed that BCG Japan was more effective than both MTB strains at inducing the expression of TNF-*α*, a gene involved in inflammation, as well as CCR7, responsible for DCs migration to lymph nodes. Furthermore, MTB H37Rv displayed, as compared to BCG Japan, an improved induction of EBI3, a IL-12p40-related polypeptide of IL-27 that may play a role in regulating cell-mediated immune response \[[@B18], [@B19]\], but, on the other hand, it resulted to be the only mycobacterial strain unable to promote a statistically significant IL-10 release as compared to uninfected cells. Moreover, IL-12 was significantly released only upon BCG Aventis infection. Finally, we also observed that BCG Japan displayed the lowest rate of intracellular replication as compared to BCG Aventis and both MTB strains. 2. Materials and Methods {#sec2} ======================== 2.1. Mycobacteria {#sec2.1} ----------------- Two pathogenic strains of MTB were chosen to infect DCs: the laboratory H37Rv strain (ATCC N° 27294) and the clinical isolate CMT97. CMT97 was isolated at the Monaldi Hospital, Naples, Italy, from a TB patient\'s sputum \[[@B20]\]. Both strains were transferred every two months to Sauton\'s medium, allowing them to grow as a layer on the medium surface. Mycobacterial layers were harvested every two months, spun down, and resuspended in phosphate-buffer saline (PBS). To get an homogeneous suspension, mycobacteria were placed in glass tubes and sonicated in a bath sonicator (UST; 20 kHz) at the maximum power of 50 W. Samples were than aliquoted and stored at −80°C. To titrate mycobacteria, few aliquots were thawed and grown on 7H10 Middlebrook plates (Becton Dickinson, Franklin Lakes, N.J.). The same frozen master batch was used for each infection experiment. As regards to BCG strains employed, we chose BCG Aventis Pasteur seed Merieux (BCG Aventis), derived from strain 1077 (Aventis Pasteur SA 2 Avenue Pont Pasteur, F-69007 Lyon), and 3-1-5, Japan BCG Laboratory Matsuyama, Kiyose-shi, Tokyo (BCG Japan). Both of them were supplied freeze-dried and intended for live inoculation. They were reconstituted according to the manufacturer\'s instructions. 2.2. Generation and Infection of DCs {#sec2.2} ------------------------------------ Peripheral blood mononuclear cells (PBMC) were isolated, by Lympholyte-H (Cederlane, Canada) gradient centrifugation, from peripheral blood of healthy donors drawn from healthy volunteers. Monocytes were selected by anti-CD14-coated magnetic beads (MACS, Milteny Biothec, Germany). The purity was \>98%, as verified by flow cytometry analysis. Cells were plated at 1.5 × 10^6^ cell/mL in RPMI-1640 medium (EuroClone, UK) supplemented with 10% heat-inactivated fetal bovine serum (FBS, HyClone, UT), L-Glutammin 2 mM, HEPES buffer 10 mM, sodium piruvate 1%, gentamicin 5 *μ*g/mL (all EuroClone, UK) adding GM-CSF 50 ng/mL and IL-4 10 ng/mL (all EuroClone, UK) to allow them to differentiate into DCs. After 5 days of differentiation, immature DCs derived from the same donor were exposed to MTB H37Rv, MTB CMT97, BCG Aventis Pasteur, and BCG Japan at a multiplicity of infection (MOI) of 1 bacterium per cell or left uninfected. Before infection bacilli were sonicated to disrupt small aggregates of bacteria. After 3 h of incubation at 37°C all DCs (both infected and uninfected ones) were harvested and centrifuged at 800 rpm for 10 min to spin down the DCs and leave extracellular bacteria in the supernatant. DCs were resuspended in fresh complete and, at the indicated time points, cells were collected and supernatants were stored at −80°C. For all the infections we chose a MOI of 1 since this condition did not result in rapid cell death and allowed us to culture DCs for at least 7 days after infection. 2.3. Flow Cytometry and Monoclonal Antibodies {#sec2.3} --------------------------------------------- The following monoclonal antibodies (mAbs), directly fluorochrome conjugated, were used for flow-activated cell sorting (FACS) analysis: anti-CD1a, anti-CD25, anti-CD80, anti-CD86, anti-CD83, anti-HLA-DR, anti-CD14, anti-CD11c, and anti-HLA-A,B,C. Negative controls were isotype-matched mAb (all from Becton Dickinson Biosciences). To determine surface cell phenotype, 24 h after infection, uninfected and infected DCs were washed in assay buffer (PBS, 0.5% BSA and 0.1% sodium azide), incubated with the above described mAbs for 15 min at +4°C, washed and then analyzed by flow cytometry. 2.4. Mycobacterial Enumeration by CFU Determination {#sec2.4} --------------------------------------------------- Following infection, cells were washed and resuspended in fresh complete medium. At the indicated time points infected DCs were incubated for 30 min with 500 *μ*L of lysis buffer (PBS, 0.1% Saponin), diluted in PBS, 0.01% Tween 80, sonicated and plated as 50 *μ*L droplet on 7H10 plates in triplicates, at different dilutions. The CFU were checked after 21 days of dish culture at +37°C in a 5% CO~2~ incubator. 2.5. cDNA Arrays {#sec2.5} ---------------- Total RNA was extracted from both the uninfected DCs and the infected cells, 24 h after infection. RNA was extracted with 4 M Guanidine iso-thio-cyanate single-step method \[[@B21]\]. The extraction was performed on an RNase-free bench, in a separate room. Absorption spectroscopy was used to measure the purity and concentration of RNA with an A~260/280~ ratio of 2.0 indicating highly purified RNA. A total RNA sample (1.5 *μ*g) was reverse transcribed using the Ampolabellig Kit (Superarray Bioscience Corporation) according to the manufacturer\'s instruction in the presence of \[*α*-^33^\] P for the generation of radio-labeled cDNA probes. The probes were used to hybridize human Dendritic and Antigen Presenting Cell Gene Arrays (GEArray, S series; Supearray Bioscience Corporation) according to the manufacturer\'s instruction. Hybridization signals were detected by Phosphor-imager Thyphoon (Molecular Dynamics) and analyzed by Array Vision 7.0 software (Imaging Research Inc., Canada). Pathogen stimulations were repeated in three donors. Our data were expressed as normalized density (nDens) of each spot, corresponding to the density value of the spot minus background density and expressed as a multiple of the reference density value. The threshold of 0.001 was attributed to any value ≤0.001. We considered upregulated (+) and downregulated genes (−) only those ones that showed at least a twofold change in the level of RNA expression of the infected *versus* uninfected DCs in two of three independent experiments (fold change ≥2). Differences in the expression were calculated by dividing the gene nDens of infected cells by uninfected cells nDens. 2.6. Quantitative Real-Time Reverse Transcriptase-PCR (q-rt RT-PCR) {#sec2.6} ------------------------------------------------------------------- One *μ*g of total RNA, treated with DNAse I Amplification Grade (Invitrogen, Paisley, UK) was reverse transcribed using random examers and SuperScript III Reverse Transcriptase (Invitrogen, Paisley, UK), according to the manufacturer\'s instruction. For quantification of PCR products ABI PRISM 7000 SDS was used (Applied Biosystem, Foster City, USA). The RealMasterMix SYBR ROX (Eppendorf AG, Germany) was used to produce fluorescently labeled PCR products, and we monitored increasing fluorescence during repetitive cycling of the amplification reaction. For all primers, the following temperature cycling profile was used: 2 min a +50°C and 10 min at +95°C followed by 30 sec at Ta, and 1 min at +68°C for 40 cycles. Primer sequences are reported in [Table 1](#tab1){ref-type="table"}. L34 were used as an internal control to normalize total RNA amounts. 2.7. Cytokine Determination by ELISA {#sec2.7} ------------------------------------ DCs supernatants were collected at the end of the 24 h of culture and stored at −80°C. The amount of IL-10, IL-12 and TNF-*α* levels was evaluated by ELISA (Pierce Endogen, Woburn, MA) according to manufacturer\'s instruction. 2.8. Graphical, Statistical and Cytofluorimetric Analysis {#sec2.8} --------------------------------------------------------- The cytometric analysis was performed on a FACScalibur flow cytometer (BD Biosciences) and data were analyzed using CellQuest software (BD Biosciences). GraphPad Prism 4 (Graphpad software, San Diego, CA) was used for graphical and statistical analysis. Statistical significance was assessed by using analysis of variance (ANOVA), followed by Bonferroni\'s *post hoc* test; differences were considered significant at *P* \< .05. 3. Results {#sec3} ========== 3.1. Phenotype Analysis of Monocyte-Derived DCs Infected with Different Mycobacteria {#sec3.1} ------------------------------------------------------------------------------------ In order to analyze DCs activation following infection with different mycobacterial strains, we evaluated the ability of MTB H37Rv, MTB CMT97, BCG Aventis and BCG Japan to induce the expression of maturation markers on DCs membrane. MTB H37Rv and MTB CMT97 were selected as both pathogenic: the first is the commonly used laboratory strain of MTB, while the second is a clinical isolate from a TB patient\'s sputum, displaying a peculiar clinical picture \[[@B20]\] and a specific ability to activate human mcrophages in comparison with MTB H37Rv \[[@B17]\]. We also chose two commonly employed vaccine strains: BCG Aventis and BCG Japan. DCs derived from the same donor were exposed to the different mycobacterial strains for 3 hours at a MOI of 1 bacteria per cell or left untreated, and, after 24 hours, infected and uninfected DCs were harvested, washed, and analyzed by flow cytometry for typical marker profiles. As shown in the representative histogram plots of [Figure 1](#fig1){ref-type="fig"}, MTB H37Rv and MTB CMT97 as well as BCG Aventis and BCG Japan were all efficient in stimulating DCs to undergo maturation when compared to uninfected DCs. This was assessed by the upregulation of activation markers such as HLA-ABC, HLA-DR, CD80, CD86, CD83, and CD25. The results obtained revealed that all the selected mycobacteria were found to induce phenotype maturation of DCs with a comparable efficacy. 3.2. Cell Vitality of DCs Following Stimulation with Different Mycobacterial Strains {#sec3.2} ------------------------------------------------------------------------------------ We further examined whether cell vitality, could be affected when comparing DC cells infection with the two pathogenic MTBs and the BCG vaccine strains. The viability of the DCs was evaluated, at different time points, in terms of percentage of living cells. In [Figure 2](#fig2){ref-type="fig"} we showed that the cell yield, 1-day-culture after mycobacterial exposure, appeared diminished as an effect of the infection and that the reduction was comparable in all the infections. Recovery of infected DCs remained almost stable for the following 4 days, when alive cells started to decrease with a comparable trend. On the contrary, the number of alive, uninfected DCs remained almost stable for 1 day, then, as expected, unstimulated immature DCs showed a reduction in viability as compared to all the infected DCs at 3, 5, and 7 days. Taken together, these data indicate that all the mycobacterial strains did not produce cell death differently in 7-day-culture after infection. 3.3. Mycobacterial Intracellular Growth {#sec3.3} --------------------------------------- To ascertain whether bacterial burden could be comparable when DCs were infected with different BCG and MTB strains, we monitored the number of intracellular mycobacteria over 7-day culture following infection. For the laboratory strain MTB H37Rv and the clinical isolate MTB CMT97, we found an increase in mycobacterial counts, and a comparable rate of growth was assessed in DCs infected with BCG Aventis ([Figure 3](#fig3){ref-type="fig"}). Specifically, enumerated colony forming units (CFU) displayed a 4-fold increase after seven days of infection. On the contrary, the intracellular ability to replicate turned out to be greatly impaired when counting the intracellular BCG Japan CFU. In fact, the mycobacterial number remained almost steady over all the monitored period and resulted statistically reduced at the 7th day of infection, as compared to the other three mycobacteria. These results suggest that DCs differently contain the intracellular growth of the mycobacteria strains analyzed. 3.4. Gene Expression Profile of DCs Infected with Different MTB and BCG Strains {#sec3.4} ------------------------------------------------------------------------------- To test whether the mycobacterial strains we selected could differently modulate the overall gene expression of DCs, we analyzed the expression of 165 genes involved in DC activation and maturation using macroarrays. We analyzed and compared the expression profile of DCs exposed for 3 h to MTB H37Rv, MTB CMT97, BCG Aventis and BCG Japan, 1 day-culture after the infection. The gene modulation, showed in [Table 2](#tab2){ref-type="table"}, was confirmed in at least two of three independent assays, where each assay was performed on DCs derived from the same donor. In general, 30 genes were detected as upregulated (+) (fold change ≥2 in infected DCs *versus* uninfected DCs), corresponding to 18% of the 165 spotted onto the membrane ([Table 2(a)](#tab2){ref-type="table"}), and 33 genes out of 165 (20%) were detected as downregulated (−) ([Table 2(b)](#tab2){ref-type="table"}). In more detail, only 17 out of 30 genes proved to be comparably upregulated in all infected DCs while the other 13 genes resulted induced upon the encounter of DCs with some of the four mycobacteria. Analogous results were obtained when analyzing the downregulated genes. In particular, despite the fact that all the different mycobacteria decreased the expression of 25 gene, other 8 genes resulted repressed only following some of the infection performed. As expected, among the commonly upregulated genes we found cytokines involved in proinflammatory immune response such as IL-1*β*, IL-6, IL-12 and TNF-*α*, the regulatory IL-10 and chemokines and their receptors able to trigger DCs migration to lymph nodes (ADAM19, CCR7 and CCL20). On the other hand, we found, as commonly downregulated, genes that are substantially involved in antigen capture, loading and presentation as DC-SIGN, that is known to be responsible for DCs-T interaction. In this group we also found IL-18, which plays an important role in enhancing IFN-*γ* production by T cells \[[@B22]\], which proved to be preferentially downregulated in BCG Japan-infected DCs. In order to exactly appreciate gene expression and modulation we showed, in [Table 3](#tab3){ref-type="table"}, nDens of all genes differentially regulated, following mycobacteria exposure. Moreover, we indicated nDens of 2 genes induced and 2 genes repressed in all infection performed and nDens of all genes discussed above (IL-1*β*, IL-6, IL-12, TNF-*α*, IL-10, ADAM19, CCR7, CCL20 and IL-18). In conclusion, the macroarray analysis suggests that different mycobacterial strains such as MTB H37Rv, MTB CMT97, BCG Aventis and BCG Japan induce, on DCs, a pathogen-specific response, both in upregulated and downregulated genes, in the face of a common set of gene that appear to be similarly modulated. 3.5. Expression of Selected Genes by q-rt RT-PCR {#sec3.5} ------------------------------------------------ The mRNA levels of 12 selected genes were further analyzed by quantitative real-time PCR using gene specific primers ([Figure 4](#fig4){ref-type="fig"}). Six of these genes (CD83, CCL20, CCR7, IL-10, IL-12, and TNF-*α*) proved to be similarly upregulated in response to all the four mycobacterial strains but, due to their important role in the modulation of immune response, we further analyzed their expression in order to appreciate some possible quantitative difference. We also chose CCL19, c-FLIP (CFLAR), EBI3, and IL-18, which appeared differently modulated following the four infections, in order to verify such dissimilarity. For its well-documented pivotal role in MTB host response, we also decided to include in this analysis IFN-*γ*, even if it was present in the array and did not show any significant modulation in DCs following infection. We also included in q-rt RT-PCR analysis IL-32, absent in macroarray, since this has been recently shown to be involved in antitubercular immunity \[[@B23]\], Moreover, it has been described that mycobacterial species such as *M. tuberculosis* or *M. bovis* BCG are potent stimuli for the production of the proinflammatory cytokine IL-32 and its production is dependent on endogenous IFN-*γ* \[[@B24]\]. The result obtained confirmed the upregulation of CCL20, IL-10, IL-12, previously observed with the macroarray assay, in DCs exposed to the MTB as well as the BCG strains. Also the similar upregulation of CD83, already described at the protein level in [Figure 1](#fig1){ref-type="fig"}, was confirmed. Interestingly, the q-rt RT-PCR pointed out a significantly higher mRNA transcript of CCR7 and TNF-*α* in BCG Japan-infected-DCs, as compared to MTB H37Rv- and MTB CMT97-infected DCs (*P* \< .01 and *P* \< .05, resp.). Concerning CCL19, c-FLIP and IL-18, whose expression resulted different in DCs after the encounter with the two MTB and two BCG strains ([Table 2](#tab2){ref-type="table"}), when we reanalyzed the mRNA levels by quantitative real-time PCR, the overall expression of these genes was comparable among the different infections although a certain degree of variability was observed. In agreement with macroarray analysis, EBI3 expression resulted clearly induced in MTB H37Rv- and MTB CMT97-infected DCs, but a lower RNA induction was observed also in BCG infected DCs. In addition, a statistical analysis of EBI3 modulation underlined a consistent upregulation of this gene only in MTB H37Rv infected-DCs as compared to BCG Japan infected-DCs. In line with the modulation of EBI3, the upregulation of IFN-*γ* and IL-32 resulted to be stronger in MTB H37Rv and MTB CMT97 infected-DCs but no statistically significant differences can be described. Collectively, these findings, consistent with the macroarray data, allow us to describe a different modulation of the selected genes, thanks to the higher sensitivity of the quantitative real-time PCR as compared to the macroarray assay. Furthermore these data suggest that the maturation of DCs, following the infection with different MTB and BCG strains, may result in a different modulation of some genes importantly involved in the response against mycobacteria. 3.6. Release of IL-10, IL-12 and TNF-*α* from Infected and Uninfected DCs {#sec3.6} ------------------------------------------------------------------------- DCs have the unique capacity to stimulate naïve T lymphocytes driving these cells into a distinct class of effector cells. To investigate whether the infection with different mycobacterial strains could result in a different cytokine release from DCs, cell supernatants were collected 24 h after infection and the presence of IL-10, IL-12 and TNF-*α* was quantified by ELISA. We chose IL-10 and IL-12 since they are crucial for driving Th-1/Th-2 response and TNF-*α*, that already proved differently modulate at the RNA level, for its well-documented protective role against MTB and for its ability to mature DCs \[[@B25], [@B26]\]. As shown in [Figure 5](#fig5){ref-type="fig"}, a significantly higher release of IL-10 was observed from both BCG-infected and MTB CMT-infected DCs as compared to uninfected cells even though DCs exposed to MTB H37Rv also release a discrete amount of this regulatory cytokine. Concerning IL-12, a consistent and significant production was found only in BCG Aventis-infected DCs while, in the other 3 infections, only a mild increase in the level of IL-12 was observed as compared to control DCs. Finally, the TNF-*α* released and accumulated in the supernatant for 24 h after infection resulted high and similar in all the infections, as compared to uninfected DCs (*P* \< .001). 4. Discussion {#sec4} ============= MTB is an extremely well adapted pathogen that coexisted with the human host for thousands of years and during this period it has learnt how to modulate potentially protective host responses, to ensure its own survival. H37Rv is the currently used MTB laboratory strain and, considering how important it is to have a good *in vitro* model, it could be necessary to assess if H37Rv is able to induce a host response at least comparable to a clinical isolate. In fact, it has been clearly shown that different MTB clinical isolates have distinct effects and produce a different response that depends on their specific virulence \[[@B27], [@B28]\]. We decided to include in our study the strain CMT97, a clinical isolate previously characterized in infected cells of bronchial lavage fluid and in human infected macrophages \[[@B17], [@B20]\]. In parallel, BCG has emerged as a vaccine that changed from the original Pasteur strain into a range of strains that immunized many people around the world, with variable results. In this context, a characterization of a strain-specific host response and in particular a critical analysis and comparison between two widely used BCG strains could lead to a more rational approach towards the improvement of the BCG vaccine. In the present study, we investigated the DCs response to mycobacterial infections and we chose two different MTB and BCG strains: the laboratory strain H37Rv, CMT79, a clinical isolate from a TB patient, and two vaccine frequently used nowadays (BCG Aventis and BCG Japan). Several studies have been published dealing with the DCs response to different mycobacteria \[[@B10], [@B27], [@B28]\] but this is the first time that two different BCGs, an MTB clinical isolate and a laboratory MTB strain are compared simultaneously for the activation induced on human monocyte-derived DCs coming from the same donor. First, we compared the selected mycobacteria for their ability to replicate inside DCs and, interestingly, it emerged that only BCG Japan was unable to enlarge the intracellular bacterial population in infected DCs. This suggests, in accordance with *in vivo* previous studies \[[@B29]\], that during the initial phase of infection with MTB H37Rv, MTB CMT97 and BCG Aventis, the growth of the bacterial population could be accompanied by an increment of the number of infected cells, which implies a high rate of cell-to-cell spread of the mycobacteria, while the cell-to-cell spread outcome following the infection with BCG Japan may possibly occur to a lesser extent. Of note, although DCs contained differently the intracellular growth of BCG Japan compared to the other three mycobacteria, cell expression of activation markers as well as cell recovery proved to be comparable unrelated to the infecting mycobacteria. This supports the fact that DCs differently contain the intracellular growth of the mycobacteria considered, while cells similarly survive after all the infections performed. It\'s important to keep in mind that, since 1921, the *in vitro* attenuation of *M. bovis* gave rise to a large number of BCG daughter strains, which have been classified as "early strains" and "later strains". The BCG Japan is an "early strain" while BCG Aventis Pasteur belongs to the "later strains", obtained before the loss of the RD14 region \[[@B9], [@B30]\]. Considering this, it is not surprising that different BCG strains could have such a different adaptive feature and that the differences existing between the BCG daughter strains may influence the activation of infected DCs. To explore globally the host gene expression differences we performed a DCs comparative *in vitro* transcriptome analysis across the two MTB and two BCG strains. The analysis of 165 genes involved in cell maturation and activation showed that 18% of these proved to be upregulated and 20% were detected as downregulated. More in detail, 17 out of 30 genes resulted to be comparably upregulated in all infected DCs and resulted to be involved in proinflammatory immune response (Il-1*β*, IL-6, IL-12 and TNF-*α*), immune regulation and migration (IL-10, ADAM19, CCR7 and CCL20) showing that all mycobacteria were able to elicit the expression of many genes already described as upregulated in common pathogen response \[[@B10]\]. The other 13 genes resulted differently induced and among these we chose CCL-19, C-Flip and EBI3 to be further investigated by q-rt RT-PCR. Accordingly with the macroarray analysis, EBI3 resulted preferentially induced in the DCs infected with MTB virulent species than BCG a-virulent strains. Interestingly, when comparing the gene-fold induction across the four mycobacteria, H37Rv displayed a significantly higher EBI3 transcription as compared to BCG Japan. EBI3 is a subunit of IL-27 that, together with IL-12, is involved in driving commitment of naïve T cells to a Th-1 phenotype \[[@B31], [@B32]\]. Intriguingly, it has been recently published that neutralization of IL-27 reduced, even if modestly, viable MTB recovered from macrophages \[[@B33]\]. In this scenario, the lower EBI3 induction in BCG Japan-infected DCs might be involved in BCG Japan failing to grow exponentially in DCs. The comparison, by q-rt RT-PCR, of gene-fold induction across the four mycobacteria also revealed that the proinflammatory TNF-*α* and the migratory chemokine receptor CCR7 were strongly induced in BCG Japan-infected DCs as compared to cells infected with both MTB strains. This suggests that BCG Japan might prove particularly efficient at promoting DCs migration into T cell-enriched areas of lymphoid tissue, where DCs are able to present Ag-derived peptides, associated with either class I or class II MHC molecules to naïve CD4 and CD8 T cells, respectively \[[@B34]\]. Moreover, previous studies have shown that TNF-*α* is fundamental in granuloma formation and maintenance and also affects cell migration upon MTB infection, since it influences the expression of adhesion molecules as well as chemokines and chemokine receptors such as MIP-1*α*, MIP-1*β*, RANTES, and CCR5 \[[@B35]--[@B38]\]. When we assayed the release of TNF-*α* from supernatants, we observed that this proinflammatory cytokine was found to be abundantly released in all infected DCs. This result strongly supports the evidence that initial interaction of MTB and BCG induces proinflammatory cytokine production \[[@B27], [@B39]\]. As previously shown \[[@B40]\], we observed a consistent release of IL-10 from BCG-infected DCs even if these cells resulted to be well activated following mycobacteria exposure, as assessed by the upregulation of maturation markers. Also, MTB CMT97-infected DCs produced significantly higher levels of IL-10 as compared to uninfected cells while, MTB H37Rv induced the release of a discrete but not significantly amount of this cytokine. When we measured the release in the supernatant of IL-12, a cytokine pivotal in directing the polarization of immune response, we observed that despite the fact that a RNA upregulation was described in all MTB and BCG infections, only BCG Aventis induced a significant IL-12 production, while BCG Japan and both MTB strains turned out to be a stimulus insufficient for the induction of IL-12 releasing DCs. The data obtained, at a protein level, confirm the previously reported ability of MTB H37Rv and different MTB clinical isolates to suppress the secretion of IL-12 by monocyte-derived DCs \[[@B27]\] and, also suggest another possible difference in the immune activation induced by the two different BCG analysed. Of note, even if there is an evident discrepancy between RNA analysis and protein detection from supernatants, it should be remembered that several experimental studies have shown that mRNA changes do not necessarily correlate with changes in the corresponding proteins, which are the ultimate determinants of cellular function \[[@B41]\]. The invalid assumption of the one-to-one correlation between the ratios of protein levels and the corresponding mRNAs \[[@B42], [@B43]\] suggests that the relation between transcription and translation, and consequently between mRNA and protein, is complex. Moreover, qRT-PCR shows what is happening in a particular moment, while the ELISA assay we performed gives us the total amount of protein, released in the course of 24 hours following infection. In conclusion, the importance of DCs in initiating immune response gave reason to investigate if these cells could discriminate between a clinical and a laboratory MTB strain and also if DCs could sense differently two distinct BCG strains. The analysis of the individual response showed that DCs exhibit stimulus-specific maturation and activation. In fact, besides a shared core reprogramming, DCs are able to modulate the expression of exclusive genes, proving that these immune regulating cells are able not only to discriminate between phylogenetically distinct pathogens \[[@B10]\] but also, elicit a specific response as respect to diverse MTB and BCG strains. The results obtained describe the contribution to pathogen-host interaction of strictly correlated mycobacterial strains. In particular, our data indicate that the specific differential response to MTB H37Rv and MTB CMT97 mimics the previously reported different cellular response to distinct clinical MTB isolate, dependently to their specific virulence \[[@B27], [@B28]\] supporting the validity of MTB H37Rv as an *in vitro* model for mycobacterial-DCs interaction. On the other side, the strain-specific modulation observed in response to BCG Aventis and BCG Japan as well as the different ability of these mycobacteria to growth in infected DCs lead to important physiological consequence that must be considered and further studies of BCG-regulated genes may thus enhance our understanding of DCs maturation and provide future indication for the design and improvement of a vaccine against tuberculosis. This study received financial support from PRIN 2007ECX29E and from FILAS, Progetti di ricerca industriale e sviluppo sperimentale, "TB VACCINE" Sviluppo di un nuovo vaccino contro la Tubercolosi, 2009. The authors would like to thank Dr. Silvia Vendetti for her critical reading of the manuscript and helpful suggestion. {#fig1} {#fig2} {#fig3} {#fig4} {#fig5} ###### Primer sequences used in the q-rt RT-PCR. Gene Sense Antisense Annealing temp. (°C) --------- ----------------------------------- ----------------------------------- ---------------------- L34 5′-GGCCCTGCTGCATGTTTCTT-3′ 5′-GTCCCGAACCCCTGGTAATAGA-3′ 64 EBI3 5′-AGAGCACATCATCAAGCCCGAC-3′ 5′-TCCCTGACGCTTGTAAGCGCATC-3′ 64 CCR7 5′-AAAAGCGTGCTGGTGGTGGC-3′ 5′-ATGATAGGGAGGAACCAGGC-3′ 64 TNF-*α* 5′-AGGCGGTGCTTGTTCCTCA-3′ 5′-GTTCGAGAAGATGATCTGACTGCC-3′ 60 CCL19 5′-CCAATGATGCTGAAGACTGC-3′ 5′-CTGGATGATGCGTTCTACCC-3′ 62 CCL20 5′-TGAAGGCTGTGACATCAATGC-3′ 5′-TGTTTTGGATTTGCGCACAC-3′ 60 IL-12 5′-GCTGCTGAGGAGAGTCGTCCC-3′ 5′-CCAGCTGACCTCGACCTGCC-3′ 62 IL-10 5′-AGGCGCATGTGAACTCCCT-3′ 5′-CACGGTCTTGCTCTTGTTTT-3′ 64 IFN-*γ* 5′-GGCTGTTACTGCCAGGACCCCATATGT-3′ 5′-GATGCTCTTGCACCTCGAAACAGCCAT-3′ 64 IL-32 5′-GACATGAAGAAGCTGAAGGCC-3′ 5′-ATCTGTTGCCTCGGCACCG-3′ 62 IL-18 5′-GACAATTGCATCAACTTTGTGG-3′ 5′-ATAGAGGCCGATTTCCTTGG-3′ 62 c-FLIP 5′-TTCATGGGAGATTCATGCCC-3′ 5′-AAGAGGCTGCTGTCCTCCA-3′ 60 CD83 5′-AGGTTCCCTACACGGTCTCC-3′ 5′-TTGAAGCTGGTAGTGTTTCG-3′ 60 ###### \(a\) Upregulated genes of DCs infected with different MTB and BCG strains. (b) Downregulated genes of DCs infected with different MTB and BCG strains. ###### \(a\) Gene MTB H37Rv MTB CMT97 BCG Aventis BCG Japan Common name Classification Function ---------- ----------- ----------- ------------- ----------- --------------- ------------------------------ -------------------------- ADAM19 \+ \+ \+ \+ Metalloproteinase Migration/Inflammation BASP1 \+ \+ \+ \+ CAP-23, NAP-2 Cell surface protein Prot-prot interaction CCL5 \+ \+ \+ \+ RANTES Chemokine ligand Migration CCL20 \+ \+ \+ \+ MIP-3*α* Chemokine ligand Inflammation CCL22 \+ \+ \+ \+ MDC Chemokine ligand Inflammation CCR7 \+ \+ \+ \+ Chemokine receptor Migration CD83 \+ \+ \+ \+ Cell surface protein Ag presentation IL-1*β* \+ \+ \+ \+ Cytokine Inflammation IL-6 \+ \+ \+ \+ Cytokine Inflammation IL-10 \+ \+ \+ \+ Cytokine Immune regulation IL-12p40 \+ \+ \+ \+ Cytokine T cell stimulation LY75 \+ \+ \+ \+ DEC-205 Cell surface receptor Ag presentation NFKB1 \+ \+ \+ \+ Trascriptional factor Signal transduction SOD2 \+ \+ \+ \+ Mytocondrial protein Oxidative stress TAP2 \+ \+ \+ \+ Ag transporter 2 Ag presentation TLR2 \+ \+ \+ \+ Toll-like receptor Pathogen assoc. receptor TNF-*α* \+ \+ \+ \+ Cytokine Inflammation ADAR \+ \+ Adenosine deaminasi RNA modification CCL3 \+ \+ \+ MIP-1*α* Chemokine ligand Inflammation CCL19 \+ \+ Cell surface receptor Migration CD80 \+ Cell surface protein Ag presentation CD86 \+ \+ \+ Cell surface protein Ag presentation CFLAR \+ \+ C-Flip Apoptosis regulator Apoptosis inhibitor CRF \+ C1q related factor EBI3 \+ \+ Secreted glicoprotein Immune-regulation IL-1a \+ \+ \+ Cytokine Inflammation LAMP3 \+ \+ \+ CD63 Lysosomal associated protein Ag capture MT2A \+ \+ Metallothionein Oxidative stress PLAUR \+ \+ CD87 Cell surface receptor Migration TNFRSF6 \+ \+ CD95, FAS Cell surface receptor Maturation ###### \(b\) Gene MTB H37Rv MTB CMT97 BCG Aventis BCG Japan Common name Classification Function ---------- ----------- ----------- ------------- ----------- ----------------- --------------------------------- ---------------------------- ARHGDIB − − − − RhoGDP dissociation inhibitor 2 Cell motility and adhesion CD1A − − − − MHC I like protein Ag presentation CD1B − − − − MHC I like protein Ag presentation CD1C − − − − MHC I like protein Ag presentation CD36 − − − − Cell surface protein Ag capture CD68 − − − − Cell surface protein Ag capture CD74 − − − − Invariant chain MHC II assoc. protein Ag loading CD209 − − − − DC-SIGN C-type lectin receptor T/DCs interaction CLECSF12 − − − − DECTIN-1 Cell surface receptor Pattern recog. receptor CLECSF6 − − − − DCIR C-type lectin receptor Ag capture CST3 − − − − Proteinase inhibitor CXCL16 − − − − Chemokine T cell stimulation FCER1A − − − − Fc receptor Inflammation FCER2 − − − − Fc receptor Inflammation GIP3 − − − − IFN-*α* inducible protein HLA-DMA − − − − MHC II accessory protein Ag loading HLA-DMB − − − − MHC II accessory protein Ag loading IFI16 − − − − IFN-*γ* inducible protein Cell cycle regulator IFITM3 − − − − IFN inducible protein Immune response ITGB2 − − − − Integrin Migration LANGERIN − − − − Cell surface protein Ag capture LIPA − − − − Lipase A Acid lipase Lipidic metabolism MX1 − − − − IFN-*α* inducible p78 Antiviral response RNASE6 − − − − Ribonuclease Ag presentation TLR4 − − − − Toll-like receptor Pathogen assoc. receptor CSF1R − − − CD115 Cell surface receptor M-CSF receptor DCP1B − − Decapping enzyme hDcp1b RNA degradation DCSTAMP − − Cell trasmemb. protein GBP3 − Guanilate binding protein GIP2 − − IFN-*α* inducible protein IL-18 − Cytokine Inflammation PFN1 − TLR6 − − Toll-like receptor Pathogen assoc. receptor Upregulated ([Table 2(a)](#tab2){ref-type="table"}) and downregulated ([Table 2(b)](#tab2){ref-type="table"}) genes of DCs exposed for 3 h to MTB H37Rv, MTB CMT97, BCG Aventis and BCG Japan as compared to uninfected DCs were assayed by macroarray 1-day culture after infections. For genes to be referred as upregulated or downregulated we considered only those that showed at least a twofold change in the level of RNA as compared to uninfected DCs expression, in at least 2 independent experiments of the 3 performed. ###### \(a\) nDens of upregulated genes of DCs infected with different MTB and BCG strains. (b) nDens of downregulated genes of DCs infected with different MTB and BCG strains. ###### \(a\) Gene MTB H37Rv MTB CMT97 BCG Aventis BCG Japan Uninfected DCs ---------- ----------------- ------------------- ------------------ ------------------ -------------------- ADAM19 0.3; 0.4; 1.5 0.3; 0.2; 1.8 0.2; 0.7; 0.1 0.3; 0.6; 1 0.2; 0.07; 0.5 CCL20 0.4; 1; 0.8 0.4; 2.5; 1.4 0.9; 2.3; 0.8 0.5; 0.8; 1.1 0.04; 0.001; 0.4 CCL22 8.5; 1.8; 11.1 6.9; 3.7; 12.4 4.3; 1.8; 12.1 8.1; 1.5; 9.8 0.3; 0.5; 0.2 CCR7 3.1; 0.001; 1.7 3.2, 0.9; 1.9 0.5; 1.8; 1.6 2.4; 0.9; 2.2 0.1; 0.2; 0.05 IL-1*β* 2.1; 2.9; 14.4 1.7; 10.5; 19.0 0.3; 6.5; 10.8 1.9; 4.8; 14 0.09; 0.9; 0.8 IL-6 0.5; 0.02; 2.9 0.1; 0.4; 3.4 0.001; 0.6; 2.3 0.1; 0.3; 3.2 0.01; 0.01; 0.1 IL-10 1.8; 3.7; 6.6 1.0; 1.2; 8.9 1.2; 1.3; 5.8 1.0; 0.9; 4 0.4; 0.3; 4.4 IL-12p40 0.6; 0.3; 0.9 0.6; 0.7; 0.7 0.7; 0.1; 1.0 0.4; 0.01; 1.5 0.1; 0.1; 0.004 SOD2 0.8; 1.6; 5.5 0.8; 3.4; 7.6 0.9; 1.5; 3.7 0.5; 1.5; 5.4 0.1; 0.7; 0.7 ADAR 0.4; 0.2; 0.2 0.2; 0.1; 0.1 0.3; 0.1; 0.1 0.4; 0.3; 0.2 0.2; 0.1; 0.1 CCL3 4.1; 7.0; 11.0 3.6; 6.5; 5.6 7.0; 4.6; 10.8 2.5; 3.7; 1.2 1.7; 3.4; 0.6 CCL19 0.1; 0.1; 0.001 0.3; 0.007; 0.001 0.2; 0.3; 0.4 0.2; 0.1; 0.5 0.001; 0.1; 0.001 CD80 3.3; 0.4; 0.2 3.2; 0.5; 0.2 3.9; 0.9; 0.2 5.8; 0.7; 0.9 2.1; 0.3; 0.2 CD86 0.3; 0.001; 0.1 0.2; 0.02; 0.4 0.1; 0.1; 0.4 0.2; 0.01; 0.4 0.05; 0.02; 0.2 CFLAR 0.4; 0.5; 0.3 0.5; 0.5; 0.3 1.1; 1.3; 0.2 0.8; 0.8; 0.4 0.5; 0.3; 0.2 CRF 0.3; 0.3; 0.001 0.3; 0.1; 0.01 1.1; 0.2; 0.001 0.1; 0.02; 0.001 0.3; 0.1; 0.001 EBI3 0.1; 0.2; 0.5 0.1; 0.001; 0.7 0.01; 0.001; 0.2 0.01; 0.001; 0.3 0.01; 0.001; 0.008 IL-1*α* 0.02; 0.01; 0.4 0.1; 0.6; 0.5 0.3; 0.6; 0.3 0.4; 0.3; 0.9 0.02; 0.01; 0.001 LAMP3 5.1; 24.0, 1.6 10.6; 31.3; 2.1 13.4; 12.9; 2.3 21.8; 26.0; 1.7 9.3; 11.8; 0.7 MT2A 0.2; 0.1; 11.1 0.4; 0.5; 11.5 0.4; 0.2; 7.1 0.08; 0.1; 8.4 0.2; 0.2; 0.1 PLAUR 0.06; 0.1; 1.5 0.1; 0.7; 0.7 0.2; 0.4; 1.2 0.06; 0.1; 1.2 0.1; 0.1; 0.1 TNFRSF6 0.6; 0.1; 0.001 0.7; 0.006; 0.004 1.6; 0.4; 0.001 1.4; 0.2; 0.001 0.6; 0.001; 0.001 ###### \(b\) Gene MTB H37Rv MTB CMT97 BCG Aventis BCG Japan Uninfected DCs --------- ------------------- ------------------- ------------------- -------------------- ----------------- CD1A 0.1; 0.2; 0.001 0.1; 0.001; 0.02 0.06; 0.001; 0.05 0.2; 0.001; 0.001 5.2; 0.3; 1.1 CD1B 0.04; 0.4; 0.001 0.05; 0.001; 0.02 0.05; 0.001; 0.06 0.008; 0.001; 0.07 1.3; 0.001; 0.7 CD209 0.05, 0.1; 0.08 0.02; 0.03; 0.001 0.04; 0.001; 0.03 0.01; 0.2; 0.03 0.3; 0.1; 0.4 CSF1R 0.05; 0.4; 0.1 0.1; 0.5; 0.001 0.5; 0.2; 0.04 0.1; 0.1; 0.3 0.3; 0.2; 0.5 DCP1B 0.2; 0.001; 0.04 0.3; 0.001; 0.1 0.3; 0.1; 0.07 0.1; 0.07; 0.01 0.4; 0.1; 0.1 DCSTAMP 1.3; 0.001; 0.001 1.6; 0.001; 0.3 0.2; 0.05; 0.02 2.1; 0.01; 0.001 3.0; 0.1; 0.05 GBP3 0.3; 0.4; 0.1 0.08; 0.5; 0.03 0.1; 0.8; 0.05 0.1; 0.3; 0.04 0.1; 0.7; 0.1 GIP2 0.05; 0.001; 0.9 0.08; 0.3; 0.6 0.07; 0.4; 0.4 0.001; 0.1; 0.7 0.3; 0.3; 0.7 IL-18 0.1; 1.9; 0.2 0.1; 1.6; 0.2 1.1; 1.8; 0.2 0.1; 0.6; 0.2 0.2; 2.7; 0.2 PFN1 0.9; 0.5; 3.1 0.6; 0.4; 5.0 0.7; 0.6; 2.3 0.7; 0.2; 3.5 0.8; 0.5; 8.5 TLR6 0.05; 0.09; 0.001 0.08; 0.3; 0.03 0.1; 0.3; 0.001 0.001; 0.1; 0.001 1.4; 0.3; 0.02 RNA expression of infected and uninfected DCs of three independent experiments. Values are reported as nDens. [^1]: Academic Editor: Nathalie Winter

Twenty-two phlebotomine sand fly species (Diptera: Psychodidae) have been reported in Algeria, 12 belonging to the *Phlebotomus* genus and 10 to the *Sergentomyia* genus ([@R1]). Those included in the *Phlebotomus* genus are of medical importance since they comprise recognized or suspected vectors of leishmaniasis and/or Phlebovirus. We report here for the first time (i) the presence of *Phlebotomus mascittii* in Algeria, and (ii) the presence of the female *Phlebotomus chadlii* in the same area. The entomological investigation was conducted in Larbaa Nath Iraten (4° 12' 05'' E, 36° 38' 10'' N at 916 m altitude), in a humid bioclimatic zone, in Kabylian area ([@R6]). Sand flies collection was performed during summer 2009 using CDC miniature light traps. A total of 883 sand flies (703 males and 180 females) were captured and morphologically identified during 16 night-CDC traps (55 sand flies/night-CDC traps). Ten distinct species were identified: one species belonging to the *Sergentomyia* genus (*S. minuta*) and nine species to the *Phlebotomus* genus including one female of *P. mascittii* and two females of *P. chadlii* ([Table 1](#T1){ref-type="table"}). Table 1.Sand fly species diversity in LNI, Kabylian area during summer 2009).SpecieMaleFemaleTotal*P. (Larroussius) perniciosus*564115679*P. (Larroussius) longicuspis*8439123*P. (Larroussius) langeroni*24024*P. (Larroussius) perfiliewi*347*P. (Larroussius) ariasi*123*P. (Larroussius) chadlii*123*P. (Paraphlebotmus) sergenti*3710*P. (Phlebotomus) papatasi*11*P. (Transphlebotomus) mascittii*11*S. (Sergentomyia) minuta*221032Total703180883 *P. mascittii* was described in Italy (Roma), then in other countries in the north shore of the Mediterranean basin, from Spain to Turkey ([@R11]). In countries of northern Europe, it was reported in Germany and Switzerland ([@R7]). However, *P. mascittii* has always been found in low density. In France, *P. mascittii* species was observed in several departments including in the north, such as Alsace ([@R2]). In southern regions, it was usually associated with the main recognized vectors of visceral leishmaniasis, *P. ariasi* and *P. perniciosus* ([@R10]; [@R8]). It was described as an anthropophylic and aggressive species ([@R8]). *P. mascittii* was suspected to be a vector of Mediterranean leishmaniasis, because it was frequently collected from human and dog leishmaniasis in endemic foci ([@R8]). However, its vector role has not been confirmed so far. Hence, we noticed for the first time the presence of *P. mascittii* female ([Fig. 1](#F1){ref-type="fig"}) in the southern part of Mediterranean. This female was collected from animal shelter localized in house basement.Fig. 1.Spermathecae of *Phlebotomus (Transphlebotomus) mascitti* (photonic microscope × 200). *P. chadlii* was described from Northwest Tunisia (El Kef) among male sand fly specimens. However, the female remained unrecognized until 2006 when it was described in specimens trapped in El Kef, Tunisia ([@R3]). In Algeria, *P. chadlii* is widely spread in humid, sub humid and arid bioclimatic zones ([@R4]). For unknown reasons, in Algeria, only male specimens have been reported so far ([@R9]; [@R4]). In our survey, three specimens of *P. chadlii* were identified, two females ([Fig. 2](#F2){ref-type="fig"}) and one male, all of them cached in animal shelters. The bioclimatic distribution of *P. chadlii* coincides with that of *P. ariasi* (Dedet *et al.*, 1985), the proven vectors of *L. infantum* and of Sand fly Fever *Phleboviruses* (SFV) in the Mediterranean basin ([@R6]). Using the mitochondrial *cyt b* gene, [@R5] reported that *P. chadlii* might be a sister group of the European and the Moroccan *P. ariasi* species. However, to date there is no confirmation neither for their vector role, nor for their trophic preferences. The two females collected in this study were not engorged, thus precluding blood meal analysis. We strongly support the idea of further studies (i) to elucidate the relationship between *P. chadlii* and *P. ariasi*, (ii) to identify their trophic preferences, and (iii) to study the relationship host/leishmaniasis parasite.Fig. 2.Spermathecae of *Phlebotomus (Larroussius) chadlii* (photonic microscope × 200). Thanks to Doctor R. Benane, veterinarian in Larbaa Nath Irathen (LNI) region, for their helpful and availability to realize this study.

Elucidating the genetic architecture of a phenotypic trait fundamentally requires that it is, to some degree, heritable. Phenotypic traits with low heritability do not generally require additional explanation as to why quantitative trait loci (QTL) cannot be robustly identified. However, while traits with high heritability increase the power to identify causative loci, the genetic architecture of traits remains a key factor in predicting success in identifying QTL. For example, a highly heritable trait that is dependent on a large number of interacting alleles or loci may still require substantial sample sizes and complex analyses to identify causative loci ([@bib25]; [@bib5]). The archetypical example of this is human height, which, by at least two distinct measures of heritability, is among the most heritable of quantitative human traits, *h*^2^ ≈ 0.4--0.7 using regression of trios ([@bib6]; [@bib8]; [@bib25]) or *h*^2^ ≈ 0.7 using twin studies ([@bib20]). However, the identification of the loci involved, and any interactions among them, has proven far from trivial, and successes to date have been dependent on utilizing very large data sets \[up to 250,000 individuals ([@bib29])\]. The current understanding being that of the 180--4000 loci implicated in impacting human height over its typical range all have additive effect sizes of \<\<1 mm ([@bib28]; [@bib25]; [@bib29]). Although studies of human height represent a valuable model with which to test how complex heritable phenotypes can be dissected, the ability to experimentally control allele frequencies and environmental conditions would be a valuable capacity that is not possible in human studies. While this is a possibility in model organisms, there are relatively few morphological traits that have been robustly demonstrated to be both highly heritable and amenable to automated high throughput phenotyping (though see below). Herein, we describe in detail a new quantitative trait---length of the pupal case---in *Drosophila melanogaster* that places it in the 15--20% most heritable morphological traits described in this species, and is amenable to reliable and automated high-throughput phenotyping ([@bib22]). The utility of automated phenotyping systems for a wide variety of organisms has increased greatly over the last 10 yr. While automated morphological phenotyping systems have been developed for *Drosophila*, as yet only automation analysis of images of wings has been widely employed ([@bib9]). However, acquiring the wing images requires manual manipulation to position flies individually and so is difficult to scale up. Likewise, a system to measure heartbeat function requires that each fly be manipulated into position. An alternative approach to phenotyping in a selection experiment for gross body size was achieved using a series of graduated sieves ([@bib23]). More recently, a sophisticated platform called the "fly cat walk" was described that has the capacity to reliably phenotype 700 flies a day for a wide range of morphological traits simultaneously ([@bib16]). This requires no user manipulation of individual flies and is nondestructive. While the cost of constructing the equipment is not detailed (<https://github.com/IMSB/FlyCatwalk/>), it is likely that it would represent a significant investment of time, expertise, and resources for most laboratories. Our setup uses an inexpensive camera in a light-proof box and the open source image analysis software Cellprofiler, with which a single user can phenotype 5000 pupae in a day. Pupae are photographed *in situ* on flattened squares of transparent film that lined the entire vertical surface of the vials. We demonstrate how the system can be used to (1) measure panels of recombinant inbred lines (RILs), and (2) generate large numbers of parent offspring trios that are particularly useful in exploring complex traits using artificial selection techniques. Furthermore, the increased throughput facilitates exploring the heritability of family means rather than single individual measurements, which are associated with increased measurement variance. Hence, pupal size could become a model phenotype that will allow deep dissection of its genetic architecture, since the availability of automated phenotyping will allow to screen very large mapping panels, or to design new complex mapping strategies. Materials and Methods {#s1} ===================== Fly stocks {#s2} ---------- The automated phenotyping system was applied to two independent datasets. The first dataset, referred to here as "eight-way" is a collection of 195 RILs that are part of the *Drosophila* Synthetic Population Resource ([http://wfitch.bio.uci.edu/∼dspr/](http://wfitch.bio.uci.edu/~dspr/)). These RILs are all originally derived from a cross between eight global stocks of *D. melanog*aster, their generation is described in detail in [@bib11],[@bib12]). Narrow sense heritability, *h*^2^, for the eight-way dataset was estimated by measuring the progeny of 67 single pair matings between individuals from 11 different RILs that spanned the full range of phenotype measurements (IDs of crossed stocks are given in Supplemental Material, [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS5.zip); six of the 67 crosses are duplicates also indicated in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS5.zip)). Broad sense heritability estimates *H*^2^ of the eight-way dataset were generated by repeated measurements of RILs (IDs given in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS5.zip)). The second dataset, termed "four-way," was initiated by a cross between two Japanese and two African stocks (see [Table S1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/TableS1.pdf)). Narrow sense heritability, *h*^2^, was estimated by measuring the progeny of single pair matings from the 2nd to 6th generation where phenotyped parents were selected randomly from different vials to form subsequent generations. The following number of single pairs were measured in each generation \[G2, 15 pairs\], \[G3, 81 pairs\], \[G4, 78 pairs\], \[G5, 88 pairs\], and \[G6, 154 pairs\]. No duplicate vials were generated from the same mating pair. Note that, unlike the eight-way dataset, levels of heterozygosity in the parents and their offspring are likely to be similar. *H*^2^ estimates of the four-way dataset were generated by measuring 83 RILs established from the four-way cross by generation 37 (see [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS5.zip) for stock IDs). Details of all progenitor stocks of the four-way, and four-way datasets with estimates of their pupal length, are given in [Table S1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/TableS1.pdf). Image acquisition {#s3} ----------------- Flies were maintained on standard food dispensed into 28.5 mm diameter, 95 mm height vials (Genesee Scientific). Once the food vials had fully cooled, 10 cm × 10.5 cm squares of overhead projector film were slid into each vial lining their entire vertical wall (nobo, plain paper copier film, 33638237). A more detailed description of the entire procedure and equipment set up is provided as [File S1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS1.docx). A custom printed semitransparent label, including a unique barcode, was affixed to the outside of each vial. Vials were incubated at 24° in a 12 hr light/12 hr dark incubator. Adults were removed after one to two nights in vials (sometimes three to four nights, if fertility appeared to be low, or due to holidays). Generally, by the 10th day after the parents were initially introduced, the majority of offspring in the vials were present as pupae attached to the transparent film; few if any larvae remained in the food. The film from each vial was removed, and placed into a purpose-made plastic frame (this frame can be 3D printed using a file provided as [File S2](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS2.zip)) that holds the film flat for photographing. Food from the lower part of the film was scraped away and any larvae, or any white puparium stage (P1), removed. The frame was then photographed using bottom illumination in a light tight box. Batches of the resulting images were then analyzed using the procedure below. Automated image analysis {#s4} ------------------------ A Cellprofiler (v2.1.0) pipeline was developed to simultaneously recognize pupae and to measure a variety of attributes, including length. The outputs of all measurements are written to .xls files that can then be viewed or imported into any database program. Cellprofiler is free, open access, software providing a suite of flexible image analysis tools ([@bib14]). In brief, the Cellprofiler pipeline first identifies "primary objects" distinct from the background without restriction on their size (module: identify primary objects). Then, applying a scalable model of pupal shape to all objects, those that are composed of multiple touching pupae are separated into distinct pupae \[module: Untangle Worms, ([@bib27])\]. The resulting putative pupae are then each shrunk, and then repropagated outwards to more precisely identify the edges of each pupa based on boundary changes in pixel intensity (module: Identify Secondary objects). Finally, pupae are crudely filtered on size attributes and the proximity of neighboring objects to place them in one of the three confidence classes described in the *Results*. The digital outlines of pupae are overlaid onto a cropped version of the original image to allow users to easily visually assess quality. Overlaid images and files of measurements were imported in batches into the database program FileMaker (v14, FileMaker Inc.). A unique barcode sticker identifying the film in the image was automatically read by the database. This enabled the image files to be automatically renamed with the barcode as their name for archiving. In addition to pupal measurements, a 1 €cent coin (16.25 mm diameter) present in all images was measured to control for camera changes, and to allow conversion of measurement in pixels to millimeter. Quality filtering of pupae and basic analyses were performed using the database. An annotated copy of the Cellprofiler pipeline is provided as [File S3](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS3.zip). Estimates of heritability {#s5} ------------------------- *H*^2^ was estimated using SPSS version 22 with the "variance components" function, with pupal length as the dependent variable, and RIL name as a random factor. This was done using the Minimum Norm Quadratic Unbiased method (though ANOVA produced identical results to two decimal places). The model for the single factor was *y~ij~* *=* *μ* + *α~j~* *+e~ij~*, where *y~ij~* is the *j*th observation of the *i*th RIL, *μ* is the overall mean, *α~j~* is the random factor, and *e* is the associated error. Estimates of *h*^2^ were made using the "linear regression" function, with pupal length as the dependent variable, and parent-midpoints as an independent variable. All other statistical analyses and graphs were also performed using SPSS or Filemaker. All data for individual pupa are available as [File S4](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS4.zip), and for vial means as [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS5.zip) (fields used in each graph are indicated in the associated "readme" files). Human height data {#s6} ----------------- Estimates of *h*^2^ for human height data were made using the same methods as for pupae, using the original data of Francis Galton for 898 individuals, transcribed from his 1880s laboratory notebooks ([@bib8]) (available <http://www.math.uah.edu/stat/data/Galton.html>). The 760 individuals in 123 families where four or more individuals were measured were used for estimates of the *h*^2^ heritability of mean family height (females were not transmuted into males). Genome scan for loci of large effect impacting pupal length using eight-way RILs {#s7} -------------------------------------------------------------------------------- Scans were performed in R (version 3.2.2) for the 195 DSPR RILs mentioned above ([@bib11],[@bib12]) using the "DSPRscan" command within the DSPRqtl-tools procedures described in [@bib11],[@bib12]). These utilize one SNP per 10 kb throughout the genome reference (with the exception of the Y chromosome and the mitochondrial genome). Significance LOD score thresholds were estimated using the "DSPRperm" command with 1000 replicates. The data input file is available as [File S6](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS6.txt). This scan was undertaken to asses whether one or few loci of large effect control pupal length, *i.e.*, the trait is not complex. Note that, if the trait is controlled by many loci of small or moderate effect size, the scan is too underpowered to realistically identify individual loci. However, the goal of the scan is to show that pupal size is not a relatively simple trait controlled by only a few major effect loci. Data availability {#s8} ----------------- All starting fly stocks are available from the sources detailed in [Table S1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/TableS1.pdf). RILs are available from the web address given above. Files of raw data are available as [File S4](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS4.zip), [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS5.zip), and [File S6](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS6.txt). Results {#s9} ======= The large numbers of individuals and vials measured, made possible through the use of the automated measuring system, provides robust insights into how phenotypic variation is partitioned for pupal length. While the method automatically identifies and measures pupae for multiple different parameters (see [File S1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS1.docx) for a full list), the length of the pupae was the principle focus of this study. Two distinct sets of biological material were used to define the biological properties of pupal length. The first dataset was all derived from an eight-way cross ([@bib11],[@bib12]); the second was from a four-way cross. None of the founding stocks are shared between the eight-way and the four-way datasets. *H*^2^ (reflecting all potentially genetic contributions to trait variance *e.g.*, additive, epistatic, dominance, maternal, and paternal effects) was estimated through the analysis of repeated measurements of RILs, whereas *h*^2^ heritability (reflecting only the impact of additive genetic effects) was measured by regression of midparent against their offspring measurements. Sampling properties of the eight-way and four-way datasets are detailed in [Table 1](#t1){ref-type="table"} and [Figure 1](#fig1){ref-type="fig"} (see also [Figure S1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FigureS1.pdf) for summary of repeat measurements of RILs). ###### Summary of sampling of eight-way and four-way datasets Dataset Heritability Vial Individuals -------------------------------------------------------------- ------------------------------------- ----------------------------------- -------------- Eight-way *H*^2^ Vials measured = 1184, RILs = 195 *n* = 83,402 Mean replicates per RIL= 6.7 (± 2.4 SD, min = 4, max = 15) Mean pupal length = 3.2 (± 0.21 SD) Mean vial pupal length = 3.2 (± 0.15 SD) Range of 95% of pupal lengths = 2.8--3.5 mm Mean number of measured pupae per vial = 70 (± 34 SD) *h*^2^ Crosses measured = 67 *n* = 3113 Mean vial pupal length = 3.5 (± 0.11 SD) Mean pupal length = 3.4 (± 018 SD) Range of 95% of pupal lengths = 3.1--3.5 mm Mean number of measured pupae per vial = 47 (± 15 SD) Four-way *H*^2^ Vials measured = 436, RILs = 81 *n* = 25,356 Mean replicates per RIL = 5.54 (± 0.78 SD, min = 3, max = 6) Mean pupal length = 3.6 (± 0.24 SD) Mean vial pupal length = 3.6 (± 0.15 SD) Range of 95% of pupal lengths = 3.4--4.0 mm Mean number of measured pupae per vial = 59 (± 24 SD) *h*^2^ Crosses measured = 363 *n* = 22,487 Mean vial pupal length = 3.4 (± 0.14 SD) Mean pupal length = 3.4 (± 0.23 SD) Range of 95% of pupal lengths = 3.1--3.7 mm Mean number of measured pupae per vial = 62 (± 17 SD) Only vials where ≥15 pupae were measured by the automated system are considered. For *H*^2^ estimates, only RILs where ≥3 replicate measurements were available are considered. {#fig1} Performance of automated pupal phenotyping {#s10} ------------------------------------------ The automated pupal recognition pipeline attempts to identify the external outlines of three classes of objects:*High confidence pupae*: conform to a set of expected properties and that have no neighboring pupae within 20 pixels (≈0.7 mm, Cellprofiler module: filter objects)*Medium confidence pupae*: conform to the same set of properties but with neighboring objects closer than 20 pixels. This often occurs as larvae select pupation sites touching each other, resulting in a more challenging target for image recognition*Low quality objects*: unlikely to be pupae.Examples of the three classes are shown in [Figure 2](#fig2){ref-type="fig"} (see [File S1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS1.docx) for further details).  for more examples of aberrant and canonical pupae). All objects are numbered in Cellprofiler output files to permit numerical measurements to be easily related to images.](1277f2){#fig2} Of the pupae retained for analysis, 47 and 53% were high and medium confidence, respectively. The probability of pupae being manually excluded as aberrantly measured was 0.02 for high quality pupae and 0.10 for medium confidence pupae ([Figure 2](#fig2){ref-type="fig"}). Note that not all images were examined manually, as, due to the robustness of the automated estimates, generally only vials with atypically high variance of the mean were examined. The precision of the automated system was assessed by remeasuring a subset of films after rotating them by 180°, and comparing 516 duplicated measurements of the same pupae ([Figure 3](#fig3){ref-type="fig"}). The difference between the two measurements was an average of 0.043 mm (±0.030 SD). Remeasuring pupae after delays of 15 or 30 hr also generated similar precision estimates, indicating that, once pupae become brown puparium (P2), there is no detectable change in pupal length. Furthermore, the exuvia of eclosed individuals can also be reliably measured (data not shown). {#fig3} The count of automatically measured pupae is a reasonable proxy for density within vials {#s11} ---------------------------------------------------------------------------------------- Unsurprisingly, not all pupae were correctly captured by the automated system, and, based on a sample of 148 vials, an average 20% ± 11.7 SD pupae per vial were missed, called as low confidence objects, or manually excluded ([Figure 4](#fig4){ref-type="fig"} and [Figure 2](#fig2){ref-type="fig"}). While there is an increased variance in the proportion of unmeasured pupae in higher density vials (*e.g.*, where the automated count is \>80), even here the automated count provides a reasonable proxy for density of individuals in each vial. This assumes that few, if any, larvae remained on the food surface after film transfer, as was generally the case (due in part to the short period parents remained in the vials). In addition, the small number of pupae removed from films that were obscured by larval food represents a constant proportion across all vials (see [File S1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS1.docx)). {#fig4} Impact of vial density on pupal length {#s12} -------------------------------------- One common major environmental covariate of many *Drosophila* traits is density of individuals within the vial in which they develop. Consequently, many *Drosophila* researchers control for this in experiments by collecting large numbers of zygotes and placing a controlled number in each vial. This is a fairly laborious process to perform routinely, and is complicated considerably where single pair crosses are required. Here, density was controlled only indirectly through limiting the number of parents used per vial, and restricting the number of nights they remained before being cleared \[generally two nights for single pair crosses, and one night for small groups (*n* = 10--20) of RIL individuals, see [File S1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FileS1.docx)\]. Throughout all experiments, two stocks were continually remeasured to act as controls (stock 335 and 329, see [Table S1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/TableS1.pdf)). The large number of repeat measurements of these two stocks across a range of densities permits the examination of any relationship between density and pupal length ([Figure 5](#fig5){ref-type="fig"}). Furthermore, in the same way, it is also possible to explore the relationship between density and pupal length using the repeated eight-way and four-way RIL measurements (see [Figure S2](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FigureS2.pdf)). All RILs and control stocks exhibit a uniformly negative relationship between density and pupal length; the slope varies from −0.0006 to −0.0041. This observed variability may reflect either variance in estimating slopes, or also that RILs exhibit different reaction norms. If the mode slope of −0.002 ([Figure S2](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FigureS2.pdf)) is used to correct the mean length of vials (or individuals) to that observed at the mean observed vial density, this can be achieved with the following equation:Figure 5Relationship between vial density and pupal length for the two control stocks. Repeated measurements of the same stocks (one short and one long) at different densities permits the slope and correlation of the relationship to be estimated. While there is a high degree of correlation for the longer stock-329 (red circles) (*R*^2^ = 0.48 and *P* = \>0.001), the correlation for the shorter stock-335 (black triangles) is minimal (*R*^2^ = 0.07 and *P* = 0.07).$$\begin{array}{l} {\mathit{M} = \text{mean}\,\text{vial}\,\text{density}\,\text{across}\,\text{whole}\,\text{experiment}} \\ {\mathit{S} = \text{slope}\,\text{of}\,\text{regression}\,\text{of}\,\text{density}\,\text{against}\,\text{mean}\,\text{vial}\,\text{length}\left( {-0.002} \right)} \\ {\mathit{D} = \text{automated}\,\text{estimate}\,\text{of}\,\text{density}\,\text{in}\,\text{vial}\,\text{to}\,\text{be}\,\text{corrected}} \\ {\mathit{Q} = \text{individual}\,\text{length}\,\text{measurement}\,\text{or}\,\text{vial}\,\text{mean}\,\text{to}\,\text{be}\,\text{corrected}} \\ {\left\lbrack {\left( {\mathit{D}-\mathit{M}} \right)\mathit{S}} \right\rbrack + \mathit{Q} = \text{Pupal}\,\text{length}\,\text{corrected}\,\text{for}\,\text{vial}\,\text{density}} \\ \end{array}$$Applying this formula to the 431 vials established as single pair crosses, \>99% of them would require a correction of \<±0.1 mm, and, of the 1620 vials established from small groups of RIL individuals, 95% would require a correction of \<±0.118 mm ([Figure 6](#fig6){ref-type="fig"}). Given the modest number of vials subjected to a large correction, and uncertainly about whether there is a truly universal linear relationship between density and pupal length, none of the measurements presented in the manuscript have been corrected analytically for density. Furthermore, with respect to repeated RIL measurements, reducing any confounding impact of density can be achieved by either experimentally increasing the number of replicate measurements closer to the mean density, or, analytically, by excluding or weighting down vials with extreme density values. {#fig6} Variance in the estimates of vial means {#s13} --------------------------------------- A script was written in Filemaker to explore the extent to which estimates of mean vial pupal length based on subsamples of individuals within a vial deviate from the vial means based on all individuals. This provides insight into at what point vials with low densities generate mean estimates with unacceptably high variance. [Figure 7](#fig7){ref-type="fig"} indicates that selecting single individuals to estimate vial means unsurprisingly results in a high degree of variance of up to 0.4 mm. While these estimates are unbiased, this is likely to prove unacceptably high given that the total observed range of pupal lengths is 1.1 and 0.8--0.9 mm within datasets. However, for random subsample sizes equal to a density of ≥15, the variance is greatly reduced to \<0.07 mm for 95% of vials ([Figure 7](#fig7){ref-type="fig"}). On this basis, only vials with \>15 measured pupae were included for analysis or presentation throughout this study. {#fig7} The impact of using smaller numbers of vials to estimate RIL mean length {#s14} ------------------------------------------------------------------------ In measuring panels of RILs, it is useful, for practical reasons, to minimize the number of replicate vials measured for each RIL to estimate an RIL mean with an acceptable degree of variance. Using eight RILs and the two control lines, all of which were measured 12 times or more, it is possible to explore the relationship between RIL means and the number of replicate vial measurements used. [Figure 8](#fig8){ref-type="fig"} indicates that six replicated measurements generally result in 95% of RIL estimates being \<0.1 mm different from that based on larger numbers of replicate vial measurements. The mean number of replicate vial measurements per RIL in this study was 6.4 ± 2.1 SD (5.8% = the minimum four replicates, 18.0% = five replicates, 61.2% = six replicates, and 15% more than six replicates). {#fig8} Estimates of h^2^ {#s15} ----------------- Estimates of the additive genetic impact on the variance of the trait can be gained from the slope of regressing the parental midpoint \[(length of father + length of mother)/2\] against the length of the progeny ([@bib6]). [Figure 9](#fig9){ref-type="fig"} indicates that, despite the eight-way (67 crosses) and four-way datasets (363 crosses) having no overlap in biological material, and having quite distinct sampling properties, they both result in remarkably similar estimates of *h*^2^, when representing all offspring of a cross as a mean. If all offspring are represented individually, then the estimate of heritability is largely unchanged for both datasets ([Table 2](#t2){ref-type="table"}). This robustness in the estimates is shared with human height ([Table 2](#t2){ref-type="table"} and [@bib8]). Likewise, regressions that use only the father or mother measurements, rather than their midpoint, confirm that paternal and maternal effects are of equal magnitude for both pupal length and human height (see [Figure S3](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FigureS3.pdf)). {ref-type="table"} for all parameter estimates). Green closed circles, eight-way crosses; blue open circles, four-way crosses.](1277f9){#fig9} ###### Summary of heritability estimates for pupal length and human height Four-Way Eight-Way Human Height -------------------------------------------------------------------------------------------------------- ---------------------- ---------------------- ----------------------------- *h*^2^ (mid-parent regression) Offspring vial means Offspring vial means Offspring family (≥4) means 0.44 ± 0.04 SE. 0.50 ± 0.09 SE. 0.68 ± 0.09 SE. *R*^2^ = 0.31 *R*^2^ = 0.33 *R*^2^ = 0.31 Offspring individual Offspring individual Offspring individual 0.42 ± 0.08 SE. 0.54 ± 0.02 SE. 0.68 ± 0.06 SE. *R*^2^ = 0.10 *R*^2^ = 0.14 *R*^2^ = 0.11 *H*^2^ Vial means 0. 58 Vial means 0. 61 Not possible Expected *H*^2^ assuming additive only model (2*h*^2^)/(1 + *h*^2^)[*^a^*](#t2n1){ref-type="table-fn"} Vial means 0.61 Vial means 0.71 --- Data for both human and *Drosophila* is uncorrected for any sexual dimorphism. [@bib15]. Estimates of H^2^ {#s16} ----------------- Estimates of all potentially genetic (*H*^2^) impacts on the variance of the trait can be obtained by estimating the proportion of the total variance in mean vial length measurements related to RIL stock. [Table 2](#t2){ref-type="table"} summarizes the results with both the eight-way (195 RILs) and four-way (81 RILs) datasets generating very similar estimates of *H*^2^ of 0.58 and 0.61, respectively. The significance of ANOVA tests reflects the strong genetic signal eight-way *P* = 2.7 × 10^−140^ (*F* = 10.86, Mean Square = 0.093, d.f. = 194, sum of squares = 18.3) four-way *P* = 8.8 × 10^−50^ (*F* = 9.026, Mean Square = 0.087, d.f. = 80, sum of squares = 6.9). Whole genome scan for loci impacting pupal length {#s17} ------------------------------------------------- Using 195 RILs from the eight-way dataset, and the DSPR tools, a genome scan for genomic regions associated with mean RIL pupal length was conducted (unweighted mean of replicate vial means). The density of markers used in the analysis is one SNP per 10 kb across almost the entire reference genome (excluding the mtDNA and Y chromosome). No SNPs of significance were identified at *α* = 0.05; this was also the case if density corrected RIL means were used ([Figure 10](#fig10){ref-type="fig"}). {#fig10} Discussion {#s18} ========== While ([@bib25]) stated that "One could argue that height in humans is the equivalent of bristle number in *Drosophila*, in terms of its role as a model phenotype." we propose that pupal length is as good, if not a better, analog, based on its biological properties described in detail for the first time here. This is in addition to the capacity to automate phenotyping using a reliable low cost system. Heritability is estimated using two distinct measures, broad sense (*H*^2^) and narrow sense (*h*^2^), for both of two biologically independent datasets. This allows the consistency of heritability estimates to be assessed between two datasets, which likely span most of the range of the trait within *D. melanogaster*. Furthermore, the contrast between *h*^2^ (additive genetic effects only) and *H*^2^ (additive + nonadditive effects) permits the magnitude of nonadditive effects to be estimated. The properties of pupal length established in this manner are potentially salient to efforts to develop resource efficient means to dissect genetically complex traits, and similar to those reported for human height ([Table 3](#t3){ref-type="table"}). The key parameter being the high and consistent heritability estimates, which, in part, reflect the reliability of automated measurements where the estimated precision of measurements of 0.04 mm is small relative to the range of the trait as a species (1.1 mm), or within datasets (0.8--0.9 mm). The reliability of pupal length measurements also likely contributes to the normality of the distributions observed at every hierarchical level in the datasets (*e.g.*, [Figure 1](#fig1){ref-type="fig"}). ###### Salient properties of pupal length relative to human height Adult Human Height *D. melanogaster* Pupal Length (Based on this Study) --------------------------------------------------------------- -------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------- Relative heritability of trait Among the most heritable reported in humans *h*^2^ = 0.4--0.7[*^a^*](#t3n1){ref-type="table-fn"} In top 80% of most heritable traits in *D. melanogaster h*^2^ = 0.4--0.5[*^b^*](#t3n2){ref-type="table-fn"} Measurement method Manual Automated Typical range within species (ignoring sexual dimorphism) ≈70 cm ≈1.1 mm Measurement error ≈1.5 cm[*^c^*](#t3n3){ref-type="table-fn"} 0.042 mm (± 0.030 SD) Ratio of measurement error: typical range of trait in species ≈0.02 ≈0.04 Paternal and maternal impact on offspring trait variance. Equal Equal Number of loci estimated to impact variance of trait ≈180--4000[*^d^*](#t3n4){ref-type="table-fn"} Unknown Experimental manipulation of allele frequencies Not possible Easy [@bib8] and [@bib25]; corrected for sexual dimorphism. Not adjusted for sexual dimorphism. [@bib26]. [@bib2], [@bib24], and [@bib29]. The ease of phenotyping large numbers of individuals at a developmental stage where they are obligatorily stationary for 2--3 d has enabled robust insight into the heritability of pupal length at 24°. The very similar estimates ([Table 2](#t2){ref-type="table"}) of heritability for two independent datasets provides increased confidence that this trait represents an excellent prospect with which to attempt to identify the genes underlying this complex trait. Furthermore, the similarity between the observed *H*^2^ estimates and those expected based on a purely additive model of inheritance (see last row in [Table 2](#t2){ref-type="table"}) indicates that epistatic, dominance, and paternal or maternal effects are relatively modest ([@bib15]). The observed similarity in the two *h*^2^ estimates (and the associated correlation *R*^2^) between the four-way and eight-way datasets ([Figure 9](#fig9){ref-type="fig"}) could also be argued to reflect a modest role for dominance. This is because, while the parents and offspring of the four-way crosses have similar levels of heterozygosity, the eight-way parents are inbred RILs while their offspring are heterozygous. Under some forms of dominance effects, both the slope and the correlation of the regression could be reduced in the latter case, which is not the case for pupal length. With respects to maternal and paternal effects, the data presented here ([Figure S3](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FigureS3.pdf)) provide evidence that any nonadditive paternal or maternal effects are small, as the size of the effect mothers and fathers exert on their offspring is symmetrical \[which is also a property of human height ([@bib8])\]. It has been detailed above that the impact of density within vials on pupal length is relatively modest (mostly \<0.1 mm, [Figure 5](#fig5){ref-type="fig"} and [Figure 6](#fig6){ref-type="fig"}), and can be conveniently controlled for. While other environmental variables were not systematically explored, limited data collected for the two control lines indicates that between 18° and 24° pupal length changes by 0.2--0.4 mm ([Figure S4](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FigureS4.pdf)), with pupal length being greater at cooler temperatures. The correlation between adult body weight and pupal length was also briefly examined and found to be *R*^2^ = 0.6 for both males and females (see [Figure S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1/FigureS5.pdf)). Any correlation between adult body length and pupal length was not examined, though it is potentially noteworthy that individuals possessing the Tb^1^ mutation resulting in a strikingly short tubby pupa are not readily distinguished from wildtype as adults based on their body length. Given the large numbers of genes impacting human height ([Table 1](#t1){ref-type="table"}), it is unsurprising that it can be correlated with disparate traits of particular interest (*e.g.*, cancer risk, cardiovascular disease and longevity \[[@bib18]; [NCD Risk Factor Collaboration](https://www.ncbi.nlm.nih.gov/pubmed/?term=NCD%20Risk%20Factor%20Collaboration%20(NCD-RisC)%5BCorporate%20Author%5D) [@bib17]\]. Future examination of traits correlated with pupal length (including their reaction norms) may prove informative with respect to their genetic architecture and that of pupal length. Currently, the automated system does not attempt to distinguish male pupae from female pupae, but it is likely that it could be extended to generate probabilistic sex assignments based on the fact that male pupae are on average ≈8% smaller than females (data not shown). When human height data are corrected for sexual dimorphism \[also an 8% average difference between sexes ([@bib6]; [@bib8])\], there is a corresponding increase of the heritability estimate by 9% (see Table 1 in [@bib8]). This implies that analyzing sexed pupal length data will generate heritability estimates that exceed the values reported here for unsexed pupae ([Table 2](#t2){ref-type="table"}). The observation that an average 20% ± 11.7 SD of pupae in a vial are not measured by the automated system ([Figure 4](#fig4){ref-type="fig"}) is generally not an issue for most experiments (even for select and resequence approaches), as it is often necessary that only a substantial proportion of individuals are measured, which is sufficient to provide a range of parents for the next generation, and to reduce the variance of mean vial estimates ([Figure 7](#fig7){ref-type="fig"}). Furthermore, parameters provided in [Figure 7](#fig7){ref-type="fig"} and [Figure 8](#fig8){ref-type="fig"} will provide future researchers with the capacity to readily design maximally resource-efficient experimental strategies. The observation that the genome scan using 195 RILs (eight-way dataset) failed to identify any significant loci impacting pupal length should be considered in light of the fact that, with this modest number of RILs, the power to detect SNPs of 10 or 5% effect size is only ≈0.37 and ≈0.08, respectively \[based on simulations incorporating these exact RILs, see [Figure 9](#fig9){ref-type="fig"} ([@bib12])\]. Consequently, the results presented here should be viewed as only potentially sufficient to indicate that few, if any, loci of large effect of size are likely to exist for this trait among the RILs used. It is of course conceivable that phenotyping more of the \>1600 RILs currently available would identify significant loci ([@bib11]; [@bib10]), and there are several in [Figure 10](#fig10){ref-type="fig"} that are close to the 0.05 significance threshold. Alternatively, an expanding variety of other approaches could be applied *e.g.*, select and resequence ([@bib23]; [@bib19]; [@bib13]), or comparisons between parallel selected lines ([@bib30]). While this needs to be further explored in the future, it is clear that the automated phenotyping system will be a key to designing mapping strategies that should allow the identification of many low effect size genes for this trait. The apparatus used to photograph pupae can be rapidly assembled for ∼500 € (including a camera), or may already be present in many laboratories as geldoc systems. Furthermore the necessary software is open source and free, and runs on any standard desktop PC or MAC. The capacity of the automated system to provide a large number of phenotyped individuals provides increased power to examine the genetic architecture of traits by most approaches. This is likely to prove key to identifying genes or alleles influencing mean pupal length and its variance ([@bib29]). This is in addition to potentially facilitating exploring the poorly understood genetic basis of sexual dimorphism ([@bib21]), and loci which impact the variance of traits \[rather than their central values ([@bib3])\]. In addition, the capacity to select from large numbers of individuals from which to establish future generations has the potential to enhance select and resequence approaches ([@bib23]; [@bib19]; [@bib13]). Furthermore, the large numbers of powerful approaches based on single pair matings developed by animal and plant breeders ([@bib7]) may also become amenable to examine complex traits in *Drosophila* through the use of this simple, low-cost, system. As with human height, over its typical range the genetic architecture of pupal length is of limited practical interest. However, for the last 130 yr, the former continues to provide key insights into the methods through which complex traits can be best understood, this is in part due to the ease and reliability of human height measurement. The high heritability of pupal length, and the capacity to easily automate phenotyping, combined with the small and well-described genome of *D. melanogaster* could make pupal length a similarly valuable model trait. Supplementary Material {#S19} ====================== Supplemental material is available online at [www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.039883/-/DC1). ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. We thank the following individuals: Stuart Macdonald for his help with the Drosophila Synthetic Population Resource (DSPR) stocks and tools and for maintaining this valuable common resource, Masayoshi Watada for his invaluable advice and access to his personal fly stocks, Predrag Kalajdzic for karyotyping the founder lines of the four-way cross, Oliver Hirsch for help with the development of the semi-transparent barcode stickers, Carolina Wählby for help with extending the use of the "untangle worms" module of Cellprofiler, and Anita Moeller for her excellent technical help and advice with conducting this experiment. Communicating editor: B. Oliver

Introduction {#sec1-1} ============ Stroke is the second leading cause of adults\' death and disability in most Western countries.\[[@ref1]\] In 2013, there were almost 25.7 million stroke survivors and 6.5 million deaths from stroke of whom 71% and 51% were with ischemic and hemorrhagic strokes, respectively. During the last two decades, in high-income countries, the incidence and mortality of both ischemic and hemorrhagic stroke decreased, in contrast, the incidence of hemorrhagic and ischemic stroke increased by 22% and 6% in low- and middle-income countries, respectively.\[[@ref2]\] According to the studies conducted throughout the world, the age-standardized ratio of stroke varies from 145.9 (130.3--162.8)/100,000 people in Japan,\[[@ref3]\]76 (59--94) in Australia,\[[@ref4]\] and 57.9 (54--61.8) in France.\[[@ref5]\] According to studies conducted in Iran, the overall crude incidence rate of first ever-stroke is 139 (128--149)/100,000 people; 144 in men (128--159) and 133 (111--148) in women. The incidence rate of stroke in Iran is significantly higher than that of Western countries, and regarding age, brain stroke occurs a decade earlier compared with Western countries.\[[@ref6]\] A study reported short-, intermediate-, and long-term mortality following stroke and factors affecting it in different parts of the world, including the WHO MONICA project,\[[@ref7]\] European countries,\[[@ref8][@ref9][@ref10]\] the USA,\[[@ref11][@ref12]\] and Australia.\[[@ref13][@ref14]\] Age, gender, type and severity of the stroke, history of previous stroke or transient ischemic attack (TIA), diabetes, and heart disease are among risk factors affecting mortality following a stroke that is reported in various studies.\[[@ref8][@ref15][@ref16][@ref17][@ref18]\] Studies have also been conducted in Iran on mortality following a stroke and its related factors.\[[@ref19][@ref20][@ref21][@ref22]\] In Mashhad study, factors associated with death within a year except for cases dying within the 1^st^ week included age, type of stroke, NIHSS score, and a history of TIA.\[[@ref23]\] In the study by Khosravi *et al*., age, a Glasgow Coma Scale Score of less than 13 and diabetes had a significant relationship with survival.\[[@ref20]\] However, there is little knowledge of the extent of fatality and the factors affecting long-term mortality following a stroke in Iran, especially in the East Azerbaijan province. Many previous studies have been focused on ischemic stroke, whereas this study covers both ischemic and hemorrhagic types of stroke. The results of this study will enable clinicians to improve the management and clinical care of stroke patients. The current research was designed and implemented with the aim of investigating the extent of 2-year mortality following a stroke and the associated affecting factors. Materials and Methods {#sec1-2} ===================== This study was conducted using a prospective cohort study designed in Tabriz, Northwest of Iran. All hospitalized cases in Razi and Emam Reza (referral centers of Tabriz) from October 2013 to March 2015 were included and followed until 2 years after the onset of the stroke. Patients with first-ever ischemic stroke (embolic and thrombotic) and hemorrhagic stroke (intracerebral hemorrhage and subarachnoid hemorrhage), according to the International Classification of Diseases-10 system with definitive diagnosis based on computed tomography (CT) and magnetic resonance imaging (MRI) scans, were included in the study. The inclusion criteria were first-ever stroke, residence in East Azerbaijan Province, consciousness, and voluntary consent to participate in the study, and the exclusion criteria were TIA and patients with a previous history of stroke. Census sampling method was used, and all cases that met the criteria for the study and those willing to participate were included in the study.\[[@ref24]\] Clinical examinations were conducted by a neurologist, and the related measures were obtained for all hospitalized patients. Next, the blood pressure and anthropometric measurements were taken and recorded in the patient\'s file. As suggested by a neurologist, CT scan and MRI were requested so that the disease could be definitively diagnosed. Following the definitive diagnosis of stroke and obtaining the informed consent of the patient, patients\' data were recorded in a checklist through an interview or using their file by an instructed expert. The data were composed of two parts: general characteristics, including age, gender, marital status, education, weight, height, monthly income, address, as well as the section regarding main stroke risk factors such as TIA history, history of heart attacks and stroke, family history of heart disease and stroke, family history of diabetes and high blood pressure, history of hypertension, diabetes, hyperlipidemia, heart disease, smoking, hookah or taking drugs and history of having contraceptive pills in women, the patient\'s blood pressure while being hospitalized and laboratory tests (fasting blood sugar, cholesterol, hematocrit, triglyceride, high-density lipoprotein, and low-density lipoprotein \[LDL\]). In cases where the patients were not able to talk, their close relatives were interviewed as a proxy. To measure the severity of the stroke, the modified National Institutes of Health Stroke Scale (mNIHSS) with high reliability and validity\[[@ref25][@ref26]\] was administered by a neurology resident at the hospital. At least two contact numbers were obtained from patients and their companions and were recorded in their files, and they were instructed on how to follow-up until 2 years after the occurrence of the stroke, and the patient\'s consent was obtained. As the cases measured in this study were patient\'s death or survival during the follow-up period, the Modified Rankin Scale\[[@ref27]\] was administered by an instructed expert until 2 years after the occurrence of the stroke. In case of the patient\'s death, the exact date and cause of death were asked and recorded. Cases who did not reply to several follow-ups and cases who had died due to reasons other than stroke were considered right censored in the survival analysis. Hypertension was defined as systolic blood pressure \>139 mmHg and/or diastolic blood pressure \>89 mmHg.\[[@ref28]\] People with hemoglobin A~1~C ≥6.5 and/or fasting plasma glucose level ≥126 mg/dl and/or 2-hour plasma glucose level ≥200 mg/dl during an oral glucose tolerance test/and/or random plasma glucose ≥200 mg/dl or those with medical treatments for diabetes were considered diabetes.\[[@ref28]\] Hyperlipidemia was defined as triglyceride ≥150 mg/dl or LDL ≥130 mg/dl or total cholesterol ≥240 mg/dl.\[[@ref29]\] Statistical analysis {#sec2-1} -------------------- Descriptive statistics were used to describe variables as frequency and percentage. The Chi-square test and if the observed frequency was limited, Fisher\'s exact test was used. To estimate the survival function, Kaplan--Meier analysis was used, and for between groups comparison, the Log-rank method was applied. To identify the factors predicting 2-year mortality, semiparametric Cox regression analysis was used. Variables with *P* \< 0.1 were included in the final, multivariable Cox analysis. The data were analyzed using the IBM SPSS software (version 24) and StataCorp STATA (version 14). *P* \< 0.05 was considered the statistical significance level. Results {#sec1-3} ======= A total of 1036 first-ever stroke patients were included in the study, among whom 394 (38%) died due to stroke within 2 years and 24 (2.3%) died of causes other than stroke. Most deaths occurred within the 1^st^ month after the occurrence of the stroke (56.8%). Follow-up was not completed for 64 patients (6.2%). The mortality rate of stroke in 6-month, 1-year, and 2-year follow-up periods was 31.6%, 34.5%, and 38%, respectively. Eight hundred and fifty-two patients (82.2%) had ischemic stroke, and 184 patients (17.8%) had hemorrhagic stroke. The mean age of the patients was 69.09 years with a standard deviation of 12.79 years, and 54.6% of participants were male. Regarding comorbid condition, 85.8% of the patients suffered from hypertension, 47.8% from diabetes, 27.8% from cardiovascular diseases, and 46.9% from hyperlipidemia. According to [Table 1](#T1){ref-type="table"}, as the age increases, mortality also increases in such a way that the age group \>75 years (53.4%) had the highest mortality compared to other age groups. Mortality was significantly higher in illiterate patients than in literate ones (*P* \< 0.001). Mortality was also higher in single patients (*P* = 0.01). No statistically significant difference was seen between the two groups regarding gender and place of living \[[Table 1](#T1){ref-type="table"}\]. ###### Sociodemographic characteristics of study participants Variable All case, *n* (%) Death within 2-year of stroke *P*\* ------------------------- ------------------- ------------------------------- ------------ --------- Age  \<55 136 (13.1) 117 (86) 19 (14) \<0.001  55-64 214 (20.7) 115 (72.4) 59 (27.6)  65-74 290 (28) 186 (64.1) 104 (35.9)  ≥75 395 (38.2) 186 (46.6) 211 (53.4) Sex  Male 566 (54.7) 363 (64.1) 203 (35.9) 0.12  Female 469 (45.3) 279 (59.5) 190 (40.5) Marital status  Yes 845 (81.60) 538 (63.7) 307 (36.3) 0.01  No 191 (18.4) 104 (54.5) 87 (45.5) Education  Illiterate 697 (67.3) 398 (57.1) 229 (42.9) \<0.001  Under diploma 241 (23.3) 171 (71) 70 (29)  Diploma and university 98 (9.5) 73 (74.5) 25 (25.5) Location  Rural 288 (28) 178 (6108) 110 (38.2) 0.85  Urban 740 (72) 462 (62.4) 278 (37.6) \*The level of statistical significance *P*\<0.05 Two-year mortality rate was 33.6% in ischemic stroke patients and 58.7% in hemorrhagic stroke patients (*P* \< 0.001). Two-year mortality rate following the stroke was 42% in patients with diabetes and 31.7% in patients without diabetes (*P* \< 0.001). Mortality of patients with a body mass index (BMI) \<25 (42.3%) was significantly higher than patients with BMI \> 25 (*P* \< 0.001). According to mNIHSS scores, the mortality rate in severe strokes with an index of mNIHSS ≥ 20 was the highest (80%), and it had a statistically significant relationship with a 2-year mortality rate (*P* \< 0.001). No statistically significant difference was observed between the two groups regarding factors such as smoking, history of heart disease, heart attack and TIA, hypertension, and use of contraceptives \[[Table 2](#T2){ref-type="table"}\]. ###### Major risk factor for study participants Variable All case, *n* (%) Death within 2 years of stroke *P*\* -------------------------- ------------------- -------------------------------- ------------ --------- BMI  \<25 482 (49.5) 278 (57.7) 204 (42.3) \<0.001  25-29/9 342 (35.1) 231 (67.5) 111 (32.5)  ≥30 150 (15.4) 110 (7303) 40 (26.7) Stroke type  Ischemic 852 (82.2) 566 (66.4) 286 (33.6) \<0.001  Hemorrhagic 184 (17.8) 76 (41.3) 108 (58.7) History of heart attack  No 982 (95.2) 612 (62.3) 370 (37.7) 0.74  Yes 50 (4.8) 30 (60) 20 (40) Previous TIA  No 980 (96.7) 615 (62.8) 365 (37.2) 0.91  Yes 33 (3.3) 21 (63.6) 12 (36.4) History of heart disease  sNo 744 (72.2) 456 (61.3) 288 (38.7) 0.34  Yes 287 (27.8) 185 (64.5) 102 (35.5) Family history of stroke  No 819 (79.4) 487 (59.5) 332 (40.5) \<0.001  Yes 212 (20.6) 155 (73.1) 57 (26.9) Current smoker  No 870 (85.1) 541 (62.2) 329 (37.8) 0.31  Yes 152 (14.9) 101 (66.4) 51 (33.6) Passive smoker  No 897 (88.3) 564 (62.8) 334 (37.2) 0.89  Yes 119 (11.7) 74 (62.5) 45 (28.1) Opium consumer  No 986 (97) 618 (62.7) 368 (73.3) 0.65  Yes 30 (3) 20 (66.7) 10 (33.3) Oral contraceptive  No 369 (85) 222 (60.2) 147 (39.8) 0.10  Yes 65 (15) 46 (70.80) 19 (29.2) Diabetes mellitus  No 511 (52.2) 349 (68.3) 162 (31.7) \<0.001  Yes 467 (47.8) 271 (58) 196 (42) Hypertension  No 146 (14.20) 96 (65.8) 50 (34.2) 0.34  Yes 879 (85.80) 542 (61.7) 337 (38.3) Hyperlipidemia  No 517 (53.1) 289 (55.9) 228 (44.1) \<0.001  Yes 456 (46.9) 323 (70.8) 133 (29.2) mNIHSS  Minor ≤5 337 (36.3) 295 (89.1) 36 (10.9) \<0.001  Moderate 6-19 472 (51.7) 281 (59.5) 191 (40.5)  Moderate to severe ≥20 110 (12) 22 (20) 88 (80) BMI=Body mass index, mNIHSS=modified National Institutes of Health Stroke Scale, TIA=Transient ischemic attack. \*The level of statistical significance *P*\<0.05 Survival rate of patients based on all variables was compared using the Log-rank test and showing a statistically significant difference between the mean 2-year survival of patients in terms of age, education, marital status, type of stroke, family history of stroke, diabetes, hyperlipidemia, BMI, and severity of stroke \[[Figure 1](#F1){ref-type="fig"}\]. {#F1} In a univariate analysis, factors such as age, marital status, education, BMI, type of stroke, family history of stroke, diabetes, hyperlipidemia, and the mNIHSS had a statistically significant relationship with a 2-year mortality rate. Two-year mortality risk in hemorrhagic stroke was 2.37 times higher than patients with ischemic stroke. People with diabetes and hypertension had a 1.41 and 1.15 higher risk of mortality than those with no diabetes and hypertension. However, this difference for diabetes was statistically significant at *P* = 0.001 and not significant for hypertension (*P* = 0.34). Female sex and a history of heart attack increased mortality risk, though not statistically significant. Regarding marital status, patients without spouse had 1.27 mortality risk higher than married patients. Illiterate patients showed 1.95 higher risk of mortality than patients holding a diploma or a higher degree \[[Table 3](#T3){ref-type="table"}\]. ###### Univariable and adjusted by sex, age group and multivariate survival analysis of suspected for death (Cox proportional hazards) Variable Univariable Adjusted for sex, age group Multivariable ---------------------------------------- --------------------- ----------------------------- --------------------- --------- -------------------- --------- Age (ref \<55)  55-64 2.15 (1.28-3.62) 0.004 2.06 (1.08-3.91) 0.02  65-74 2.92 (1.79-4.76) \<0.001 2.75 (1.47-5.15) 0.001  ≥75 4.80 (3.00-7.67) \<0.001 3.98 (2.17-7.29) \<0.001 Sex (female) 1.19 (0.97-1.45) 0.07 1.04 (0.79-1.37) 0.77 Marital status (single/divorced/widow) 1.27 (1.00-1.61) 0.04 1.01 (0.79-1.31) 0.88 0.94 (0.68-1.30) 0.74 Education (ref diploma and university)  xsUnder diploma 1.13 (0.72-1.79) 0.57 0.91 (0.57-1.45) 0.71 1.37 (0.73-2.57) 0.31  Illiterate 1.95 (1.29-2.93) 0.001 1.24 (0.8-1.90) 0.32 1.44 (0.79-2.61) 0.23 BMI (ref \<25)  25-29/9 0.69 (0.55-0.87) 0.002 0.76 (0.60-0.96) 0.02 0.96 (0.73-1.26) 0.79  ≥30 0.52 (0.37-0.73) \<0.001 0.55 (0.39-0.78) 0.001 0.85 (0.57-1.24) 0.40  Location (Urban) 0.97 (0.78-1.21) 0.80 0.96 (0.77-1.19) 0.72  Stroke type (hemorrhagic) 2.37 (1.90-2.96) \<0.001 2.64 (2.11-3.30) \<0.001 1.44 (1.06-1.97) 0.01  Hitory of heart attack 1.11 (0.70-1.74) 0.64 1.18 (0.75-1.85) 0.46  Pervious TIA 0.85 (0.48-1.51) 0.58 0.84 (0.47-1.50) 0.57  History of heart disease 0.87 (0.69-1.09) 0.24 0.84 (0.67-1.05) 0.14  Family history of stroke 0.59 (0.44-0.78) \<0.001 0.67 (0.50-0.88) 0.006 0.77 (0.55-1.07) 0.12  Current smoker 0.83 (0.62-1.11) 0.22 1.12 (0.82-1.52) 0.46  Passive smoker 1.00 (0.73-1.36) 0.98 1.05 (0.76-1.45) 0.73  Opium consumer 0.87 (0.46-1.64) 0.68 1.25 (0.66-2.37) 0.48  Oral contraceptive user 0.72 (0.44-1.16) 0.18 1.04 (0.63-1.70) 0.87  Diabetes mellitus 1.41 (1.14-1.73) 0.001 1.47 (1.19-1.82) \<0.001 1.40 (1.09-1.81) 0.009  Hypertension 1.15 (0.85-1.55) 0.34 0.93 (0.68-1.26) 0.65  Hyperlipidemia 0.57 (0.43-0.70) \<0.001 0.61 (0.48-0.76) \<0.001 0.94 (0.72-1.22) 0.67 mNIHSS (ref minor ≤5)  Moderate to severe 6-19 4.62 (3.23-6.60) \<0.001 3.90 (2.72-5.59) \<0.001 3.62 (2.46-5.33) \<0.001  Severe ≥20 17.63 (11.90-26.11) \<0.001 15.06 (10.08-22.50) \<0.001 12.92 (8.20-20.35) \<0.001 HR=Hazard ratio, CI=Confidence interval, BMI=Body mass index, mNIHSS=modified National Institutes of Health Stroke Scale, TIA=Transient ischemic attack. \*The level of statistical significance *P*\<0.05 After adjusting age and gender, variables such as BMI, type of stroke, family history of stroke, diabetes, hyperlipidemia, and the mNIHSS score showed a statistically significant relationship with a 2-year mortality rate. Smoking, taking medication, and contraceptives and passive smoking increased the risk of 2-year mortality 1.12, 1.25, 1.04, and 1.05 times, respectively, although none of these variables had a significant relationship with 2-year mortality. In this model, illiterate patients showed 1.24 times higher risk of mortality than patients holding a diploma or a higher degree, though not statistically significant. In the multivariate Cox model, variables such as age, type of stroke, diabetes, and severity of stroke, according to the mNIHSS, were identified as factors predicting 2-year mortality following the stroke. As the age increases, the risk of 2-year mortality also increases in a way that patients in the age group of 55--64 years showed 2.06 times, the age group of 65--74 years showed 2.06 times, and the age group ≥75 years showed 3.98 times risk of higher mortality than the age group of below 55 years. Having diabetes and hemorrhagic stroke increased the risk of 2-year mortality 1.40 and 1.44 times, respectively. Patients with a mNIHSS score of 6--19 and ≥20 had 3.62 and 12.92 times higher mortality, respectively, compared to people with mNIHSS ≤5 \[[Table 3](#T3){ref-type="table"}\]. Discussion {#sec1-4} ========== The current study aimed at investigating 2-year mortality following the first-ever stroke and identifying its predictors in the East Azerbaijan Province, Northwest of Iran. According to the results of this study, the rate of 2-year mortality was 38% (33.6% ischemic stroke and 58.7% hemorrhagic stroke), respectively. These results are almost in line with the study conducted in the Northeast of Iran.\[[@ref23]\] Various studies have been conducted in different parts of the world to investigate the long-term mortality of stroke. In a study by Chang *et al*. carried in Taiwan, the mortality rate of ischemic stroke was 12.2%, 15.8%, 20.5%, and 25.6%, respectively, for 1--4 years after the occurrence of stroke.\[[@ref30]\] In a study conducted in Denmark, mortality rates 1 year and 5 years following the occurrence of stroke were 41% and 60%, respectively.\[[@ref31]\] In a study conducted in Brazil, cumulative mortality rates 1 year, 2 years, and 3 years after the stroke were 27.8%, 31.2%, and 33.4%, respectively.\[[@ref32]\] Compared with the above-mentioned studies, the mortality rate in Iran was less than Taiwan and Brazil but more than Denmark. That might be because of the difference in patient\'s characteristics or health-care system. In age- and sex-adjusted Cox analysis, BMI, type of stroke, family history of stroke, hyperlipidemia, diabetes, and severity of the stroke were associated with 2-year mortality. Patients with hyperlipidemia and those with higher BMI had a lower risk of mortality suggesting a protective role against mortality. Markaki *et al*. showed that the risk of long-term mortality increased in patients with low cholesterol that is in line with the results of the present study.\[[@ref33]\] Recently, a prospective cohort study investigating the relationship among consuming fats, cardiovascular diseases, and stroke has shown that the use of various types of fat causes a reduction in mortality following a stroke.\[[@ref34]\] Other studies also showed reduced mortality among patients with higher BMI.\[[@ref35]\] The mechanism to explain how increased BMI might have a protective effect against stroke morality is not clear, further longitudinal studies are required to examine dietary habits after stroke, calorie intake, and the role of adipocytokines.\[[@ref36]\] In our study, the risk of mortality in patients with a family history of stroke was lower that was different from the results of previous studies in Iran.\[[@ref23]\] In Cox multivariate analysis, age, type of stroke, the severity of stroke, and diabetes were main factors affecting 2-year mortality. As the age increased, mortality risk also increased in a way that ≥75-year-old patients had 3.98 times higher risk of mortality. In other studies done in Iran, age has been recognized as a factor affecting mortality.\[[@ref20][@ref23]\] This is also confirmed in studies carried out in other parts of the world.\[[@ref30][@ref36][@ref37][@ref38]\] In the present study, the risk of 2-year mortality following hemorrhagic stroke was 1.44 times higher than ischemic stroke. The relationship between long-term mortality following a stroke has been confirmed by various studies conducted in Iran and other parts of the world. In other studies, hemorrhagic stroke with HR = 1.86\[[@ref23]\] and HR = 1.56\[[@ref39]\] was factors affecting mortality from stroke. In this study, diabetes increased the risk of 2-year mortality following stroke 1.40 times. This has been shown in many studies, for example, Goulart *et al*. showed having diabetes increased the risk of 2-year mortality 1.45 times.\[[@ref36]\] In a study by Wang *et al*., diabetes mellitus with HR = 1.20 had a significant relationship with stroke mortality.\[[@ref40]\] The severity of stroke at admission is a crucial factor that predicts mortality. In this study, patients with a mNIHSS score of 6--19 had 3.62 times and patients with mNIHSS ≥20 and 12.92 times higher risk of mortality than patients with mNIHSS ≤5. The severity of stroke in various studies has been identified as a factor predicting mortality. For example, a study in Iran, the NIHSS with HR = 1.13 had a significant association with 1-year mortality.\[[@ref23]\] In other studies that have been carried out in other parts of the world,\[[@ref16][@ref18]\] the association between the severity of stroke and mortality was shown. Conclusion {#sec1-5} ========== The present study is the first long-term study conducted in the East Azerbaijan Province, Northwest of Iran. It provides useful information about an epidemiologic picture of this disease and factors affecting mortality from stroke in the region. Being a hospital-based study is a limitation of this research that makes it difficult to generalize the findings to the population. Therefore, it is suggested that the results be interpreted with caution. Further population-based longitudinal studies are recommended. Financial support and sponsorship {#sec2-2} --------------------------------- This study was funded by Research Council, Tabriz University of Medical Science and reviewed and approved by the Ethics committee of Tabriz University of Medical Sciences (Ethical approval number: IR.TBZMED.REC.1395.127). Informed consent was obtained from all the participants. Conflicts of interest {#sec2-3} --------------------- There are no conflicts of interest. The researchers should thank the Research Council of Tabriz University of Medical Sciences and the honorable personnel of Imam Reza and Sina Hospitals of Tabriz for the sincere help they provided.

This work was supported by the Welcome Trust International Senior Research Fellowship in Biomedical Science in Central Europe (Reg. No. 070255/Z/03/Z) and by grants from Grant Agency of Charles University in Prague (No. 0073/2010 and No. 260501). Institutional support was provided by the Research Projects from Ministry of Education of the Czech Republic (No. MSM0021620806 and No. MSM6046137305). J.P.K. was supported by NIH Grant HL065217, by American Heart Association Grant-in-Aid 09GRNT2110159, and by a grant from the Jerome Lejeune Foundation. Cystathionine β-synthase (CBS,[1](#fn3){ref-type="fn"} EC 4.2.1.22) is a pyridoxal 5′-phosphate (PLP) dependent enzyme which catalyzes the first step of the transsulfuration pathway, namely, the condensation of serine with homocysteine to cystathionine ([@ref1]). Its deficiency due to missense mutations in the *CBS* gene is the most common cause of inherited homocystinuria, a treatable multisystemic disease affecting to various extent vasculature, connective tissues, and central nervous system (<http://www.ncbi.nlm.nih.gov/omim/236200>). More than 100 different pathogenic amino acid substitutions in the CBS protein were described, and the missense mutations represent 86% of all analyzed patient alleles (<http://www.uchs.edu/cbs/cbsdata/cbsmain.htm>).[^1] Human CBS is a homotetrameric protein, and each subunit (61 kDa) consists of 551 amino acids. The protein sequence comprises three regions: the N-terminal heme-binding domain (1−69), a highly conserved catalytic core (70−413), and the C-terminal regulatory domain (414−551) ([@ref2]), an autoinhibitory module with binding site for the allosteric activator, AdoMet ([@ref3]). CBS activity can be stimulated *in vitro* by several processes: by allosteric binding of *S*-adenosyl-[l]{.smallcaps}-methionine (AdoMet) ([@ref3]), by proteolytic cleavage yielding the C-terminally truncated dimer containing identical subunits with molecular mass of 45 kDa ([@ref4]), or by heat activation ([@ref3],[@ref5]). Proteolytic activation of CBS was observed also *in vivo* in rat liver extract ([@ref6]) and in HepG cell lines ([@ref7]). The spatial arrangement of CBS molecule was solved by X-ray crystalography for the truncated 45 kDa enzyme lacking the C-terminal regulatory domain (amino acids 1−413, 45CBS) only ([@ref8],[@ref9]); the 3-D structure belongs to the β-family of PLP enzymes such as *O*-acetylserine sulfhydrylase or tryptophan synthase. However, the 3-D structure of the full-length CBS (wtCBS) has not yet been determined, and therefore the atomic basis of the enzyme regulation is still unclear. While hydrophobicity of the C-terminal module and putative interdomain motions prevented successful crystallization of wtCBS, alternative techniques can yield at least partial information about the allosteric communication in the wtCBS protein. Using H/D exchange, Sen et al. showed that the region 356−385 exhibited significantly slower rate of deuterium incorporation for wtCBS compared to 45CBS ([@ref10]). The data were used for evaluation of a protein−protein docking exercise, and a structural model of the full-length CBS was proposed. However, this model has not yet been supported and/or refined by other structural techniques. In this study, we developed a procedure for covalent labeling of solvent-accessible amino acid residues ([@ref11]) in purified CBS. Using this technique, we compared reactivity of the side chains in 45CBS and wtCBS with six modifiers. These commonly used compounds specifically react with histidines (diethyl pyrocarbonate; DEP), tyrosines (*N*-acetylimidazole; NAI), cysteines (*N*-ethylmaleimide; NEM), lysines (sulfo-*N*-hydroxysuccinimido acetate; NHS), tryptophans (*N*-bromosuccinimide; NBS), and arginines (4-hydroxyphenylglyoxal; HPG) ([@ref12]). Surface mapping provided data which faciliated development of the refined model for wtCBS spatial arrangement and enabled insight into the structural basis of the enzyme allosteric regulation. Experimental Procedures {#sec2} ======================= Materials {#sec2.1} --------- If not specified otherwise, all chemicals were purchased from Sigma-Aldrich. Preparation of 45CBS and wtCBS {#sec2.2} ------------------------------ The 45CBS and the wtCBS were expressed in *Escherichia coli* and purified to homogenity as previously described ([@ref13],[@ref14]). Pulse Proteolysis {#sec2.3} ----------------- Pulse proteolysis was performed as described previously ([@ref15],[@ref16]) with some modifications. Purified 45CBS or wtCBS (0.5 mg/mL) was equilibrated overnight at 4 °C in 20 mM Tris-HCl (pH 8.0) containing 10 mM CaCl~2~ and urea (0−7 M) and then digested by thermolysin from *Bacillus thermoproteolyticus* (0.1 mg/mL). To carry out pulse proteolysis of wtCBS in the presence of AdoMet, wtCBS was incubated with 300 μM AdoMet at room temperature for 10 min prior to equilibration in urea. The proteolytic pulse (1 min) was quenched in 20 mM EDTA. Protein samples (7.5 μg) were analyzed by SDS−PAGE using Tris−acetate SDS running buffer with 3−8% gradient Tris−acetate precast gels (Invitrogen) and visualized by Coomassie blue solution. Experiments were repeated three times. Band intensities were quantified using GeneTools software (Syngene) and were fitted into the sigmoidal equation:using Origin 8.0 (Originlab); *f*~fold~ represents a fraction of folded proteins remaining intact after proteolytic pulse, *c*~m~ urea concentration at which *f*~fold~ is 0.5, and *c* urea concentration. Value of *p* is a slope of curve at *c*~m~, and it reflects unfolding cooperativity. Proteolytic Kinetics under Native Conditions {#sec2.4} -------------------------------------------- Purified proteins (0.5 mg/mL) were diluted in 20 mM Tris-HCl (pH 8.0) containing 10 mM CaCl~2~ and digested by thermolysin (0.1 mg/mL). At the chosen time point, proteolysis was quenched in 20 mM EDTA. SDS−PAGE and band quantification were performed as described for pulse proteolysis. First-order kinetic constant of proteolysis (*k*~p~) for each protein was determined by nonliner curve fitting ([@ref17]). Preparation of Modified Protein Samples {#sec2.5} --------------------------------------- CBS proteins (1 mg/mL) were diluted in modification buffer and covalently labeled. Each labeling procedure ([@ref18]−[@ref23]) (Table S1 in the [Supporting Information](#notes-1)) was repeated three times. Reaction was quenched by buffer exchange using Zeba Desalt spin columns (ThermoFischer Scientific) with elution by 50 mM NH~4~HCO~3~. Analysis of Modified Proteins {#sec2.6} ----------------------------- ### (i) Native Electrophoresis {#sec2.6.1} Labeled proteins (3 μg) were separated using Laemmli buffer system without sodium dodecyl sulfate and with 3−8% gradient Tris−acetate precast gel at 4 °C and visualized by silver staining kit (Promega) according to manufacturer's manual. ### (ii) CBS Activity Assay {#sec2.6.2} Enzyme activity of the proteins was determined in the absence or presence of 1 mM AdoMet by radiometric assay using \[^14^C\]-[l]{.smallcaps}-serine (Moravek Biochemicals); the previously described method ([@ref24]) was slightly modified. The reactants and products were separated by thin-layer chromatography using cellulose−HPTLC sheets (Merck) and subsequently visualized using PhosphorImager system (Molecular Dynamics); amount of radioactive cystathionine as the reaction product was determined by ImageQuant 5.0 software (Molecular Dynamics). ### (iii) In-Solution Proteins Digestion and Mass Spectrometric Analysis {#sec2.6.3} Labeled proteins were reduced in 5 mM dithiothreitol at 50 °C for 30 min; reduced cysteines were acetamidated in 25 mM iodoacetamide in the dark at room temperature for 30 min. Subsequently, they were digested by trypsin (Promega), chymotrypsin, endoprotease Glu-C, and protease double combinations ([@ref25]) at 37 °C for 1 h. The CBS:protease ratio (w/w) was 20:1. The protein digest was fractionated by ZipTip (Millipore), and each fraction was mixed with the matrix solution (saturated solution of α-cyano-4-hydroxycinnamic acid supplied by Bruker Daltonics, sample:matrix ratio of 1:1 v/v) and measured using Autoflex II (Bruker Daltonics) mass spectrometer equipped with a nitrogen laser (337 nm) in reflector positive mode (*m*/*z* range from 500 to 4500). The mass spectrometer was externally calibrated by peptide calibration standard II (Bruker Daltonics). All spectra were processed by Flex Analysis, Biotools 3.0 and mMass 3.0 ([@ref26]); mass accuracy tolerance was set at 50 ppm for MS and ±0.5 Da for MS/MS analyses ([@ref22]). With the exception of labeling with NBS, all other modification sites were identified by detection of labeled peptides that were not detected in unmodified controls ([@ref27]); expected mass shifts for each reaction are shown in [Supporting Information](#notes-1) Table S1. The labeling with NBS induces tryptophan oxidation ([@ref19]) which is also considered to be a common artifact of sample handling ([@ref28]). Since we observed tryptophan oxidation even in the unmodified controls, the residues labeled with NBS were determined by comparing peak intensities of the modified and the unmodified peptides ([@ref29]). Tryptophan residues were classified as labeled if the relative intensity of modified peptide increased at least 1.5-fold compared to the unmodified control. Identity of the modified peptides generated from all labeling experiments was confirmed by MS/MS measurements (method LIFT). In general, mass spectrometric measurements were satisfactorily reproducible; i.e., modification sites were determined identically in the repeated experiments. Thermal Activation of wtCBS {#sec2.7} --------------------------- The wtCBS diluted in the reaction buffer was incubated at 55 °C for 10 min and then chilled on ice ([@ref3],[@ref5]). Thermally activated proteins were labeled and analyzed by native electrophoresis, activity assay, and mass spectrometry as described above. Protein Structure Modeling {#sec2.8} -------------------------- Model of the C-terminal regulatory module was built by homology modeling package Modeller 9v3 ([@ref30]) using the structure of CBS-domain containing protein MJ0100 from *Methanocladococcus jannaschii* (PDB ID [3KPB](3KPB)) ([@ref31]) as a template. The initial sequence−sequence alignment was processed by the web services of PHYRE ([@ref32]) and PSI-BLAST ([@ref33]) and further modified manually. The resulting model was evaluated using Prosa web service ([@ref34]) and statistical coupling/protein sector analysis ([@ref35]). For this purpose, 6983 protein sequences from CBS subfamily were taken from the Pfam database ([@ref36]) and analyzed using a Python script based on the procedure introduced by Halabi and co-workers ([@ref35]). Model of wtCBS dimer was obtained by docking of a single C-terminal regulatory domain to 45CBS dimer (PDB ID [1JBQ](1JBQ), with missing loops reconstructed by Modeller package) using the program ZDOCK ([@ref37]). This was followed by addition of the C-terminal domain to the second subunit that was driven by symmetry. Differentially modified residues from the experimental results were forced to be involved in an interdomain interaction during the docking process. Structural models generated by this approach were visually inspected on the basis of several criteria, namely, involvement of differentially modified resides in the interaction, dimer symmetry, and protein stereochemistry. The full-length dimer was built by Modeller using the best-suited structure from docking procedures. Results {#sec3} ======= Pulse Proteolysis and Proteolytic Kinetics under Native Conditions {#sec3.1} ------------------------------------------------------------------ Using pulse proteolysis we determined the global conformational stability and unfolding cooperativity of CBS proteins (Figure [1](#fig1){ref-type="fig"} and Table [1](#tbl1){ref-type="table"}). The wtCBS exhibited lower resistance to urea-induced denaturation and lower degree of unfolding cooperativity compared to 45CBS. Binding of AdoMet to wtCBS moderately increased the protein stability toward urea, although it remained lower than the 45CBS resistance. On the other hand, unfolding cooperativity of wtCBS did not differ from wtCBS in the presence of AdoMet. These data show that CBS proteins adopt variant conformational states characterized by different degree of the stability. {#fig1} ###### Results from Pulse Proteolysis in Urea Gradient and Proteolytic Kinetics under Native Conditions[a](#tbl1-fn1){ref-type="table-fn"}   pulse proteolysis proteolytic kinetics under native conditions ---------------- ------------------- ---------------------------------------------- --------------- wtCBS 2.70 ± 0.08 1.51 ± 0.15 0.026 ± 0.005 wtCBS + AdoMet 3.26 ± 0.08 1.48 ± 0.13 0.056 ± 0.005 45CBS 4.08 ± 0.07 2.20 ± 0.28 0.075 ± 0.008 The CBS proteins (0.5 mg/mL) were digested with thermolysin (0.1 mg/mL). Data were evaluated by nonlinear curve fitting. Value of *c*~m~ reflects conformational stability, and value of *p* is informative about unfolding cooperativity; the constant *k*~p~ was acquired from the equation of first-rate kinetics. Proteolytic kinetics by thermolysin under native conditions (Figure [1](#fig1){ref-type="fig"}B) revealed slower cleavage of wtCBS compared to the 45CBS. AdoMet binding to wtCBS accelerated proteolysis; however, it was still slower than cleavage of 45CBS. These results are concordant with previously proposed regulation mechanisms; i.e., catalytic core is sterically hindered in the full-length protein by the C-terminal regulatory domain, and the hindrance is partially cleared off upon AdoMet binding ([@ref3],[@ref38]). To verify this hypothesis at the atomic level, we compared three-dimensional structures of 45CBS and wtCBS using protein surface mapping. Surface Mapping of CBS {#sec3.2} ---------------------- ### (i) Sequence Coverage {#sec3.2.1} To reach high degree of protein sequence coverage, 45CBS and wtCBS were digested by three proteolytic enzymes, namely, chymotrypsin, endoprotease Glu-C, and trypsin and by their double combinations; we obtained the sequence coverage of unmodified proteins 89% and 94% for 45CBS and wtCBS, respectively. For each labeling experiment, we selected the digests that yielded the highest amount of reliably identified modified amino acid residues (mass spectrometric data set available in the [Supporting Information](#notes-1)). ### (ii) Modifier Concentrations {#sec3.2.2} In the next step, we optimalized conditions of each labeling reaction as the excess of a modification agent may disrupt the spatial arrangement of a protein ([@ref39]). Therefore, the lowest concentration of the modifier that enabled efficient mass spectrometric detection of the modified residues was chosen. The integrity of modified proteins was monitored by disruption of their structure manifested by smears and lack of sharp bands on native gels along with complete loss of enzymatic activity. These effects were observed in the case of labeling with tyrosine modifiers tetranitromethane and iodine, and tryptophan modifier 2-hydroxy-5-nitrobenzyl bromide (data not shown). Six other modification agents ([Supporting Information](#notes-1) Table S1) were feasible for this study since modified CBS proteins migrated as sharp bands on native gels and retained high levels of enzymatic activity (Figure S1 and Table S2 in the [Supporting Information](#notes-1)). These data showed that most of the modification reactions did not even partially disturb integrity of CBS proteins, with the exception of the 45CBS labeled with NBS. In this case, modification procedure decreased enzyme activity to 43% of the unmodified control. This observation indicated that the labeling reaction may partially affect the catalytic activity. Despite this obstacle, we utilized NBS labeling since it was the only suitable compound for the detecting of solvent-exposed tryptophans. The eventual impact of the modification procedure on the protein integrity should be thus taken into account during structural interpretation. Modification Sites in CBS {#sec3.2.3} ------------------------- Modification reactions were examined by mass spectrometry, and residues labeled by six different agents were determined. The labeling was monitored qualitatively; i.e., the evaluation was based on the presence/absence of the modified peptides in 45CBS and wtCBS. This approach is commonly known as chemical footprinting ([@ref40]), a suitable technique for study of protein/protein and protein/DNA interactions ([@ref41]). Mass spectrometric analysis revealed 50 and 70 modification sites in 45CBS and wtCBS, respectively (Table [2](#tbl2){ref-type="table"}). Identity of the modified peptides was verified by MS/MS sequencing. However, several sites could not be confirmed due to insufficient fragmentation of the modified peptides (see Table [2](#tbl2){ref-type="table"}). The majority of the unconfirmed peptides contain a modified arginine residue since their tagging may affect the fragmentation process as previously reported ([@ref42]). Nevertheless, these peptides were included in the data set, since their observed masses were unambiguously assigned against *in silico* generated digests. ###### Modification Profile of CBS: List of Modified Amino Acids[a](#tbl2-fn1){ref-type="table-fn"} modifier ------------- -------------------------------------- ------ ------------------------------------------- ------------------ ------------------------------------------------ G1 W43 C15 G1 G1 R18 C15 W54 C52 K25 K25 R45 H17 W208 C272 K39 K39 R182/R190/R196[b](#t2fn1){ref-type="table-fn"} H22 [W408/W409/W410]{.ul} C370 K72 K72 R209 H65/H66/H67 M505[b](#t2fn1){ref-type="table-fn"} C431 K137 K137 [R336]{.ul}[b](#t2fn1){ref-type="table-fn"} H203 M529[b](#t2fn1){ref-type="table-fn"}   K172/K177[c](#t2fn2){ref-type="table-fn"} [K172/K177]{.ul} R369[b](#t2fn1){ref-type="table-fn"} K211     K211 K211 R389 K406     K271 K271 R413 H411     Y308 K322 R439 H433     K322 K405 R491[b](#t2fn1){ref-type="table-fn"} H501     K359 K406 R498 H507     [K384]{.ul} K441 R527       K398/K405[b](#t2fn1){ref-type="table-fn"} K472 R548       K406 K481         K441 K485         K472 K488         Y484/K485 K523         K488 K551         K523           K551     Differentially reactive residues (modified in 45CBS but not in wtCBS) are underlined. Identity of modified peptide could not have been confirmed by MS/MS due to insufficient fragmentation. Reactivity of these residues could have been confirmed by MS/MS only in the case of 45CBS. In wtCBS, 46 labeled residues were identified in the active core (region 1−413), and 24 sites were located in the regulatory domain (414−551). Comparing the side-chain reactivity of 45CBS and wtCBS in the region 1−413, we found four sites that were differentially labeled, i.e., modified in 45CBS and not wtCBS (Figure [2](#fig2){ref-type="fig"}; MS/MS spectra are shown in Figures S2−S4 in the [Supporting Information](#notes-1)). Differentially modified amino acid side chains were found in the peptide 164−181 (residue K172 and/or K177 modified by NHS), in the peptide 326−345 (residue R336 modified by HPG), in the peptide 380−389 (residue K384 modified by NAI), and in the peptide 406−413 (residue W408 and/or W409 and/or W410 modified by NBS). {#fig2} Differentially Reactive Peptides in Thermally Activated CBS {#sec3.2.4} ----------------------------------------------------------- Lack of reactivity of the above four peptides in wtCBS could be explained by interdomain sterical hindrance that is independent of the regulatory motions or by conformational changes which modulate the enzymatic activity. Thus, we tested whether the reactivity of the residues can be restored by stimulating activity of wtCBS. If so, such a result would suggest that the residues are involved in conformational motions; on the other hand, persistent unreactivity of these side chains would indicate their location at the fixed interdomain interface. Since the surface mapping in the presence of AdoMet could not be performed due to this ligand's reactivity toward most of the modifiers, thermally activated wtCBS was analyzed as a surrogate. This approach is feasible since allosteric changes due to AdoMet binding and partial heat denaturation share a common mechanism ([@ref3]). For covalent labeling of the stimulated wtCBS we applied only the modifiers and the digestions which provided differentially reactive peptides. Structural integrity of thermally activated wtCBS was preserved after the labeling reactions to an extent similar to the nondenatured wtCBS at the same concentration of modifier (enzyme activities are shown in Table S3, [Supporting Information](#notes-1)). The restoration of the residue reactivity was observed only for peptide 164−181 labeled by NHS, while the other three peptides were not labeled in thermally activated wtCBS (Figure [2](#fig2){ref-type="fig"}). These findings indicate that both sterical hindrance and regulatory motions are responsible for the differential reactivity of the residues. Structural Prediction Using Computational Modeling {#sec3.2.5} -------------------------------------------------- Initially, homology model of the C-terminal regulatory domain was built using archeal CBS domain as a template ([@ref31]). The amino acid sequence identity was 16% for the template-model pair. Nevertheless, CBS domains form a conserved tertiary structure despite rather low sequence identity of individual proteins ([@ref43]). Since this structural module contains many flexible loops, additional restraints were applied; namely, residues 466−472 and 537−549 of the CBS were forced to α-helix formation according to secondary structure prediction, and the distance between C~α~ atom of L412 and CD2 of L539 was restrained to 4 Å according to structural data of the template.The resulting structure was evaluated by Prosa and yielded a value of −6.38 which is comparable to values usual for experimental structures of the same size and similar to the score for the single subunit of the template (−7.50). Moreover, statistical coupling/protein sector analysis was used for evaluation of the model. Protein sectors are coevolving networks of residues supposed to play a common role (i.e., catalytic, stabilizing etc.) and thus showing a spatial proximity. Two sectors with evolutionary coupled residues were identified in the autoregulatory domain (Figure [3](#fig3){ref-type="fig"}A), showing a strong coevolution within each sector but a loose one between each other (Figure [3](#fig3){ref-type="fig"}B). The residues from the particular sector were found next to each other in the homology model which indicated a high plausibility of the resulting structure. In addition, the residue I483 was located at the sector interface and revealed strong coupling with both sectors (further details on SCA/sectors can be obtained in the [Supporting Information](#notes-1)). {#fig3} In the next step, the modeled C-terminal domain (residues 410−545) was docked onto the available structure of 45CBS. The initial docking was not successful, indicating that certain conformational changes in the catalytic domain may be associated with the binding of the regulatory domain. Therefore, the sterically hindering C-terminal helical region of the catalytic domain (residues 385−397) was deleted, and the truncated structure was used as bait with residues K172, K177, and R336 being forced to interaction. Other differentially reactive residues (K384 and W408/409/410) were not involved in docking as they are located in the linker between the catalytic and autoregulatory domains; their topology was used rather for verification of the resulted models. This modified docking procedure resulted in the generation of 79 structures; model no. 32 was selected by visual inspection on the basis of the location of the differentially modified residues, protein symmetry, and general stereochemistry. Using the result from the docking procedures, the structure of the full-length dimer was built. Plausibility of the structural model is greatly supported by data from surface mapping experiments: differentially reactive residues are located at the regulatory interface while residues modified in 45CBS as well as in wtCBS are still solvent-accessible (see Figure [3](#fig3){ref-type="fig"}D,E; structural model is available in the [Supporting Information](#notes-1)). However, the resulting model represents a possible structural interpretation of our experimental data and should not be interpreted as an atom-resolved structure due to limitations of homology modeling and protein−protein docking procedures. Discussion {#sec4} ========== Regulatory Interface in CBS {#sec4.1} --------------------------- A cross-talk between the active core and the regulatory domain in CBS modulates its enzyme activity. The main aim of the study was to compare residue reactivity in 45CBS and wtCBS as the differences may reveal the regulatory network. In 45CBS, we identified 50 labeled residues in total, and we found only 4 modification sites which were not detected in wtCBS (Table [2](#tbl2){ref-type="table"}). Using the thermally activated wtCBS as a surrogate of the AdoMet activated enzyme, we tested whether the abolished side-chain reactivity could be restored by the allosteric stimulation. The only differentially reactive peptide 164−181 was labeled in the thermally activated wtCBS, suggesting that this region (namely, residues K172 and/or K177) increases surface accessibility during enzyme stimulation and that it is involved in regulatory motions of CBS. Three other differentially reactive peptides were not labeled in wtCBS even upon thermal activation. Therefore, these side chains (R336, K384, W408 and/or W409 and/or W410) are probably localized at the fixed domain interface. Contact Area between the Catalytic Core and the Regulatory Domain {#sec4.2} ----------------------------------------------------------------- Three differentially modified residues (K172 and/or K177, R336, K384) were located in the same region of the 45CBS crystal structure (see Figure [2](#fig2){ref-type="fig"}E), and docking procedure showed that these residues may form an interface between the catalytic core and the regulatory domain. The region possessing differentially reactive sites was also predicted to form interdomain contact area due to the presence of hydrophobic residues on the surface of the 45CBS crystal structure ([@ref44]). Several CBS patient-derived mutations, namely, the p.V173M ([@ref45]), the p.E176K ([@ref46]), and the p.E302K ([@ref47]), which are located at this putative interface, exhibited enzyme activity similar to wtCBS and failed to be allostericaly stimulated by AdoMet. These observations indicate that mutations of these residues affect interdomain interactions and the CBS allostery. Another differentially reactive site, the tryptophan cluster W408/9/10, was not previously assigned by the diffraction analysis of the 45CBS crystal; thus we propose that it forms a flexible region in 45CBS and a loop between the active core and the C-terminal domain which is sterically hindered in the wtCBS. As mentioned in the [Results](#sec3), findings dealing with the residues W408/9/10 should be taken with care as modification of 45CBS with NBS led to apparent decrease in enzymatic activity. On the other hand, the electromigration of modified 45CBS was undistinguishable from the unmodified control, indicating that quaternary structure was preserved after the labeling. We assume that the protein structural integrity was not essentially disrupted and that the enzyme activity was affected due to the local conformational changes. Moreover, conclusions about different microenvironment along the tryptophan cluster are also supported by changes in tryptophan flourescence spectra reported previously ([@ref3],[@ref4]). However, a previously published study involving H/D exchange ([@ref10]) revealed the interdomain contact at a different region of the CBS structure. Although the changes in microenvironment of K384 were observed by both the H/D exchange and the covalent labeling in the present study, other differentially solvent-accessible regions were found using just single technique. We observed changes in residues K172/K177, R336, and W408/W409/W410, but they were not reported by Sen et al. On the contrary, H/D exchange study revealed differences in the segment of 359−370, but our covalent labeling experiments did not confirm them; in this region, three modification sites (K359, R369, and C370) were identically observed in the both proteins, 45CBS and wtCBS. Similarly to our study, results from H/D exchange were further supported by properties of certain mutant proteins, namely of double-linked mutant p.P78R/K102N ([@ref48]). Its amino acid substitutions are located in the proximity of the differentially solvent-accessible region 359−370, and this mutant affects the protein allostery driven by AdoMet binding. The discrepancies between results of covalent labeling and H/D exchange are unclear. The inconsistency may reflect the methodological limitations of each technique. Our experimental setup was designed for identification of differentially reactive sites rather than for quantification of small changes in extent of modification. Mass spectrometry analyses of the reactions were performed qualitatively (with exception of labeling with NBS; see [Experimental Procedures](#sec2)) which enabled determination of totally blocked residues only. On the other hand, we might have lost information about subtle conformational motions that would be revealed by quantitative evaluation. Conformational study using H/D exchange has its own limitations as well. It determines the rate of deuterium incorporation to protein backbone from several seconds to hours, and consequently any differences on a short time scale of the exchange may be missed. Therefore, each of these two approaches might locate only particular changes in the CBS protein. Unfortunately, an attempt to generate a model consistent with both data sets was not successful (data not shown). We can speculate that the discrepancies between these studies might arise from different conditions and procedures during preparation of CBS proteins. Consequently, each study would have analyzed only limited set of all possible states from the conformational ensemble. Nevertheless, the inconsistency needs to be examined by additional structural techniques. Allostery of CBS Is Not Associated with Extensive Conformational Changes {#sec4.3} ------------------------------------------------------------------------ Covalent labeling as well as H/D exchange showed that autoinhibition of the active core by the regulatory domain is associated with only subtle changes at the protein surface. These observations indicate that the CBS allostery is not necessarily directed by extensive conformational motions, suggesting that other factors may play an important role. Changes in structural flexibility and "population shift" as determinants for protein allostery were proposed in the past decade ([@ref49]−[@ref51]); it has been shown that the ligand binding often leads to stabilization and/or rigidification of certain conformations ([@ref52]). As the enzyme activity of CBS proteins ([@ref14]) is directly proportional to the conformational stability, as determined by pulse proteolysis (Table [1](#tbl1){ref-type="table"}; 45CBS, wtCBS in the presence of AdoMet, wtCBS in the absence of AdoMet, in descending order), it is tempting to speculate that CBS regulation may be driven by changes in protein dynamics. However, detailed knowledge of this type of CBS allostery is limited since the 3-D structure of the protein has not yet been reliably described in sufficient resolution. AdoMet Binding Site {#sec4.4} ------------------- Furthermore, the designed model of wtCBS provides information about the structural features of several sites with putative regulatory function. Since the spatial arrangement of archeal CBS-domain pair in complex with AdoMet was solved recently ([@ref31]) and we used this structure as a template for homology modeling of the C-terminal regulatory domain, the possible AdoMet binding site can be proposed. An interesting feature of the C-terminal autoinhibitory domain is its pseudo-2-fold symmetry (the axis indicated in Figure [3](#fig3){ref-type="fig"}A) which forms the basis for two ligand binding sites in each regulatory subunit (a and b in Figure [3](#fig3){ref-type="fig"}A). The experimental structures of the template−ligand complexes showed that the ligands bind to either one of these sites. Sequence similarity does not provide enough information to precisely identify the AdoMet binding site in CBS. However, AdoMet is likely bound in site b (Figure [3](#fig3){ref-type="fig"}A) including the residue D444 that has been identified to be involved in the autoinhibitory function ([@ref38]). CXXC Oxidoreductase Motif {#sec4.5} ------------------------- CBS also contains the CXXC oxidoreductase motif which spans residues 272−275. Here we identified the C272 as a solvent-exposed residue both in the 45CBS and in the wtCBS. This observation disagrees with a previous study that used three different cysteine modifying agents and an N-terminal sequencing of carboxymethylated peptides ([@ref5]). However, our findings are in agreement with the crystal structure of the 45CBS. The solvent accessibility of CXXC motif observed in our study may thus support the notion of its possible role in redox sensing ([@ref53]), although the biological relevance of this observation remains to be answered. Residues Responsible for Aggregation and Allostery {#sec4.6} -------------------------------------------------- Other residues, which play important role in CBS function, were revealed by labeling with NAI; this modification decreased the tendency of wtCBS to form higher order oligomers and increased its catalytic activity (Table S2 and Figure S1 in the [Supporting Information](#notes-1)). Similar effect was also observed after modification by NEM as reported previously ([@ref5]). Frank et al. explained the stabilizing action of the NEM by covalent blocking of C15, the residue responsible for aggregation of wtCBS. Interestingly, wtCBS labeled with NAI failed to be fully activated upon AdoMet binding while modification of wtCBS by NHS, which exhibited similar modification pattern as NAI (Table [2](#tbl2){ref-type="table"}), did not cause such effects. These data indicate that certain modified residues are responsible for CBS aggregation and also for allosteric activation, and their function can be repressed by covalent blocking of the reactive groups. Comparing the results from labeling with NAI and NHS, we can point out three candidate residues, namely, Y308, K359, and Y484, that are modified by NAI and not by NHS. However, we could not specify the "aggregation inducing" and "regulation networking" side chains in this study. Quarternary Structure of wtCBS {#sec4.7} ------------------------------ The relevance of the structural model proposed in this paper is limited as dimeric full-length CBS does not explain the atomic basis of the protein tetramerization. Our results from surface mapping revealed a single contact area between the catalytic core and the C-terminal regulatory domain; this is in agreement with the solved structure of the dimeric protein MJ0100 from *M. jannaschii* containing CBS pairs binding AdoMet ([@ref31]) and suggests that the autoinhibitory module contains dimer−dimer interface responsible for the CBS tetramer assembly (scheme in Figure [3](#fig3){ref-type="fig"}C). This proposal is also in agreement with the previously built structural model of wtCBS derived from H/D exchange ([@ref54]). In summary, we covalently labeled solvent-exposed side chains in CBS, and we identified the interface between the active core and the regulatory domain. The data were applied for generation of the refined full-length CBS structural model. Our results also indicate that the allostery of CBS is not associated with extensive conformational conversion but rather with changes in protein dynamics. We thank Petr Prikryl, Ph.D., for helpful assistance with the mass spectrometry measurements. \(i\) Detailed results from the labeling procedures and from the statistical coupling analysis of the autoinhibitory domain; (ii) an overview of the modified peptides that were identified by mass spectrometry; (iii) the generated structural model for the wtCBS illustrating the experimental data. This material is available free of charge via the Internet at <http://pubs.acs.org>. Supplementary Material ====================== ###### bi101384m_si_001.pdf ###### bi101384m_si_002.xls ###### bi101384m_si_003.pdb [^1]: Abbreviations: CBS, cystathionine β-synthase; PLP, pyridoxal 5′-phosphate; AdoMet, *S*-adenosyl-[l]{.smallcaps}-methionine; 45CBS, C-terminally truncated CBS; wtCBS, full-length CBS; DEP, diethyl pyrocarbonate; NAI, *N*-acetylimidazole; NEM, *N*-ethylmaleimide; NHS, sulfo-*N*-hydroxysuccinimido acetate; NBS, *N*-bromosuccinimide; HPG, 4-hydroxyphenylglyoxal.

**Related research article** Takahashi N, Ogita N, Takahashi T, Taniguchi S, Tanaka M, Seki M, Umeda M. 2019. A regulatory module controlling stress-induced cell cycle arrest in *Arabidopsis*. *eLife* **8**:e43944. doi: [10.7554/eLife.43944](http://doi.org/10.7554/eLife.43944) When something goes awry during the cell cycle -- for example, if DNA gets broken during replication -- checkpoint mechanisms put the cycle on pause so that the cell can repair the damage before dividing. In mammals, failure to activate these checkpoints can lead to cancer. The p53 tumor suppressor is a mammalian transcription factor which controls the genes that stop the cell cycle, repair DNA, and even trigger cell death in response to DNA damage ([@bib4]). Many cell cycle and DNA repair genes are conserved between vertebrates and plants, yet a p53 ortholog has never been found in any plant genome sequence. Instead, plants use SOG1 (short for suppressor of gamma-response 1), a plant-specific transcription factor that also arrests the cell cycle, coordinates DNA repair and promotes cell death. Recently, two independent studies have demonstrated that SOG1 regulates the expression of almost all the genes that are induced when DNA is damaged, including other transcription factors from the same family ([@bib1]; [@bib5]). Now, in eLife, Masaaki Umeda and colleagues from the Nara Institute of Science and Technology, the RIKEN Center for Sustainable Resource Science and the RIKEN Cluster for Pioneering Research -- with Naoki Takahashi as first author -- report on the roles of two of these SOG1-like transcription factors, ANAC044 and ANAC085 ([@bib7]). In plants, SOG1 can bind to the promoter regions of these factors, and it encourages the transcription of these genes upon DNA damage. Knockout experiments show that the ANAC044 and ANAC085 proteins are not necessary to repair DNA; instead, they stop the cell cycle just before division by increasing the levels of transcription factors called Rep-MYBs (where Rep is short for repressive). Once stabilized, these factors can bind to and inhibit genes involved in the progression of cell division ([@bib3]). When the cells are ready to divide, Rep-MYBs are marked for destruction, freeing up the genes that promote division so that they can be activated by other transcription factors ([@bib2]). Rep-MYBs do not accumulate when the genes for ANAC044 and ANAC085 are knocked out. The roots of mutant plants that lack both of these genes can therefore keep growing when agents that damage DNA are present. However, these double knockouts do not show a difference in the levels of RNA transcripts of Rep-MYBs. This prompted Takahashi et al. to speculate that an intermediate molecular step allows ANAC044 and ANAC085 to control the levels of Rep-MYBs after transcription, possibly by inhibiting the machinery that labels and degrades these proteins. Upon DNA damage, two kinases called ATM and ATR phosphorylate specific sites on SOG1 so that it can bind to DNA and perform its regulatory role ([@bib6]; [@bib8]; [@bib5]). Both ANAC044 and ANAC085 have sequences that are very similar to those of SOG1, but they appear to lack these phosphorylation sites. Moreover, overexpression of ANAC044 only inhibits the cell cycle if the DNA is damaged. It is therefore possible that this transcription factor only works in the presence of ANAC085, or that its activity is controlled by other kinases. Overall, the work by Takahashi et al. shows that plants have harnessed SOG1-like transcription factors to regulate the network of genes that respond to DNA damage. These results represent a major step in unraveling the hierarchical control of the DNA damage response in plants. So far, SOG1 appears to be the master regulator, delegating downstream responses among various regulators ([Figure 1](#fig1){ref-type="fig"}), with ANAC044 and ANAC085 stopping the cell cycle before division. Takahashi et al. also report that when plants are exposed to high temperatures, ANAC044 and ANAC085 help to halt the cell cycle. Therefore, these two transcription factors could be part of a central hub that delays cell division in response to a diverse set of stresses. ![Hierarchical control of the DNA damage response in plants.\ In plant cells, the kinases ATM and ATR are activated by different types of DNA damage. These enzymes go on to phosphorylate and activate the SOG1 transcription factor, which then binds to and switches on its target genes. These include (**i**) genes involved in DNA repair through homologous recombination (HR); (**ii**) the genes for ANAC044 and ANAC085, the newly identified transcription factors that help to stop the cell cycle; (**iii**) genes that trigger a cell death program (for when damage is too severe). ANAC044 and ANAC085 work by increasing the levels of Rep-MYB transcription factors. If stabilized, these proteins maintain the cells in the phase just before division (G2/M arrest) by binding to and repressing the genes essential for cell division to proceed. It is still unclear how Rep-MYBs are stabilized, or how SOG1 and ANAC044/ANAC085 may trigger cell death ([@bib7]).](elife-46781-fig1){#fig1} No competing interests declared. **Thomas Eekhout** is in the Department of Plant Biotechnology and Bioinformatics, Ghent University, and the VIB Center for Plant Systems Biology, Ghent, Belgium **Lieven De Veylder** is in the Department of Plant Biotechnology and Bioinformatics, Ghent University, and the VIB Center for Plant Systems Biology, Ghent, Belgium

Background {#Sec1} ========== In surgery, post-operative mortality is the most commonly cited outcome variable and to date has been considered the standard measure of quality of care. However, while cardiac operative mortality has fallen (currently 2.1% in the USA ([@CR26]) and 1.5% in the UK (Bridgewater et al. [@CR3])), post-operative morbidity remains common affecting between 4.3% (Fortescue et al. [@CR10]) and 36% (Magovern et al. [@CR21]) of cardiac patients and significantly prolonging length of stay (LOS) (Dupuis et al. [@CR8]). Such morbidity has substantial impact on healthcare resources, with the average in-hospital incremental cost of experiencing any complication at \$15,468 per patient (Brown et al. [@CR4]). Thus, strategies to identify and reduce post-operative morbidity might reduce both patient well-being and healthcare costs. However, few countries reliably record morbidity outcome data (Weiser et al. [@CR29]). Previously, morbidity definitions included death (Fortescue et al. [@CR10]), focused on major morbid events only (Huijskes et al. [@CR15]), or used surrogate markers of morbidity (for example post-operative LOS (Magovern et al. [@CR21]; Dupuis et al. [@CR8])). Likewise, many national cardiac surgical registers collect only 30-day mortality outcome and in some cases hospital LOS. Contrastingly, although the Society of Thoracic Surgeons national database contains 49 variables related to post-operative events, a composite metric aimed at defining post-operative morbidity includes in-hospital death and only five severe specific morbidities (Shahian et al. [@CR25]). The Cardiac Post-Operative Morbidity Score (C-POMS) (Sanders et al. [@CR23]) is a simple, validated score (0--13) by which to identify and quantify total morbidity burden (TMB) after adult cardiac surgery (Table [1](#Tab1){ref-type="table"}) on multiple post-operative days. We have previously reported that every unit increase in C-POMS is associated with a 1.7, 2.2 and 4.5-day increase in subsequent LOS on days 3 (D3), 5 (D5) and 8 (D8) (Sanders et al. [@CR23]), which has significant associated health-care costs, organisational and resource implications. Efforts to identify the risk factors associated with this post-operative TMB score may not only assist in identifying modifiable risk factors for which therapeutic interventions can be implemented to reduce the post-operative risk, but also serve to potentially further validate C-POMS as a useful clinical tool in cardiac surgery post-operative morbidity assessment.Table 1The Cardiac Post-Operative Morbidity Score (C-POMS) (as reported in Sanders et al. [@CR23])Morbidity typeC-POMS criteriaPulmonaryPresence of one or more of the following:\ ▪ New requirement for oxygen or respiratory support (including nebuliser therapy or request for chest physiotherapy on or after D5)\ ▪ Pleural effusion requiring drainageInfectiousPresence of one or more of the following:\ ▪ Currently on antibiotics\ ▪ Has had a temperature of \>38 °C in the last 24 h\ ▪ Has a white cell count/CRP level requiring in-hospital review or treatmentRenalPresence of one or more of the following:\ ▪ Presence of decreased urine output requiring intervention (including IV furosemide)\ ▪ Increased serum creatinine (\>30% from pre-operative level)\ ▪ Urinary catheter in situ\ ▪ New urinary incontinence\ ▪ Serum potassium abnormalities requiring treatmentGastrointestinalPresence of one or more of the following:\ ▪ Unable to tolerate an enteral diet for any reason including nausea, vomiting and abdominal distension\ ▪ Nasogastric tube\ ▪ Diagnosis of a gastrointestinal bleed\ ▪ DiarrhoeaCardiovascularPresence of one or more of the following:\ ▪ The use of inotropic therapy for any cardiovascular cause\ ▪ Pacing wires (on or after D5) and/or requiring temporary or new permanent pacing\ ▪ Diagnostic tests or therapy within the last 24 h for any of the following: (1) new MI or ischaemia, (2) hypotension (requiring fluid therapy, pharmacological therapy or omission of pharmacological therapy, (3) atrial or ventricular arrhythmias, (4) cardiogenic pulmonary oedema, thrombotic event (requiring anticoagulation), (5) hypertension (pharmacological therapy or omission of pharmacological therapy)NeurologicalNew neurological deficit (including confusion, delirium, coma, lack of coordination, drowsy/slow to wake, poor swallow, blurred vision, sedated, changing loss of consciousness)HaematologicalPresence of one or more of the following:\ ▪ Untherapeutic INR requiring pharmacological therapy or omission of pharmacological therapy\ ▪ Requirement for any of the following within the last 24 h: packed erythrocytes, platelets, fresh-frozen plasma, or cryoprecipitateWoundPresence of one or more of the following:\ ▪ Wound dehiscence requiring surgical exploration or drainage of pus from the operation wound with or without isolation of organisms\ ▪ Chest drains\ ▪ Wound pain significant enough to require continuing or escalating analgesic interventionPainPostoperative pain significant enough to require parenteral opioids and/or continuing or additional analgesiaEndocrineNew or additional requirements for blood sugar managementElectrolyteElectrolyte (including sodium, urea, phosphate) imbalance requiring oral or intravenous intervention (not including potassium as included in renal category)ReviewRemaining in hospital for further review, investigation and/or procedureAssisted ambulationA new or escalated post-operative requirement for mobility assistance (including wheelchair, crutches, zimmer frame, walking sticks or assistance)*CRP* C-reactive protein, *IV* intravenous, *MI* myocardial infarction, *INR* international normalised ratio, *OPA* out-patient appointment, *OT* occupational therapy Thus, we sought to identify risk factors associated with post-operative TMB, as assessed by the C-POMS tool, with an aim of identifying potentially modifiable risk factors that could be therapeutic targets to reduce post-operative cardiac surgical morbidity. Methods {#Sec2} ======= The National Research Ethics Committee London-Bentham (Chair Professor David Katz) gave ethics permission for this study (protocol amendment 7) on 6 September 2011 (reference 04/Q0502/73). All patients included in this study gave written informed consent to participate. Participants {#Sec3} ------------ Patients were drawn from the development and validation of the C-POMS study, detailed elsewhere (Sanders et al. [@CR23]). In brief, patients undergoing any form of adult cardiac surgery (excluding cardiac surgery for a congenital heart condition or a cardiomyopathy) between January 2005 and November 2007 at the Heart Hospital, University College London Hospitals NHS Trust, UK, and who gave written informed consent were eligible for inclusion. Excluded were those \<18 years old, undergoing emergency surgery, who were enrolled in clinical intervention trials or who died within 5 days of surgery. C-POMS {#Sec4} ------ The development and validation of the C-POMS tool and score is detailed elsewhere (Sanders et al. [@CR23]). In brief, The McMaster Framework (Kirshner and Guyatt, [@CR18]; Guyatt et al. [@CR12]) for constructing and assessing health indices for discriminative instruments, comprising item selection, item scaling, item reduction and determination of reliability and validity processes, was used. The C-POMS represents TMB as a summary score (0--13), derived by noting the presence or absence of 13 morbidity domains on days 3 (D3), 5 (D5), 8 (D8) and 15 (D15) after cardiac surgery (Table [1](#Tab1){ref-type="table"}). Pre- and intra-operative clinical data {#Sec5} -------------------------------------- A protocol development group, comprising 15 representatives from cardiac nursing, surgery, intensive care and anaesthesia, determined the pre-, intra-, and post-operative variables to be collected prior to commencement of the study. In brief, these included demographic details, past and current medical history, coronary heart disease risk factors, routine biochemistry and haematology measurements, anaesthesia details, operative details and a detailed record of the first 24 h in the intensive care unit. Variables were either obtained from the Society of Cardiothoracic Surgery national database or prospectively from the medical and nursing records by a dedicated, experienced research nurse using a standardised proforma. Identification and categorisation of potential risk factors {#Sec6} ----------------------------------------------------------- To identify pre-operative risk prediction models of post-operative morbidity following cardiac surgery, a systematic literature review was conducted using the basic framework for conducting systematic reviews from the Centre for Reviews and Dissemination (Dissemination Centre for Reviews and Dissemination [@CR7]). Three methodological quality filters were utilised. Firstly, the study population was defined as an adult population undergoing any form of cardiac surgery (excluding transplantation and grown-up congenital heart surgery); only methodologies that constructed a pre-operative risk assessment tool were included; valid outcomes were mortality and morbidity. There were no exclusions on the basis of the definition of either outcome. Search terms included cardiac surgery score, cardiac surgery risk score, pre-operative risk; cardiac surgery and risk prediction score; cardiac surgery, coronary artery bypass graft (CABG), surgery morbidity and surgery outcome. In addition to publication databases (the National Centre for Biotechnology Information, Entrez retrieval system and the Web of Science ISI Citation Databases), sources of on-going and recently completed studies (The National Research Register, The Cochrane Library of Systematic Reviews) were also interrogated to identify eligible papers. In total, the abstracts of 1067 papers were scrutinised. Backward and forward citation searches were conducted on all identified eligible papers. Overall, a total of 21 pre-operative risk prediction models were identified. All pre- and intra-operative variables obtained within the C-POMS study were classified with respect to these models into one of two tiers (results in Additional file [1](#MOESM1){ref-type="media"}): Tier 1 variables were those which had been associated with morbidity risk in three or more separate papers (significant evidence), while tier 2 variables were those identified in one or two papers (some evidence). Statistical methods {#Sec7} ------------------- Baseline characteristics are presented as mean ± SD or *n* (%). For the univariate analysis, C-POMS is presented as the median score and compared over categories using the Kruskal-Wallis test. The association of continuous variables with C-POMS was assessed using the Spearman rank correlation. For the multivariate analysis, variables with *p* \< 0.25 on univariate analysis were considered for inclusion into the models and stepwise regression with backwards elimination and a threshold of *p* \< 0.05 was run. Validation of the models was performed by running 1000 bootstrap samples. Variables selected in at least 60% of the bootstrap samples were included in the final model. For the data on all time points combined, the models used were random intercept models with time fitted as a fixed effect. Results {#Sec8} ======= Baseline participant characteristics {#Sec9} ------------------------------------ Of 748 potentially eligible patients undergoing cardiac surgery, 520 (69.5%) were screened (due to researcher availability) and 464 (89.2%) consented to participate. Fourteen participants subsequently became ineligible, leaving 450 who completed the study. Six participants declined for their data to be used outside the development of C-POMS. Table [2](#Tab2){ref-type="table"} summarises participant characteristics. The majority were White British (379, 85.4%), male (351, 79.1%) with a mean age of 66.6 years (range 19--91 years). Most had triple vessel disease (243, 54.7%), a good (\>50%) left ventricular ejection fraction (LVEF) (323, 74.1%) and were of moderate mortality risk (mean EuroSCORE 4.1). The majority had elective surgery (69.6%), using cardiopulmonary bypass (412, 93.4%) and stayed on intensive care unit (ICU) for an average of 2.0 days while remaining in the operating hospital for 9.5 days. The observed in-hospital mortality rate was 1.3%. Overall, 444 (100.0%) were in-patients on D1 and D3, 420 (94.6%) on D5, 178 (40.1%) on D8 and 45 (10.1%) on D15. Subsequent risk factor analysis was only appropriate on D3, D5 and D8 due to low numbers on D15.Table 2Baseline characteristics (*n* = 444). All values *n*(%) unless otherwise statedFrequency/mean ± SDDemographics Age (mean/years)66.6 ± 10.7 Female gender93 (20.9) Ethnicity (White British)379 (85.4)Medical history Renal (dialysis)7 (1.6) History of previous MI148 (33.3) Re-operation18 (4.1)Symptoms NYHA Class  I115 (26.0)  II205 (46.3)  III101 (22.8)  IV22 (5.0)Cardiac risk factors Smoking  Current49 (11.0)  Ex245 (55.2)  Never150 (33.8) Hypertension303 (68.2) Hypercholesteraemia343 (77.4) Diabetes103 (23.2) Body mass index (kg/m^2^)/mean28.5 ± 5.6Examination and investigation LVEF  Good323 (74.1)  Fair90 (20.6)  Poor23 (5.3)Pre-operative risk assessment EuroSCORE4.1 ± 2.8Intra-operative details Operative priority---elective309 (69.6) Operation performed  CABG299 (67.3)  AVR61 (13.7)  MVR10 (2.3)  CABG + AVR36 (8.1)  CABG + MVR0 (0.0)  AVR + MVR3 (0.7)  CABG + AVR + MVR2 (0.5)  Other33 (7.4) Cardiopulmonary bypass used412 (93.4)Outcome Length of ICU stay (mean/days)2.0 ± 3.5 Return to theatre21 (4.8) Total length of hospital stay (mean/days)11.8 ± 11.7 ### Tier 1 and 2 analysis {#Sec10} #### Univariate analysis {#FPar1} Fifty-six variables were identified in tiers 1 and 2. The incidence of seven (12.5%) pre-operative variables (cardiogenic shock, catheter-induced coronary closure, intra-aortic balloon pump, intubation, permanent pacemaker, immunosuppressant medications and inotropes) was too small for analysis, resulting in 49 variables (23 tier 1; 26 tier 2) for analysis. Thirty-three of the 49 (67.3%) variables previously identified to be associated with post-operative morbidity were found to be associated with C-POMS on at least one post-operative day (Table [3](#Tab3){ref-type="table"}). Of those, 10 variables were associated with C-POMS summary score on 1 day only, on either D3 (smoking, body mass index, urgency of operation, use of cardiopulmonary bypass, total drainage within the first 12 h and D1 inotropes) or D5 (ethnicity, neurological history, pre-operative heart rate and dialysis). No variables were solely predictive of D8 C-POMS summary score. Twenty variables (40.8%) were associated with C-POMS summary score on 2 days, while 3 variables (pre-operative albumin and New York Heart Association (NYHA) class and LVEF) were associated with C-POMS summary score on all days.Table 3Univariate analysis: tier 1 and 2 predictors of C-POMS summary score on D3, D5 and D8VariableD3 (*n* = 441)D5 (*n* = 419)D8 (*n* = 177)Median\ C-POMS/Rho*p*Median\ C-POMS/Rho*p*Median\ C-POMS/Rho*p*Tier 1 variablesDemographicsAge0.1880.0000.1100.0250.1330.078Age quartiles 1 (0--59)2.00.0022.00.0102.00.300 2 (60--68)3.02.03.0 3 (69--74)3.02.04.0 4 (≥75)4.02.53.0Age grp 2 \<652.00.0002.00.0442.50.083 65--743.02.03.0 ≥754.02.53.0Age grp 3 \<703.00.0012.00.4263.00.225 70--793.02.03.0 \>805.02.03.0Gender M3.00.0242.00.0163.00.215 F3.03.02.5Medical historyDiabetes Y4.00.0002.50.0153.00.308 N3.02.03.0Cerebrovascular disease Y4.00.0283.00.0234.00.219 N3.02.03.0Neurological history Y4.00.1033.00.0164.00.553 N3.02.03.0Congestive heart failure Y4.00.0094.00.0254.00.310 N3.02.03.0COPD/lung disease Y4.00.0053.00.0123.00.380 N3.02.03.0Renal disease Y5.00.0645.00.0015.00.025 N3.02.03.0POSSUM ECG Normal3.00.0002.00.0003.00.200 Sinus abnormal5.04.0-- AF4.03.03.0 Any other abnormal5.02.52.0Paced Y5.00.0054.00.0142.50.875 N3.02.03.0Dialysis Y5.50.0737.50.0065.50.064 N3.02.03.0Pre-operative measurementsBody mass index(kg/m^2^)0.1000.0430.0770.1300.0990.206 LVEF Good3.00.0132.00.0113.00.013 Fair3.02.04.0 Poor5.05.04.0Intra-operativeType of surgery CABG3.00.0002.00.0003.00.718 AVR4.03.03.0 MVR5.03.03.5 CABG + AVR4.02.03.0 AVR + MVR3.03.04.5 CABG + MVR + AVR5.05.06.0 Other5.03.03.0Urgency of op Elective3.00.0022.00.0613.00.327 Urgent4.02.03.0Within first 12 hSystolic blood pressure (highest)0.0020.9590.0260.599−0.0890.238Tier 2 variablesDemographicsEthnicity Caucasian3.00.0522.00.0083.00.532 Asian2.01.03.0 Black5.03.03.0 Other2.03.01.0Medical history Smoking Current2.00.0482.00.3953.00.753 Ex3.02.03.0 Never3.02.03.0Family Hx CAD Y3.00.0022.00.0193.00.686 N4.02.03.0Atrial arrhythmia Y4.00.0213.00.0073.00.781 N3.02.03.0Pre-operative measurements Albumin−0.2100.000−0.1440.004−0.1830.019 Haemoglobin−0.2790.000−0.1800.000−0.1430.058 Heart rate0.0900.0580.1790.0000.0800.294 Weight0.0440.3690.0420.3990.1010.186No diseased vessels 04.00.0113.00.0003.00.203 13.02.02.0 23.02.02.0 33.02.03.0NYHA class 12.00.0011.00.0023.00.001 23.02.03.0 34.03.03.0 45.02.55.0CCSC class 03.00.0413.00.0423.00.317 12.02.02.0 23.02.03.0 33.02.03.0 44.02.04.0Cardiomegaly Y5.00.0003.00.0004.00.141 N3.03.03.0 Not stated3.03.03.0Extracardiacarteriopathy Y5.00.0033.00.0044.50.147 N3.02.03.0Current medicationsDiuretic Y4.00.0033.00.0033.00.780 N3.02.03.0Intra-operativeCardiopulmonary bypass Y3.00.0012.00.0723.00.098 N2.01.01.0Within 1st 12 h after surgery Total drainage0.1390.0040.0300.5420.0310.679Post-operative (D1)Heart rhythm Sinus rhythm3.00.0002.00.0023.00.473 Sinus tachycardia2.52.02.5 Sinus bradycardia4.02.04.0 Atrial fibrillation5.03.04.0 Other3.02.03.0Inotropes Y4.00.0043.00.0703.00.923 N3.02.03.0 Overall, 8/23 (34.7%) tier 1 variables and 8/26 (34.7%) tier 2 variables were not associated with C-POMS summary score on any post-operative day (results in Additional file [2](#MOESM2){ref-type="media"}). #### Multivariate analysis {#FPar2} Of the 49 variables, 16 (32.7%) were independent risk factors of C-POMS summary score on one or more post-operative days, while no variables were associated with C-POMS outcome on all three post-operative days (Table [4](#Tab4){ref-type="table"}).Table 4Tier 1 and 2 (combined) independent predictors of C-POMS summary scoreVariableD3D5D8Combined over all time pointsEffectB(se)*p*EffectB(se)*p*EffectB(se)*p*B(se)*p*Pre-operative haemoglobin1 SD increase−0.42 (0.11)\<0.0001−0.29 (0.12)0.0112-h total drainage1 SD increase0.47 (0.10)\<0.0010.35 (0.10)0.0010.31 (0.09)0.001Pre-operative weight1 SD increase0.34 (0.10)0.0010.27 (0.11)0.020.26 (0.09)0.004Extra-cardiac arteriopathyYes: no0.73 (0.34)0.030.87 (0.35)0.010.66 (0.31)0.03AgePer year0.03 (0.009)0.0010.024 (0.009)DiabetesYes: no0.75 (0.23)0.0020.64 (0.27)0.020.92 (0.22)\<0.0001Left ventricular ejection fractionPoor: good/fair0.88 (0.43)0.041.38 (0.45)0.003CABG surgeryYes: no−1.17 (0.24)\<0.0001−1.02 (0.25)\<0.0001−1.06 (0.21)\<0.0001MVR surgeryYes: no1.40 (0.54)0.01Operative urgencyUrgent: elective0.48 (0.22)0.03Chronic obstructive pulmonary diseaseYes: no0.77 (0.30)0.01NYHA class1 category increase0.47 (0.24)0.05RenalYes: no2.46 (0.74)0.001Pre-operative cardiomegalyYes: no0.66 (0.32)0.040.66 (0.28)0.02Pre-operative albumin1 SD increase−0.43 (0.20)0.04−0.38 (0.09)\<0.0001Number of diseased vesselsIncrease of 11.04 (0.36)0.005Number of saphenous vein grafts1 SD increase−0.66 (0.31)0.04 There were eight tier 1 variables, all of which (with the exception of renal dysfunction) were associated with TMB on D3. Diabetes, LVEF, CABG surgery and renal dysfunction were also associated with D5 C-POMS summary score. However, only tier 2 variables (pre-operative albumin, number of diseased vessels and number of saphenous vein grafts) were independently predictive of D8 score and these variables were not independent risk factors for any other post-operative day. Pre-operative haemoglobin, 12 h drainage, pre-operative weight and extra cardiacarteriopathy were tier 2 risk factors associated with D3 and D5 morbidity score, while pre-operative cardiomegaly and NYHA class were associated with D5 score only. Considering all time points, 12-h total drainage, pre-operative weight, extra-cardiac arteriopathy, age, diabetes, CABG surgery, pre-operative cardiomegaly and pre-operative albumin level were independently predictive of C-POMS-defined post-operative morbidity outcome. Discussion {#Sec11} ========== We found that over two thirds (67.3%) of variables, previously reported as a risk factor for post-operative morbidity, were associated with the new C-POMS (denoting TMB) on at least one post-operative day, with 40% being significant risk factors for two post-operative days and three (6.1%) associated on all three post-operative days. From these results, there are four main findings of note. Firstly, we aimed to identify independent modifiable C-POMS risk factors amenable to therapeutic intervention. These were found to be pre-operative albumin and haemoglobin levels and weight---all tier 2 variables. Previously, pre-operative hypoalbuminaemia has been identified as a risk factor for post-operative delirium (Rudolph et al. [@CR22]), reoperation for bleeding (Engelman et al. [@CR9]), requirement for renal replacement therapy need (Engelman et al. [@CR9]; Sato et al. [@CR24]), increased ICU and hospital LOS (Engelman et al. [@CR9]; Koertzen et al. [@CR19]) and increased mortality (Engelman et al. [@CR9]; Koertzen et al. [@CR19]) cardiac surgery. While hypoalbuminaemia may be directly harmful, it may also mark other pathological states, such as anaemia, since pre-operative hypoalbuminaemia and anaemia are independently associated (Carrascal et al. [@CR5]). Similarly, in cardiac surgery, pre-operative anaemia is associated with in-hospital (De Santo et al. [@CR6]; Hung et al. [@CR16]) or 30-day (Boening et al. [@CR2]) mortality, post-operative blood transfusion rate (De Santo et al. [@CR6]; Hung et al. [@CR16]; Boening et al. [@CR2]), ICU (De Santo et al. [@CR6]; Hung et al. [@CR16]) and in-hospital LOS (De Santo et al. [@CR6]), major adverse cardiovascular events (Boening et al. [@CR2]), and renal complications (De Santo et al. [@CR6]; Boening et al. [@CR2]) than non-anaemic patients. However, evidence relating to whether pre-operative anaemia is associated with infection (Boening et al. [@CR2]) or not (De Santo et al. [@CR6]) is conflicting. There is less evidence relating pre-operative weight to post-operative outcome. Weight loss after bariatric surgery improves hypertension, diabetes and dyslipidaemia (Batsis et al. [@CR1]), although unintended pre-operative weight loss (≥10%) is also associated with prolonged hospital LOS (van Venrooij et al. [@CR28]). Evidence of association is, of course, not the same as proof of causation. However, overall, taken with associated literature, our findings suggest pre-operative albumin, haemoglobin and weight to be candidates for pre-operative interventional studies with the aim of improving post-operative morbidity. Secondly, we identified 17 *independent* risk factors for TMB on at least one post-operative day. Interestingly, except for renal dysfunction, the other seven tier 1 variables were all associated with D3 morbidity, with only three (diabetes, LVEF and CABG surgery) also being associated on D5 and none with D8 morbidity. Independent risk factors for D8 TMB lay entirely in tier 2. Such findings perhaps suggest that different risk factors are associated with outcome on different post-operative days and that well-accepted risk factors may only be useful for predicting morbidity risk in the first few days of recovery. Patterns of morbidity are well-recognised to differ with time after surgery, and it is likely that their drivers (whether they be intrinsic, environmental, or related to nascent morbidities) will also vary. Such factors are likely to explain the fact that correlations between D3 and D5 with D8 data differed. Furthermore, this study adds to the evidence that pre-operative haemoglobin concentration, pre-operative weight, extra-cardiac arteriopathy, pre-operative cardiomegaly, NYHA class, pre-operative albumin, number of diseased vessels, and number of saphenous vein grafts are independently associated with morbidity outcome. Thirdly, as expected, the majority of tier 1 variables (those with significant evidence of association with post-operative morbidity) were associated with TMB on at least one post-operative day. While this is not clinically surprising, it does further validate C-POMS as a useful clinical tool in outcome assessment following cardiac surgery. In relation to the eight tier 1 variables that were not associated with TMB on any post-operative day, combining cerebrovascular accident and transient ischaemic attack to a combined cerebrovascular disease variable was associated with a higher C-POMS on D3 and D5, in line with other morbidity risk assessment models (Magovern et al. [@CR21]; Huijskes et al. [@CR15]; Higgins et al. [@CR14]; Tuman et al. [@CR27])^.^ Furthermore, while hypertension has been independently associated with post-operative morbidity in some studies (for example, Fortescue et al. ([@CR10]) and Ivanov et al. ([@CR17])), this has been disputed by others (Higgins et al. [@CR14]; Hattler et al. [@CR13]). However, our results pertaining to systolic blood pressure, previous cardiac surgery and diagnosed peripheral vascular disease are at odds with the literature. Finally, this study also adds to the more limited data relating to some variables (those in tier 2) and their association with post-operative morbidity after cardiac surgery. Of these, 65.4% were associated with C-POMS TMB score on at least one post-operative day, while pre-operative albumin measurement and NYHA class were risk factors for all three post-operative days. Furthermore, this study also confirmed the findings of previous studies suggesting that unstable angina or recent myocardial infarction (Tuman et al. [@CR27]; Hattler et al. [@CR13]) and pre-existing liver disease (Higgins et al. [@CR14]) are not associated with post-operative morbidity. However, while our study corroborated that of Magovern et al. (Magovern et al. [@CR21]) showing that pre-operative atrial arrhythmia and cardiomegaly are associated with morbidity outcome (Magovern et al. [@CR21]), such results conflict with those found by Hattler and colleagues (Hattler et al. [@CR13]). There are four main limitations of this study. Firstly, there was no consistent definition of post-operative morbidity used in the pre-operative risk assessment models reported by others. Thus, a wide-range of variables to predict such diversely described outcomes were identified. However, over two-thirds of all variables were found to be associated with TMB, as defined by C-POMS, on at least one post-operative day suggesting that C-POMS is a valid measure of morbidity. Secondly, from the pre-operative risk assessment models, it is difficult to assess what variables were not found to be associated with post-operative morbidity. Most studies did not report variables for which no association with morbidity was identified, due to the often large number (\>100) of variables included. We have made our statistically non-significant results available to redress this balance and to aid in the evaluation of this (and other) tools. Thirdly, 14 risk factors identified in the pre-operative risk assessment models were available within this study dataset. Some variables, such as transplantation and ventricular-septal defect, were not available as these surgery types were not included in the study, while the others were non-routinely recorded items. Aside from catastrophic states, the variables not included were all tier 2 variables and in the main were only associated with outcome in one previous study. Finally, analysis to identify risk predictors for C-POMS TMB on D15 could not be conducted due to there being too few participants remaining in the hospital on D15 (*n* = 45). However, this could be the subject of future work. Conclusions {#Sec12} =========== Post-operative morbidity is increasingly accepted as an independent quality of care indicator, with approximately 80% of patients wanting to be informed of all the risks associated with surgery (Larobina et al. [@CR20]). Thus, in obtaining operative consent, patients should be told about 'less serious side effects and complications' (General Medical Council [@CR11]). C-POMS is a tool that permits such morbidity assessment and TMB scoring at several time points after cardiac surgery, allowing both broad and detailed tracking of morbidities. We have found that pre-operative albumin, haemoglobin and weight are potentially modifiable risk factors for which the investigation of the effect of therapeutic interventions on C-POMS outcome is warranted. Furthermore, we have identified, for the first time, that risk factors differ for different post-operative days. Further work should include the identification of novel risk factors of C-POMS TMB score for each post-operative day, and the identification of risk factors associated with each C-POMS morbidity type to identify the risk factors associated with D15 C-POMS summary score. Additionally, further validation of these results could be sought in a subsequent C-POMS dataset. Such identification of risk factors for C-POMS TMB may aid patient group and individual risk stratification and potentially reduce healthcare costs. Additional files {#Sec13} ================ Additional file 1:Variables included in pre-operative prediction models of post-operative morbidity. (DOCX 241 kb) Additional file 2:Univariate analysis: tier 1 and 2 not predictive of C-POMS summary score on D3, D5 and D8. (DOCX 20 kb) AVR : Aortic valve replacement CABG : Coronary artery bypass graft CAD : Coronary artery disease CCSC : Canadian cardiovascular score COPD : Chronic obstructive pulmonary disease C-POMS : Cardiac Post-Operative Morbidity Score CRP : C-reactive protein D3 (D5, D8, D15), : Day 3, (day 5, day 8, day 15) ICU : Intensive care unit INR : International normalised ratio IV : Intravenous LOS : Length of stay LVEF : Left ventricular ejection fraction MI : Myocardial infarction MVR : Mitral valve replacement NYHA : New York Heart Association OPA : Out-patient appointment OT : Occupational therapy TMB : Total morbidity burden UK : United Kingdom USA : United States of America The authors would like to thank all members of the protocol development group (PDG) and the patients who generously gave their time and consent to participate in the C-POMS study. Funding {#FPar3} ======= This work was unfunded, but Professors Hugh Montgomery and Michael Mythen were supported by the National Institute for Health Research University College London Hospitals Biomedical Research Centre. Availability of data and materials {#FPar4} ================================== Informed consent was not obtained for publication of patient data as publication of the dataset was not anticipated at the time of the initial C-POMS study. Thus, the data that support the findings of this study are available from JS but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of an appropriate research ethics committee and information governance (where appropriate) approvals. Authors' contributions {#FPar5} ====================== Each author has fulfilled the ICMJE guidelines to qualify as an author. According to the ICMJE guidelines, to qualify as an author, one should have made substantial contributions to conception and design (JS, MM, HM) or acquisition of data (JS), or analysis (JC) and interpretation of data (JS, MM, HM); been involved in drafting the manuscript or revising it critically for important intellectual content (ALL); and given final approval of the version to be published (ALL). Each author has participated sufficiently in the work to take public responsibility for the appropriate portions of the content and have agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Competing interests {#FPar6} =================== The authors declare that they have no competing interests. Authors' information {#FPar7} ==================== Not included. Consent for publication {#FPar8} ======================= Not applicable. Ethics approval and consent to participate {#FPar9} ========================================== The National Research Ethics Committee London-Bentham (Chair Professor David Katz) gave ethics permission for this work in the Cardiac Post-Operative Morbidity Score (C-POMS) study (protocol amendment 7) on 6 September 2011 (reference 04/Q0502/73). All patients included in this study gave written informed consent to participate.

Hyperhomocysteinemia (HHcy) has been described as a risk factor for diabetic retinopathy (DR),\[[@ref1]\] especially proliferative DR (PDR) in patients with both type 1 diabetes mellitus (DM)\[[@ref2]\] and type 2 DM.\[[@ref3]\] The study "Role of hyperhomocysteinemia in proliferative diabetic retinopathy: A case--control study" done by Gupta *et al*.\[[@ref4]\] is well-appreciated. They have shown higher prevalence of HHcy and higher mean serum levels of homocysteine (Hcy) in the cases with PDR when compared with the controls with no retinopathy and also have given possible reasons for not getting statistically significant differences. However, some points can be commented from the study by Gupta *et al*.\[[@ref4]\] The small sample size in this study (39 cases and 39 controls) which involves a largely prevalent disease could have affected the results. In statistical analysis, paired *t*-test could have yielded better results. Considering the number of factors included in the analysis, a multiple logistic regression analysis would have yielded better results on covariance and confounding factors. Some new perspectives can be explored from this article. Studies evaluating association of Hcy levels on DR in type 1 DM cases from India and comparison to those from outside India can further validate HHcy in DR cases, since geographical location has been mentioned to affect this association.\[[@ref2]\] There is evidence suggesting that Hcy activates vascular inflammation through mediators, including vascular endothelial growth factor,\[[@ref1]\] and the correlation of diabetic macular edema with HHcy has been mentioned.\[[@ref5]\] The effects of folate and vitamin B-12 supplementation on the level of Hcy remains to be studied. HHcy in PDR can modulate dual enzymatic activity of paraoxonase (PON), that is, esterase activity (PON-AREase) and lactonase activity (PON-HCTLase) which can be evident from elevated levels of vitreous homocysteine thiolactone (HCTL) and PON-HCTLase activity in PDR.\[[@ref6]\] This supports the association of HHcy with DR. Methylenetetrahydrofolate reductase (MTHFR) is an enzyme involved in remethylation of Hcy to methionine, and MTHFR gene polymorphism leads to impaired enzyme activity, resulting in HHcy and can contribute to the progression of DR.\[[@ref1]\] Genetic studies involving MTHFR gene polymorphism can further validate this association. Finally, association of HHcy with DR is an evolving topic. Future studies and discussion among peer group will enrich the collective academic knowledge and may help in future management of patients with DR.

{#sp1 .309}

Introduction {#s1} ============ A variety of agents, including environmental toxins or drugs, can cause DNA damage and lead to arrest of DNA replication forks. Arrested forks are among the most serious threats to genomic integrity because they can collapse, break, or rearrange [@pone.0013379-Branzei1], [@pone.0013379-Heller1], [@pone.0013379-Lambert1]. To circumvent these problems, cells are equipped with a DNA replication stress response pathway, termed the DNA replication checkpoint or the S-phase checkpoint. This checkpoint is activated by impeded replication forks and arrests the cell cycle while reducing the rate of DNA synthesis in order to coordinate with DNA repair and preserve genomic integrity [@pone.0013379-Branzei2], [@pone.0013379-Paulsen1], [@pone.0013379-Aguilera1]. In the fission yeast *Schizosaccharomyces pombe*, atop the replication checkpoint system stands a protein kinase, Rad3, which is homologous to human ATM and ATR [@pone.0013379-Carr1], [@pone.0013379-Nyberg1], [@pone.0013379-Boddy1]. Rad3 controls downstream effector kinases Cds1 (functional homolog of human Chk1) and Chk1 (functional homolog of human Chk2), both of which are also conserved throughout evolution [@pone.0013379-Carr1], [@pone.0013379-Nyberg1], [@pone.0013379-Boddy1]. Chk1 promotes the DNA damage checkpoint pathway while Cds1 acts as the master kinase for activation of the replication checkpoint to phosphorylate Cdc25, thereby inhibiting the Cdc2 (Cdk1) kinase and facilitating DNA repair and recombination pathways [@pone.0013379-Carr1], [@pone.0013379-Nyberg1], [@pone.0013379-Boddy1], [@pone.0013379-Walworth1], [@pone.0013379-Zeng1], [@pone.0013379-Boddy2], [@pone.0013379-Lindsay1]. Another important function of the replication checkpoint is to stabilize replication forks by maintaining proper assembly of replisome components and preserving DNA structures when problems are encountered during DNA replication [@pone.0013379-Lopes1], [@pone.0013379-Paciotti1], [@pone.0013379-Sogo1], [@pone.0013379-Tercero1], [@pone.0013379-Tercero2]. In fission yeast, we have demonstrated that Cds1 prevents fork collapse in response to hydroxyurea (HU) [@pone.0013379-Noguchi1], a compound that arrests replication forks, indicating that Cds1 is required for stabilization of stalled replication forks in a replication competent state. However, the precise molecular mechanisms by which stalled forks activate the replication checkpoint are not completely understood. In our previous studies concerning the mechanisms of the replication checkpoint, we found that Swi1 is required for proper activation of Cds1 in response to HU and for stabilization of replication forks in fission yeast [@pone.0013379-Noguchi1]. Further investigation has revealed that Swi1 interacts with Swi3 and travels with the replication fork as a replisome component [@pone.0013379-Noguchi2]. In the absence of Swi1 or Swi3, cells accumulate Rad22 DNA repair foci in S-phase [@pone.0013379-Noguchi1], [@pone.0013379-Noguchi2]. These foci correlate with the Rad22-dependent appearance of Holliday junction (HJ)-like structures [@pone.0013379-Noguchi2]. Rad22 is a Rad52 homolog and is known to bind single-stranded DNA (ssDNA) regions at the site of DNA damage [@pone.0013379-Lisby1], [@pone.0013379-Lisby2]. Thus, our results suggest a high rate of fork abnormalities in *swi1Δ* and *swi3Δ* mutant cells, generating ssDNA regions near the replication fork, which induces accumulation of HJ-like structures [@pone.0013379-Noguchi1], [@pone.0013379-Noguchi2]. Based on our results, we have referred to the Swi1--Swi3 complex as "the Replication [F]{.ul}ork [P]{.ul}rotection [C]{.ul}omplex" (FPC) [@pone.0013379-Noguchi2]. The Swi1--Swi3 complex is evolutionarily conserved and is homologous to the Tof1-Csm3 complex in *Saccharomyces cerevisiae* and the Timeless-Tipin complex in humans [@pone.0013379-Noguchi2], [@pone.0013379-Sommariva1], [@pone.0013379-Lee1], [@pone.0013379-Gotter1], [@pone.0013379-Mayer1]. Tof1-Csm3 has been shown to be part of the replisome or the replisome progression complex (RPC) and is involved in Rad53 activation [@pone.0013379-Calzada1], [@pone.0013379-Katou1], [@pone.0013379-Nedelcheva1], [@pone.0013379-Gambus1]. In humans, Timeless-Tipin interacts with Chk1 and ATR to control activation of checkpoint kinase Chk1 [@pone.0013379-Chou1], [@pone.0013379-Gotter2], [@pone.0013379-UnsalKacmaz1], [@pone.0013379-YoshizawaSugata1]. We have also demonstrated that Timeless-Tipin moves with replication forks, functions to stabilize replication forks, and facilitates sister chromatid cohesion in human cells [@pone.0013379-Leman1]. However, it remains unclear how Swi1--Swi3 related complexes interact with and stabilize replication forks and coordinate with multiple genome maintenance processes. Therefore, it is important to understand the functions of Swi1--Swi3, by dissecting molecular pathways that require this protein complex. In the present studies, we have carried out a mutational analysis of *S. pombe* Swi3 to further understand the functions of the Swi1--Swi3 replication fork protection complex. We identified separation-of-function mutations of Swi3, which leads us to propose that Swi3 utilizes different molecular mechanisms to regulate the replication checkpoint and sister chromatid cohesion. Swi3 appears to use the replication checkpoint pathway to stabilize stalled replication forks. However, when broken forks are present, Swi3 functions to restore forks using a checkpoint-independent pathway, which is also important for proper establishment of sister chromatid cohesion. Results {#s2} ======= Isolation of *swi3* mutants {#s2a} --------------------------- To understand the roles of the Swi1--Swi3 complex in the S-phase stress response, we isolated a number of *swi3* mutants using error prone PCR (*swi3* E-series). The wild-type *swi3* gene was replaced with mutagenized *swi3-5FLAG* genes at the *swi3* genomic locus, and mutants were tested for their viability in YES medium containing a high dose of hydroxyurea (HU, 10 mM) or camptothecin (CPT, 10 µM). HU depletes the dNTP pool and causes an arrest of replication fork progression, while CPT traps topoisomerase I on DNA and induces replication fork breakage. Among 20 HU and/or CPT-sensitive mutants, 12 mutants failed to express Swi3 as a 5FLAG fusion protein, suggesting that these mutants contain non-sense or frame-shift mutations that cause early termination of Swi3 translation (data not shown). Therefore, we decided to further characterize the remaining 8 mutants and *swi3-NBT7*, which was individually isolated as a mating-type switching defective mutant (see [Materials and Methods](#s4){ref-type="sec"}). These mutants were more carefully examined for sensitivities to HU and CPT. For sensitivity assays, we also included methyl methanesulfonate (MMS), which causes replication fork arrest by alkylating template DNA. The 9 mutants were categorized into four groups according to their drug sensitivity. Class I mutants (*swi3-E40* and *NBT7*) showed strong sensitivity to 2 mM HU, 0.0025% MMS and 2 µM CPT ([Figure 1A](#pone-0013379-g001){ref-type="fig"}), which was comparable to that of *swi3*Δ cells. Class II mutant (*swi3-E31*) was sensitive to 5 mM HU, 0.005% MMS and 5 µM CPT ([Figure 1A](#pone-0013379-g001){ref-type="fig"}). Class III mutants (*swi3-E1*, *E39, E59* and *E68*) were not significantly sensitive to HU and MMS, but did show significant sensitivity to 5 µM CPT ([Figure 1A](#pone-0013379-g001){ref-type="fig"}). Class IV mutants (*swi3-E10*, and *E42*) were only sensitive to HU, MMS or CPT at very high doses (10 mM HU, 0.01% MMS and 10 µM CPT, data not shown) where wild-type cells start to decrease their viability. Drug sensitivities of *swi3* mutants are summarized in [Table 1](#pone-0013379-t001){ref-type="table"}. {#pone-0013379-g001} 10.1371/journal.pone.0013379.t001 ###### Summary of *swi3* mutants characterized in this study. {#pone-0013379-t001-1} Growth rate -------------- ----- ------------------ ------------- ----- ----- ----- ----- ------ ----- wild-type none +++ +++ +++ +++ +++ ++++ − *swi3*Δ deletion +++ − − − N/A \+ +++ *swi3-E1* III R125H,A170V +++ +++ +++ \+ +++ ++++ N/D *swi3-E10* IV K78R,Y111N,R124L +++ +++ +++ +++ +++ ++++ N/D *swi3-E31* II D84H, F171L +++ \+ \+ \+ − +++ \+ *swi3-E39* III W128R +++ +++ +++ \+ +++ ++++ +/− *swi3-E40* I N17I, W95R +++ − − − − ++ +++ *swi3-E42* IV M91I +++ +++ +++ +++ +++ ++++ N/D *swi3-E59* III I94K,K68E,D177N +++ +++ +++ \+ +++ ++++ N/D *swi3-E68* III K47N,Y111C +++ +++ +++ \+ +++ ++++ N/D *swi3-NBT7* I L112R +++ − − − − \+ +++ *swi3-D84H* D84H +++ \+ \+ \+ − N/D +++ *swi3-F171L* F171L +++ +++ +++ +++ +++ N/D \+ *swi3-N17I* N17I +++ +++ +++ +++ +++ N/D \+ *swi3-W95R* W95R +++ − − − − N/D +++ *swi3-Y111C* Y111C +++ − − − − N/D +++ *swi3-K78R* K78R +++ +++ +++ +++ +++ N/D +/− *swi3-K47N* K47N +++ +++ +++ +++ +++ N/D \+ *swi3-L112R* L112R +++ − − − − N/D +++ *swi3-Y111N* Y111N +++ +++ +++ +++ +++ N/D +/− *swi3-R124L* R124L +++ +++ +++ +++ +++ N/D \+ Effects of *swi3* mutations on the formation of the Swi1--Swi3 complex {#s2b} ---------------------------------------------------------------------- Swi1 is known to co-purify with Swi3 from *S. pombe* cell extracts [@pone.0013379-Noguchi2], [@pone.0013379-Lee1]. Therefore, to address the effect of Swi3 mutations on Swi1--Swi3 complex formation, we performed immunoprecipitation assays to examine the ability of the Swi3 mutant proteins to interact with Swi1. Cells expressing Swi3-5FLAG mutant proteins were engineered to produce Swi1-13Myc from its genomic locus. As shown in [Figure 2A](#pone-0013379-g002){ref-type="fig"}, all mutant cells expressed Swi1-13Myc and Swi3-5FLAG proteins from their endogenous promoters. Swi1-13Myc consistently showed a series of degraded bands possibly due to proteolysis at specific sites in Swi1 ([Figure 2](#pone-0013379-g002){ref-type="fig"}). Interestingly, *swi3-E31*, *E40* and *NBT7* mutant cells reproducibly expressed reduced amounts of the Swi3 protein compared to *swi3* ^+^ cells, although they are readily detectable ([Figure 2A](#pone-0013379-g002){ref-type="fig"}). Accordingly, Swi3-5FLAG was immunoprecipitated, and Swi1 associated with Swi3 was examined by immunoblotting using the anti-FLAG and Myc antibodies. As shown in [Figure 2A](#pone-0013379-g002){ref-type="fig"}, considerable amounts of Swi3 mutant proteins were recovered from all mutants except for *swi3-E10*. Although the amount of Swi3 recovered from *swi3-E10* cells was much less than other mutants, it was still detectable. Notably, there was no detectable interaction of Swi1-13Myc and Swi3-5FLAG in *swi3-E31*, *E40* and *NBT7* (Classes I and II) cells, whereas other mutants retained significant levels of Swi1--Swi3 complex formation ([Figure 2A](#pone-0013379-g002){ref-type="fig"}). Considering that *swi3-E31*, *E40* and *NBT7* are significantly sensitive to HU, MMS and CPT ([Figure 1A](#pone-0013379-g001){ref-type="fig"} and [Table 1](#pone-0013379-t001){ref-type="table"}), these data suggest that Swi1--Swi3 complex formation is required for tolerance to replication fork arrest and damage. We also observed that *swi3-E1*, *E39*, *E59* and *E68* (Class III), which retained Swi1--Swi3 complex formation, were only sensitive to CPT ([Figures 1A](#pone-0013379-g001){ref-type="fig"} and [2A](#pone-0013379-g002){ref-type="fig"}); suggesting that CPT sensitivity is not caused uniquely by a defect of formation of the Swi1--Swi3 complex, and that Swi1--Swi3 possesses at least two separate functions in the preservation of genomic integrity. ![Effects of *swi3* mutations on the formation of the Swi1-Swi3 complex.\ (**A**) Protein extracts were prepared from cells expressing the indicated fusion proteins. Swi3-FLAG (Swi3-FL) was precipitated, and associated proteins were probed with the anti-Myc 9E10 and anti-FLAG M2 antibodies. Classes (C I to C IV) of Swi3 mutants are indicated in parentheses. The appearance of two to three bands in Swi1-Myc Western blots is due to degradation of the fusion protein [@pone.0013379-Noguchi2], [@pone.0013379-Lee1]. The Swi3-E40 mutant protein showed slower mobility, which is possibly due to mutational effects. Although only small amount of the Swi3-E10 protein was recovered by immunoprecipitation, Swi1-13Myc was efficiently co-precipitated with Swi3-E10. Western blotting of tubulin was performed as a loading control. (**B**) Protein extracts from the indicated strains were subjected to Swi3-TAP precipitation experiments, and associated proteins were probed with the anti-FLAG M2 and PAP antibodies. Original *swi3* alleles from which the single point mutations were derived are also indicated in parentheses. Although reduced amount of Swi3 were recovered by immunoprecipitation in *swi3-D84H*, *L112R* and *R124R*, they were all readily detected. Asterisk indicates non-specific bands. Representative results of repeat experiments are shown. IP, immunoprecipitation; WB, Western blotting; WCE, whole cell extract.](pone.0013379.g002){#pone-0013379-g002} DNA sequencing analysis of *swi3* mutants isolated by error prone PCR (*swi3-E* series) revealed that many of them contained multiple mutations in *swi3* ([Table 1](#pone-0013379-t001){ref-type="table"}). Therefore, we employed site-directed mutagenesis to introduce single-point mutations at sites found in *swi3-E10*, *swi3-E31*, *swi3-E40*, and *swi3-E68* ([Table 1](#pone-0013379-t001){ref-type="table"}). These mutants and *swi3-NBT7* (L112R) were expressed from the *swi3* promoter as TAP fusion proteins in *swi3*Δ *swi1-3FLAG* cells. As shown in [Figure 2B](#pone-0013379-g002){ref-type="fig"}, *swi3-D84H* (from *swi3-E31*), *swi3-W95R* (from *swi3-E40*), and *swi3-L112R* (from *swi3-NBT7*) mutant cells expressed somewhat lower amounts of Swi3-TAP protein. Moreover, Swi3-D84H, Swi3-W95R, and Swi3-L112R proteins failed to interact with Swi1 ([Figure 2B](#pone-0013379-g002){ref-type="fig"}). These data are consistent with the results of the original mutants (*swi3-E* series) that showed strong sensitivity to genotoxic agents ([Figure 1A](#pone-0013379-g001){ref-type="fig"} and [2A](#pone-0013379-g002){ref-type="fig"}, and [Table 1](#pone-0013379-t001){ref-type="table"}). Interestingly, when the two mutations (Y111C and K47N) present in Swi3-E68 were characterized individually, we found that the expression level of Swi3-Y111C was lower than wild-type and that Swi3-Y111C failed to interact with Swi1 ([Figure 2B](#pone-0013379-g002){ref-type="fig"}). In contrast, Swi3-K47N expression and its ability to interact with Swi1 were indistinguishable from the wild-type Swi3 protein ([Figure 2B](#pone-0013379-g002){ref-type="fig"}). We obtained similar results when the single-point Swi3 mutants were expressed as FLAG-fusion proteins (data not shown). Taken together with the fact that the original mutant (Swi3-E68) retained ability to interact with Swi1 ([Figure 2A](#pone-0013379-g002){ref-type="fig"}), these results suggest that the conformational change induced by Y111C abolishes the interaction with Swi1, which is compensated by the K47N mutation. More importantly, all of the single-point mutations that eliminate Swi1--Swi3 complex formation are located within the central "Swi3 domain" region (52--116 amino acids), which shows significant homology throughout evolution ([Figures 3A](#pone-0013379-g003){ref-type="fig"}) [@pone.0013379-Noguchi2]. Consistently, Swi3-D84H, W95R, Y111C and L112R mutants were all highly sensitive to HU, MMS and CPT ([Figure 1B](#pone-0013379-g001){ref-type="fig"}), suggesting that complex formation is important for cellular tolerance to S-phase stressing agents. {#pone-0013379-g003} Structural prediction of Swi3 {#s2c} ----------------------------- To understand the molecular basis of the Swi1-Swi3 replication fork protection complex, we performed structural analyses of the Swi3 protein at the amino acid sequence level. We used ClustalW multiple Sequence Alignment of Swi3-related proteins, including human Tipin, *Drosophila* Swi3 (dmSwi3), *C. elegans* Swi3 (ceSwi3), *S. pombe* Swi3 and *S. cerevisiae* Csm3. This analysis predicted that dmSwi3 and Csm3 have stretches of amino acid sequences that may divide Swi3-related proteins into at least 4 functional domains ([Figure 3B and 3C](#pone-0013379-g003){ref-type="fig"}). The N-terminal domain (Domain I: 1--34 amino acids) had weak similarity among the species and contained acidic amino acid-rich sequences. The central domain (Domain II: 35--117 amino acids) possessed significant homology throughout evolution. We have also found a putative nuclear localization signal (NLS: 42--49 amino acids) using the PredictNLS program provided by Columbia University. Although the NLS was only found in *S. pombe* Swi3, the corresponding regions from other species were rich in basic amino acids. Interestingly, using the Jpred3 secondary structure prediction program provided by University of Dundee, we found that Domain II contained three alpha helices, which were also conserved among the species. Although, the third domain (Domain III: 118--149 amino acids) was only weakly conserved, Jpred3 found that N-terminal part of this domain contained a conserved alpha helix structure. The fourth domain (Domain IV: 150--181 amino acids) appeared not to be conserved and varied in their length between species. Interestingly, the RPA-binding motif found in mammalian Tipin [@pone.0013379-Gotter1], [@pone.0013379-UnsalKacmaz1] was not conserved in *S. pombe*, *S. cerevisiae*, *C. elegans* and *Drosophila* ([Figure 3A and 3C](#pone-0013379-g003){ref-type="fig"}). It is important to note that all the mutations that disrupted Swi1--Swi3 complex formation (D84H, W95R, Y111C and L112R) were found in one of the alpha helices within the central conserved Swi3 domain, suggesting that alpha helix structures in Domain II play a role in interacting with Swi1 ([Figure 3B and 3C](#pone-0013379-g003){ref-type="fig"}). Cellular phenotypes of *swi3* mutants {#s2d} ------------------------------------- We have previously shown that *swi1*Δ and *swi3*Δ cells are moderately elongated with mild growth defect and that this mitotic delay requires Chk1 but not Cds1 [@pone.0013379-Noguchi1], [@pone.0013379-Noguchi2]. Therefore, we determined growth rates and cell lengths of *swi3* mutants. The growth rates of *swi3-E1*, *E10*, *E31*, *E39*, *E42*, *E59* and *E68* (Classes II, III and IV) cells were comparable to that of wild-type cells, whereas *swi3-E40* (Class I) showed mild growth defects similar to *swi3*Δ ([Figure 4A](#pone-0013379-g004){ref-type="fig"}). Interestingly, *swi3-NBT7* (Class I) had slower growth rate than *swi3*Δ ([Figure 4A](#pone-0013379-g004){ref-type="fig"}). Consistent with these results, *swi3-E40* and *NBT7* cells (Class I) showed moderate but statistically significant cell elongation phenotype in the absence of genotoxic agents, which was similar to that of *swi1*Δ and *swi3*Δ ([Figure 4B](#pone-0013379-g004){ref-type="fig"}). We then treated *swi3* mutants with CPT and measured their dividing cell length ([Figure 4C](#pone-0013379-g004){ref-type="fig"}). Wild-type cells showed mild elongation, probably due to a cell cycle delay provoked by replication fork breakage ([Figure 4B and 4C](#pone-0013379-g004){ref-type="fig"}. non-treated: 11.80 µm; CPT-treated: 14.12 µm; *p*-value = 0.0013). Consistent with the fact that CPT activates the Chk1-dependent checkpoint pathway [@pone.0013379-Wan1], *chk1*Δ cells failed to show a significant elongation phenotype (non-treated: 12.07 µm; CPT-treated: 12.57 µm). Rad3, which is known to activate Chk1, also appear to be important for this cell cycle delay (non-treated: 11.76 µm; CPT-treated: 11.50 µm). In contrast, *cds1*Δ cells showed mild elongation phenotype similar to wild-type (non-treated: 12.13 µm; CPT-treated: 14.84 µm; *p*-value = 0.0006), indicating that Cds1, a master kinase required for the replication checkpoint, does not have a major role in CPT-dependent cell cycle delay. When treated with CPT, Class I mutants (*E40* and *NBT7*) were significantly more elongated than wild-type cells. This elongation was similar to that of *swi3*Δ and *swi1*Δ cells ([Figure 4C](#pone-0013379-g004){ref-type="fig"}), suggesting that Class I mutant cells experience severe difficulty in recovering broken replication forks. Class IV mutants (*E10* and *E42*) were similar to wild-type. However, in response to CPT, *swi3-E39* (Class III) and *E68* (Class III) also displayed statistically stronger elongation phenotype, and *swi3-E31* (Class II) and *E59* (Class III) reproducibly showed somewhat more elongated phenotype when compared to wild-type. These results suggest that Class II and III mutants might have difficulty in recovering broken replication forks after CPT exposure, and they are consistent with the camptothecin sensitivity of the *swi3* mutants ([Figure 1C](#pone-0013379-g001){ref-type="fig"}). {#pone-0013379-g004} Effects of *swi3* mutations on the recovery of broken replication forks {#s2e} ----------------------------------------------------------------------- We have previously shown that Swi1 and Swi3 are required for stabilization of replication forks [@pone.0013379-Noguchi1], [@pone.0013379-Noguchi2], [@pone.0013379-Ansbach1]. To investigate the effect of Swi3 mutations on replication fork stability, we examined the recovery of DNA replication after fork breakage induced by CPT treatment. We chose representative *swi3* mutant(s) from each *swi3* mutation class, including *swi3-NBT7* and *swi3-E40* (Class I), *swi3-E31* (Class II), and *swi3-E39* (Class III). Class IV mutants were not included because they were not significantly sensitive to genotoxic agents ([Figure 1](#pone-0013379-g001){ref-type="fig"}). Chromosome samples of wild-type and *swi3* mutant cells were prepared before and at 3 h after CPT treatment, and at different time points during recovery after the removal of CPT. These chromosomes were then resolved by PFGE, which allows only a fully replicated chromosomes to appear in the gel ([Figure 5A](#pone-0013379-g005){ref-type="fig"}, the top and middle panels). Intact chromosomes from exponentially growing cells (log) in wild-type and all mutant strains migrated into the gel. CPT treatment causes replication fork breakage, leading to the reduction in the amount of intact chromosomes that migrated into the gel in wild-type and all *swi3* mutant cells. When cells were returned into fresh medium without CPT, intact chromosomes from wild-type cells re-appeared in the gel at 1.5 h after CPT removal due to the completion of DNA synthesis. However, intact chromosomes from all *swi3* mutant cells failed to migrate into the gel at 1.5 h and 3 h during recovery, indicating that Swi3 is required for the recovery of DNA replication after fork breakage. In addition, all *swi3* mutants contained excessive amounts of fragmented chromosomes during and after CPT exposure ([Figure 5A](#pone-0013379-g005){ref-type="fig"}, the top and middle panels), suggesting that Swi3 might be involved in efficient repair of broken replication forks. ![Effects of *swi3* mutations on the recovery of replication forks.\ (**A, B, C**) Chromosome samples from cells of the indicated genotypes were examined by PFGE. Cells were grown until mid-log phase and then incubated in the presence of 30 µM CPT (A), 20 mM HU (B) or 15 µM CPT (C) for 3 h at 30°C. Cells were then washed and released into fresh medium. Chromosomal DNA samples were prepared at the indicated times. *swi3* (except for *swi3-E39*) and *cds1* mutants appeared to harbor a shorter chromosome III, which is probably due to recombination at rDNA repeats [@pone.0013379-Sommariva1], [@pone.0013379-Ansbach1], [@pone.0013379-Noguchi4]. Representative results from repeat experiments are shown. (D) Five-fold serial dilutions of cells of the indicated genotypes were incubated on YES agar medium supplemented with the indicated amounts of CPT for 3 days at 32°C.](pone.0013379.g005){#pone-0013379-g005} Swi3 plays a role in recovery of broken replication forks in a manner independent of checkpoints {#s2f} ------------------------------------------------------------------------------------------------ It is known that Swi3 is important for efficient activation of the Cds1-dependent replication checkpoint [@pone.0013379-Noguchi2]. Therefore, we compared the replication recovery defect of *swi3* mutants with that of checkpoint mutants ([Figure 5A](#pone-0013379-g005){ref-type="fig"}, the bottom panel). *cds1*Δ cells failed to show significant defects in replication recovery after CPT exposure, indicating that Cds1 does not have a major role in the recovery of broken replication forks. It is known that Chk1 has a major function in the DNA damage checkpoint but also plays a redundant role with Cds1 in DNA replication checkpoint [@pone.0013379-Boddy2]. When *chk1*Δ cells were tested, a mild defect in replication recovery was observed in response to CPT. This is consistent with the fact that CPT activates the Chk1-dependent DNA damage checkpoint [@pone.0013379-Wan1]. However, *chk1*Δ cells were able to recover replication more efficiently than any of the *swi3* mutants tested. In addition, there was much less accumulation of CPT-dependent fragmented chromosomes in both *cds1*Δ and *chk1*Δ cells compared to the *swi3* mutants. We also examined chromosomal DNA isolated from *rad3*Δ cells ([Figure 5A](#pone-0013379-g005){ref-type="fig"}, the bottom panel). *rad3*Δ cells failed to recover replication and accumulated fragmented chromosomes as expected from the role of Rad3 in activation of both Cds1 and Chk1. These results suggest that the replication checkpoint function of Swi3 does not have a major role in the recovery of broken replication forks induced by CPT. To further address this possibility, we directly compared *swi3*Δ and *rad3*Δ cells in the recovery of broken forks, using a lower dose of CPT and longer recovery time points ([Figure 5C](#pone-0013379-g005){ref-type="fig"}). In this condition, *rad3*Δ cells were able to recover broken replication forks more efficiently than *swi3*Δ cells ([Figure 5C](#pone-0013379-g005){ref-type="fig"}). In addition, *swi3*Δ cells accumulate significantly more fragmented DNA during recovery when compared to *rad3*Δ cells. Furthermore, *swi3*Δ *rad26*Δ cells were much more sensitive to CPT than either single mutant ([Figure 5D](#pone-0013379-g005){ref-type="fig"}). We also obtained similar results with *swi1*Δ *rad26*Δ cells in a CPT sensitivity assay ([Figure 5D](#pone-0013379-g005){ref-type="fig"}). Rad26 is essential for activation of Rad3, which is required for both Cds1 and Chk1 activities [@pone.0013379-Edwards1], [@pone.0013379-Wolkow1]. Therefore, our results suggest that Swi3 has a specific role in replication recovery after fork breakage, which is independent of Cds1 or Chk1 activation. Effects of *swi3* mutations on the replication checkpoint {#s2g} --------------------------------------------------------- The Cds1-dependent replication checkpoint is required for the resumption of stalled replication forks in response to HU [@pone.0013379-Boddy1], [@pone.0013379-Lindsay1], [@pone.0013379-Noguchi1]. Since Swi3 is important for the full activation of Cds1 and for the stabilization of stalled replication forks in response to HU that activates Cds1 [@pone.0013379-Noguchi2], we also monitored replication recovery after fork arrest due to HU exposure. *swi3-NBT7* (Class I), *swi3-E31* (Class II), *swi3-E39* (Class III), and *swi3-E40* (Class I) cells were treated with HU for 3 h and released into fresh medium to allow resumption of replication. As expected, *swi3*Δ, *cds1*Δ and *rad3*Δ cells, which all have defects in Cds1 activation, were not able to properly resume stalled forks after HU exposure ([Figure 5B](#pone-0013379-g005){ref-type="fig"}). *swi3-NBT7* and *E40* cells also showed resumption defects similar to *swi3*Δ ([Figure 5B](#pone-0013379-g005){ref-type="fig"}), suggesting the failure in Cds1 activation in these mutants. *swi3-E31* had mild defect in recovery from HU, which is consistent with its mild sensitivity to HU ([Figure 1](#pone-0013379-g001){ref-type="fig"}). Interestingly, *swi3-E39* cells were able to resume replication ([Figure 5B](#pone-0013379-g005){ref-type="fig"}) at the wild type level, suggesting that the Cds1-dependent replication checkpoint is still functional in this mutant. Taken together, our present data indicate that Swi3 has a replication function that is independent of Cds1 activation. Therefore, we have examined the effects of *swi3* mutations on Cds1 activity. As shown in [Figure 6A](#pone-0013379-g006){ref-type="fig"} and [Table 1](#pone-0013379-t001){ref-type="table"}, although there was a variation, Class I *swi3* mutants (*swi3-NBT7* and *E40*) had the most significant defects in Cds1 activation, which is consistent with the results of PFGE after HU treatment ([Figure 5B](#pone-0013379-g005){ref-type="fig"}). Class II mutants (*swi3-E31*) also displayed a slight decrease in Cds1 activation. However, Class III (*swi3-E1, E39, E59* and *E68*) and Class IV (*swi3-E10* and *E42)* appeared to have wild-type levels of Cds1 activity ([Figure 6A](#pone-0013379-g006){ref-type="fig"} and [Table 1](#pone-0013379-t001){ref-type="table"}). These results indicate that Class I and II mutants but not Class III and IV mutants have a defect in the Cds1-dependent replication checkpoint. Taken together with the fact that *swi3-E39* (Class III) mutants failed to recover replication after fork collapse provoked by CPT ([Figure 5A](#pone-0013379-g005){ref-type="fig"}, middle panel), our results also indicate that Swi3\'s role in Cds1 activation is independent of the function of Swi3 in the recovery of broken replication forks. {#pone-0013379-g006} Replication abnormalities in *swi3* mutants {#s2h} ------------------------------------------- To further evaluate replication abnormalities in *swi3* mutants, we also monitored the formation of Rad22-YFP DNA repair foci in the absence of genotoxic agents. Rad22 is a homolog of budding yeast Rad52 and has been shown to bind ssDNA at the site of DNA damage [@pone.0013379-Lisby1], [@pone.0013379-Lisby2]. Depletion of *swi3* was shown to be associated with replication fork abnormalities, resulting in the strong accumulation of spontaneous Rad22-YFP DNA repair foci during unperturbed S-phase [@pone.0013379-Noguchi2] ([Figure 6B](#pone-0013379-g006){ref-type="fig"}). Therefore, we monitored the formation of spontaneous Rad22-YFP foci in the *swi3* mutants. As shown in [Figure 6B](#pone-0013379-g006){ref-type="fig"}, we observed dramatically elevated levels of Rad22-YFP foci formation in *swi3-NBT7* (Class I) and *swi3-E40* (Class I) and significantly increased levels in *swi3-E31* (Class II) ([Figure 6B](#pone-0013379-g006){ref-type="fig"}), suggesting that these mutants accumulate DNA damage probably during normal DNA replication. It is important to note that these mutants are defective in Swi1-Swi3 complex formation ([Figure 2A](#pone-0013379-g002){ref-type="fig"}). Interestingly, *swi3-E39* (Class III) cells failed to show a significant increase in spontaneous DNA damage foci ([Figure 6B](#pone-0013379-g006){ref-type="fig"}), suggesting that these cells are proficient in normal DNA replication. Since *swi3-E39* (Class III) is largely defective in the recovery of broken replication forks ([Figure 5A](#pone-0013379-g005){ref-type="fig"}), the results suggest that Swi3 has a specific function in facilitating repair of broken forks. We have also monitored Rad22-YFP in single-point *swi3* mutants and found that *swi3-D84H* (E31), *swi3-W95R* (E40), *swi3-Y111C* (E68) and *swi3-L112R* (NBT7) cells have greatly increased DNA repair foci formation ([Figure 6B](#pone-0013379-g006){ref-type="fig"}). All these mutants were defective in Swi1-Swi3 complex formation, suggesting the importance of the Swi1--Swi3 complex in suppression of spontaneous DNA damage during unperturbed DNA replication. Sister chromatid cohesion abnormalities in *swi3* mutants {#s2i} --------------------------------------------------------- We have previously found that Swi1 and Swi3 are required for proper establishment of sister chromatid cohesion [@pone.0013379-Ansbach1]. Therefore, we examined the effect of *swi3* mutations on sister chromatid cohesion. To monitor cohesion defects in *swi3* mutants, we used a strain that has the bacterial *LacO* tandem repeat sequences inserted at the *lys1* locus located in the vicinity of the centromere on chromosome I. This strain is engineered to express the LacI repressor fused to GFP-NLS, which is recruited to *LacO* repeat sequences, allowing us to visualize centromere 1 [@pone.0013379-Ansbach1], [@pone.0013379-Ding1]. If sister chromatids are properly adhered to one another, the GFP signal should resolve as a single focus in the nuclei until cells enter anaphase when cells separate two sister chromatids. However, if sister chromatids are prematurely separated, two distinct GFP foci would occur before cells enter anaphase. Using this system, we determined the effect of *swi3* mutations on cohesion at the centromere region. For synchronization, we used *nda3-KM311* cold-sensitive background to arrest cells at prophase/metaphase by culturing cells at 20°C [@pone.0013379-Hiraoka1]. Because sister chromatids are still attached to one another at prophase/metaphase, the majority of wild-type cells showed a single centromere focus in nuclei ([Figure 7A](#pone-0013379-g007){ref-type="fig"}). In contrast, the experiments revealed a significant increase in the number of nuclei with two foci in *swi3-NBT7*, *swi3-E31*, *swi3-E39* and *swi3-E40* cells ([Figure 7A and 7B](#pone-0013379-g007){ref-type="fig"}). This indicates that these mutants have a defect in efficient establishment of sister chromatid cohesion. Moreover, considering the fact that *swi3-E39* has defects in replication recovery after fork breakage but not in Cds1 activation ([Figure 5A](#pone-0013379-g005){ref-type="fig"} and [6A](#pone-0013379-g006){ref-type="fig"}), our results are consistent with the notion that the checkpoint role of Swi3 is not sufficient for proper establishment of sister chromatid cohesion. The data also suggest that *swi3-E39* has a defect in a specific function that is required to coordinate with cohesion processes. {#pone-0013379-g007} We have previously shown that *swi3*Δ is synthetically lethal with deletion of *ctf18*, which encode the largest subunit of an alternative replication factor C complex (RFC^Ctf18^) required for establishment of sister chromatid cohesion [@pone.0013379-Ansbach1], [@pone.0013379-Hanna1], [@pone.0013379-Mayer2]. *swi3-NBT7* and *swi3-E40* were found to be synthetically lethal with *ctf18*Δ (data not shown). This is consistent with the fact that these mutants displayed phenotypes similar to those of *swi3*Δ cells. Although *swi3-E31 ctf18*Δ and *swi3-E39 ctf18*Δ cells were viable, these double mutants were much more sensitive to CPT compared to either single mutant ([Figure 7C](#pone-0013379-g007){ref-type="fig"}). Importantly, although *swi3-E31* and *swi3-E39* were not sensitive to 15 µg/ml of thiabendazole (TBZ), *swi3-E31 ctf18*Δ and *swi3-E39 ctf18*Δ showed TBZ hypersensitivity ([Figure 7C](#pone-0013379-g007){ref-type="fig"}). TBZ sensitivity is found among mutants that affect general sister chromatid cohesion and segregation [@pone.0013379-Silverstein1], [@pone.0013379-Tatebayashi1], [@pone.0013379-Wang1], [@pone.0013379-Williams1]. Therefore, these results strengthen the fact that the cohesion function of Swi3 is defective in *swi3-E31* and *E39*. Discussion {#s3} ========== Programmed fork pausing and replication termination events near the mating-type (*mat1*) locus are needed to create an imprint and initiate a gene conversion event that switches mating-type in fission yeast [@pone.0013379-Dalgaard1], [@pone.0013379-Kaykov1]. These events require *swi1* ^+^ and *swi3* ^+^ genes [@pone.0013379-Dalgaard1]. Since mutations in these genes were found to be synthetically lethal with a mutation in DNA polymerase α, the role of *swi1* ^+^ and *swi3* ^+^ in DNA replication was proposed [@pone.0013379-Dalgaard1]. Accordingly, Swi1 and Swi3 have been identified and shown to form a stable complex that plays critical roles in stabilization of replication forks, activation of the replication checkpoint, and coordination of leading- and lagging-strand DNA synthesis [@pone.0013379-Noguchi1], [@pone.0013379-Noguchi2], [@pone.0013379-Sommariva1], [@pone.0013379-Lee1], [@pone.0013379-Dalgaard1], [@pone.0013379-Krings1], [@pone.0013379-Pryce1]. In addition, Swi1 and Swi3 are required for proper establishment of sister chromatid cohesion [@pone.0013379-Ansbach1]. However, the molecular mechanisms by which Swi1 and Swi3 stabilize replication forks and contribute to various replication-associated events remain elusive. Therefore, in this report, as an initial step toward dissecting the molecular pathways that require the Swi1--Swi3 replication fork protection complex, we performed mutational analyses of Swi3. Accordingly, we found separation-of-function mutations that led us to the conclusion that Swi3 utilizes different pathways to regulate the replication checkpoint and replication-dependent sister chromatid cohesion. Roles of Swi1--Swi3 complex as a checkpoint mediator {#s3a} ---------------------------------------------------- Our investigation suggest that the central conserved region of Swi3 is essential for interacting with Swi1 ([Figure 2](#pone-0013379-g002){ref-type="fig"} and [3](#pone-0013379-g003){ref-type="fig"} and [Table 1](#pone-0013379-t001){ref-type="table"}) and that Swi1--Swi3 complex formation is required for S-phase stress response ([Figures 1A and 1B](#pone-0013379-g001){ref-type="fig"}). Mutations that abolish Swi1--Swi3 complex formation sensitize cells to many different S-phase stressing agents. *swi3-NBT7*, *E40* and *E31* mutants (Classes I and II), all of which have a defect in Swi1-Swi3 complex formation, showed significant sensitivity to HU, MMS and CPT ([Figure 1A](#pone-0013379-g001){ref-type="fig"}). HU and MMS cause an arrest of the replication fork, which in turn activates the Cds1-dependent replication checkpoint. Consistently, *swi3-E31*, *swi3-NBT7* and *E40* had impaired Cds1 activity ([Figure 6A](#pone-0013379-g006){ref-type="fig"}) and had significant defects in resumption of replication after HU treatment ([Figure 5B](#pone-0013379-g005){ref-type="fig"}). Since replication resumption from HU arrest requires Cds1 [@pone.0013379-Boddy1], [@pone.0013379-Lindsay1], [@pone.0013379-Noguchi1], our data suggest that Swi1--Swi3 complex formation plays a critical role in activation of the replication checkpoint and stabilization of stalled replication forks in response to HU. In addition to the checkpoint defect in *swi3-NBT7*, *E40* and *E31*, these cells showed strong accumulation of spontaneous Rad22 DNA repair foci, indicative of DNA damage ([Figure 6B](#pone-0013379-g006){ref-type="fig"}). Consistently, when we examined single-point mutants defective in Swi1--Swi3 complex formation, cells showed dramatic accumulation of Rad22-YFP DNA repair foci in the absence of genotoxic agents ([Figure 6B](#pone-0013379-g006){ref-type="fig"} and [Table 1](#pone-0013379-t001){ref-type="table"}). Therefore, although there is a possibility that these mutants might not be solely defective in Swi1--Swi3 complex formation, our results are consistent with the notion that Swi1--Swi3 complex formation is also important to prevent DNA damage, probably during normal DNA replication. Intriguingly, all mutations affecting Swi1--Swi3 complex formation were located in one of the putative alpha helices found in the central conserved domain ([Figure 3B](#pone-0013379-g003){ref-type="fig"}, Domain II), suggesting that such alpha helix structures play an important role in protein-protein interaction. Interestingly, Swi3-E68 (K47N, Y111C) retained the ability to interact with Swi1 ([Figure 2A](#pone-0013379-g002){ref-type="fig"}), and corresponding mutant cells were sensitive to CPT, but not HU and MMS ([Figure 1A](#pone-0013379-g001){ref-type="fig"}). In contrast, *swi3-Y111C* mutant cells were highly sensitive to HU, MMS and CPT ([Figure 1B](#pone-0013379-g001){ref-type="fig"}), and the Swi3-Y111C protein failed to interact with Swi1 ([Figure 2B](#pone-0013379-g002){ref-type="fig"}). This indicates that the K47N mutation alleviates the defect of *swi3-E68* cells in Swi1--Swi3 complex formation and restores tolerance to HU and MMS, agents that activate the replication checkpoint. These results further support the idea that Swi1--Swi3 complex formation is essential for its function as a mediator of the replication checkpoint. Roles of Swi3 in the recovery of broken replication forks {#s3b} --------------------------------------------------------- It is important to note that some of the *swi3* mutants (Class III mutants: *swi3-E1, E39*, *E59* and *E68*) were only sensitive to CPT, which causes replication fork breakage ([Figure 1](#pone-0013379-g001){ref-type="fig"}). In these mutants, Swi1--Swi3 complex formation was unaffected, and cells failed to show significant HU sensitivity ([Figures 1A](#pone-0013379-g001){ref-type="fig"}, [2A](#pone-0013379-g002){ref-type="fig"} and [Table 1](#pone-0013379-t001){ref-type="table"}). Consistently, all Class III mutants had robust Cds1 activation in response to HU ([Figure 6A](#pone-0013379-g006){ref-type="fig"} and [Table 1](#pone-0013379-t001){ref-type="table"}). In addition, *swi3-E39* cells were able to normally resume DNA replication after HU-dependent fork arrest ([Figure 5B](#pone-0013379-g005){ref-type="fig"}). Since *swi3-E39* cells were not able to recover damaged replication forks provoked by CPT ([Figure 5A](#pone-0013379-g005){ref-type="fig"}), our results suggest that Swi3 regulates at least two separate pathways. The first pathway is checkpoint-dependent, which is to promote Cds1 activation and stabilize stalled replication fork in response to HU-dependent fork arrest ([Figure 8](#pone-0013379-g008){ref-type="fig"}). The second pathway is to promote efficient DNA replication and/or replication recovery after CPT-dependent fork breakage, which is independent of the Cds1-dependent replication checkpoint ([Figure 8](#pone-0013379-g008){ref-type="fig"}). This model is consistent with the previous study that reported the role of Swi3 in survival of MMS, which is also independent of Cds1- and Chk1-mediated checkpoints [@pone.0013379-Sommariva1]. It has been known that Cds1 is involved in replication fork stabilization in response to HU in *S. pombe* [@pone.0013379-Boddy1], [@pone.0013379-Lindsay1], [@pone.0013379-Noguchi1]. It has also been reported in *S. cerevisiae* that Rad53 (Cds1 homolog) is required to prevent accumulation of unusual DNA structures at the replication fork in response to fork arrest induced by HU or MMS [@pone.0013379-Lopes1], [@pone.0013379-Sogo1], [@pone.0013379-Tercero1]. Since Swi1--Swi3 is required for the chromatin association of Mrc1, which is essential for Cds1 activation [@pone.0013379-Shimmoto1], Swi1--Swi3 may regulate the replication checkpoint pathway by recruiting Mrc1 to activate Cds1 and promote fork stabilization in response to HU ([Figure 8](#pone-0013379-g008){ref-type="fig"}). However, in the presence of CPT, Cds1 is dispensable when cells restore broken replication forks ([Figure 5A](#pone-0013379-g005){ref-type="fig"}). Therefore, fork stabilization function of Cds1/Rad53 may be important when the fork is arrested by dNTP depletion (HU) or alkylation of template DNA (MMS), and this function is checkpoint-dependent. However, when cells are treated with CPT, replication forks must be recovered by a different mechanism that utilizes Swi1--Swi3, but is independent of the replication checkpoint ([Figure 8](#pone-0013379-g008){ref-type="fig"}). It is possible that Swi1--Swi3 facilitates efficient repair of broken replication forks, although further investigation is needed to address this possibility. It is also feasible that Swi1--Swi3 promotes DNA replication after DSBs at forks have been repaired. Therefore, we propose a model in which Swi1--Swi3 is involved in at least two processes during fork recovery. First, Swi1--Swi3 is required to resume arrested replication fork in a replication checkpoint-dependent manner. This process can be referred to as "fork stabilization" ([Figure 8](#pone-0013379-g008){ref-type="fig"}). Second, Swi1--Swi3 may also be important to re-capture replication fork and/or re-assemble replisome components when forks are broken. This "fork regeneration process" is independent of the replication checkpoint ([Figure 8](#pone-0013379-g008){ref-type="fig"}). Our results are consistent with the idea that Class I and II mutants have defects in both "fork stabilization" and "fork regeneration" processes, while *swi3-E39* mutant (Class III) is proficient in "fork stabilization" but defective in "fork regeneration". In budding yeast, it is proposed that Tof1-Csm3-Mrc1 form a "fork pausing complex", which is required to stabilize stalled replication forks [@pone.0013379-Katou1]. In this model, the fork pausing complex is involved in coupling of polymerases and helicases at stalled replication forks. However, in *S. pombe*, Swi1--Swi3 (Tof1-Csm3 homolog) only weakly associates with Mrc1, while the interaction between Swi1 and Swi3 is tight [@pone.0013379-Noguchi2], [@pone.0013379-Shimmoto1]. In addition, our data strongly support the idea that Swi1--Swi3 also plays a role in fork-recapture and/or -reassembly when forks are actually broken. Therefore, we prefer the model in which Swi1--Swi3 functions as a "fork protection complex" that promotes both fork-stabilization and fork-regeneration processes in response to various genotoxic agents ([Figure 8](#pone-0013379-g008){ref-type="fig"}). {#pone-0013379-g008} Roles of Swi3 in replication-coupled sister chromatid cohesion {#s3c} -------------------------------------------------------------- The present studies revealed that a separation-of-function mutation of Swi3, which render cells sensitive specifically to CPT, also caused sister chromatid cohesion defects comparable to *swi3* deletion mutants. This mutant (*swi3-E39*) also had defects in recovery of broken replication forks but not in resumption of arrested forks ([Figure 5](#pone-0013379-g005){ref-type="fig"}), the latter of which is dependent on the Cds1-dependent replication checkpoint. It has been thought that proteins involved in replication checkpoint safeguard sister chromatid cohesion [@pone.0013379-Warren1]. While this is true, our present results are consistent with the notion that the checkpoint and cohesion roles of Swi3 are separable, and that the replication checkpoint function of Swi3 is not sufficient for cohesion process. Intriguingly, fork-regeneration function of Swi3 is coupled with sister chromatid cohesion ([Figure 8](#pone-0013379-g008){ref-type="fig"}). Therefore, we propose that the replication checkpoint and chromosome cohesion function in separate pathways. We also propose that Swi1--Swi3 has a key role in replication-coupled sister chromatid cohesion established at the replication fork. Consistently, we have shown that Timeless interacts with cohesin subunits in human cells [@pone.0013379-Leman1]. Moreover, Timeless downregulation led to dissociation of cohesin subunits from chromatin and defects in sister chromatid cohesion in human cells [@pone.0013379-Leman1]. Interestingly, we have also demonstrated in both *S. pombe* and human cells that Swi1--Swi3^Timeless-Tipin^ acts together with Chl1^ChlR1^, a DNA helicase known to be required for establishment of sister chromatid cohesion [@pone.0013379-Leman1], [@pone.0013379-Ansbach1]. Therefore, we suggest that Swi1--Swi3^Timeless-Tipin^ and Chl1^ChlR1^are in the same pathway to control fork regeneration and cohesion processes ([Figure 8](#pone-0013379-g008){ref-type="fig"}). Recent studies have shown the role of sister chromatid cohesion in the repair of DSBs [@pone.0013379-Strom1], [@pone.0013379-Unal1]. Therefore, we also speculate that improper cohesion in the absence of Swi3 can affect efficient repair of DSBs at replication forks when cells are treated with camptothecin. Therefore, it is possible that Swi1--Swi3 facilitates sister chromatid cohesion to promote efficient recapture of the fork during recombination processes, which also contribute to the regeneration of replication forks ([Figure 8](#pone-0013379-g008){ref-type="fig"}). Materials and Methods {#s4} ===================== General Techniques {#s4a} ------------------ The methods used for genetic and biochemical analyses of fission yeast have been described previously [@pone.0013379-Alfa1], [@pone.0013379-Moreno1]. PCR amplification of DNA was done using EX taq DNA polymerase (TaKaRa, Ohtsu, Japan). Accurate PCR reactions were confirmed by DNA sequencing analyses. Western blotting, Cds1 kinase assay, and drug sensitivity assays were performed as described in our earlier studies [@pone.0013379-Ansbach1], [@pone.0013379-Noguchi3]. For immunoblotting, Myc, TAP, and FLAG fusion proteins were probed with the anti-c-Myc 9E10 monoclonal antibody (Covance, Berkeley, CA), PAP (Peroxidase Anti-Peroxidase Soluble Complex antibody) (Sigma-Aldrich, St. Louis, MO), and the anti-FLAG M2 monoclonal antibody (Sigma-Aldrich), respectively. TAT-1 [@pone.0013379-Woods1] was used to detect tubulin. Microscopic analyses of green fluorescent protein (GFP) and yellow fluorescent proteins (YFP) were performed using Olympus PROVIS AX70 microscope equipped with a Retiga EXi camera (QImaging, Surrey, BC, Canada). Images were acquired with Ivision software (BioVision Technologies, Exton, PA). Plasmids {#s4b} -------- Genomic DNA was isolated from *S. pombe* cells containing the *swi3-TAP-kanMX6* gene [@pone.0013379-Noguchi2]. The 1.7 kb *swi3-TAP* genomic fragment including the *swi3* promoter region was amplified by PCR from this genomic DNA preparation, and subsequently cloned into the XbaI/KpnI site of pJK148 [@pone.0013379-Keeney1] to generate pJK148-swi3-TAP. The 1.3 kb mutant *swi3-5FLAG* fragments were amplified by PCR from genomic DNA prepared from *swi3* mutants, and cloned into the XbaI/BamHI site of pJK148 to generate pJK148-swi3-5FLAG. The 1.5 kb NotI-BglII fragment containing a C-terminal *rad22* region fused with *YFP* cDNA [@pone.0013379-Noguchi1], [@pone.0013379-DuLL1] was introduced into the NotI/BamHI site of pJK210 [@pone.0013379-Keeney1], resulting in pJK210-rad22-YFP-CT. *S. pombe* strains {#s4c} ------------------ The *S. pombe* strains used in this study were constructed using standard techniques [@pone.0013379-Alfa1], and their genotypes are listed in Supplementary [Table S1](#pone.0013379.s001){ref-type="supplementary-material"}. *swi1-13Myc* (*swi1-13Myc-hphMX6*), *swi3-13Myc* (*swi3-13Myc-hphMX6*) and *ctf18*Δ (*ctf18*::*hphMX6*) were generated by a one-step marker switch method [@pone.0013379-Sato1] using the *swi1-13Myc-kanMX6*, *swi3-13Myc-kanMX6* and *ctf18*::*kanMX6* strains, respectively. Single-point *swi3* mutants were generated by Kunkel site-directed mutagenesis [@pone.0013379-Kunkel1] in pJK148-swi3-TAP, and integrated at the *leu1* locus of the *swi3*::*kanMX6 swi1-3FLAG-kanMX6* strain. To visualize Rad22-YFP in *swi3* mutants, pJK210-Rad22YFP-CT was integrated at the *rad22* locus of the *swi3* mutant strains. To monitor cohesion defects, pJK148-swi3 (wild-type or mutants)-5FLAG was integrated at *leu1* locus of an *S. pombe* strain containing *nda3-KM311*, *swi3*::*KanMX6*, *lys1* ^+^:*LacO* repeat and *his7* ^+^:*GFP-LacI-NLS*. Mutations and epitope-tagged genes have previously been described for *swi1*Δ (*swi1*::*kanMX6*) [@pone.0013379-Noguchi1]; *swi1-13Myc* (*swi1-13Myc-kanMX6*), *swi1-3FLAG* (*swi1-3FLAG-kanMX6*), *swi3*Δ (*swi3*::*kanMX6*), *swi3-TAP* (*swi3-TAP*-*kanMX6*), *swi3-3FLAG* (*swi3-3FLAG-kanMX6*), *swi3-13Myc* (*swi3-13Myc-kanMX6*) [@pone.0013379-Noguchi2], *cds1*Δ (*cds1*::*kanMX6*), *chk1*Δ (*chk1*::*kanMX6*), *rad3*Δ (*rad3*::*kanMX4*), *ctf18*Δ (*ctf18*::*kanMX6*) [@pone.0013379-Ansbach1], *rad26*Δ (*rad26*::*ura4* ^+^) [@pone.0013379-alKhodairy1], *nda3-KM311* [@pone.0013379-Hiraoka1], and *lys1* ^+^-*LacO* repeat *his7* ^+^-*dis1promoter*-*GFP-LacI-NLS* [@pone.0013379-Ding1]. Isolation of *swi3* mutants {#s4d} --------------------------- Genomic DNA was isolated from *S. pombe* cells containing the *swi3-5FLAG-kanMX6* gene [@pone.0013379-Noguchi2]. The 2.9 kb *swi3-5FLAG-kanMX* genomic fragment was amplified from this genomic DNA preparation by PCR, and subsequently cloned into the AdhI site of pBluescript II TKS (+) [@pone.0013379-Ichihara1] to generate the pTKS-swi3-5FLAG-kanMX construct. Error-prone PCR was performed using five- and threefold higher than recommended concentrations of EX taq DNA polymerase and dNTPs, respectively. The wild-type *swi3* ^+^ gene was replaced with the mutagenized *swi3-5FLAG-kanMX6* gene at the *swi3* locus by a standard transformation method. Kanamycin-resistant colonies were isolated and their growth was examined to select for hydroxyurea- and camptothecin-sensitive mutants. This method generated eight *swi3* mutants, which are designated *swi3-E1*, *swi3-E10*, *swi3-E31*, *swi3-E39*, *swi3-E40*, *swi3-E42*, *swi3-E59* and *swi3-E68*. We also isolated the *swi3-NBT7* mutant by selecting for mating-type switching defective mutants. Precipitation of TAP and FLAG-tagged proteins {#s4e} --------------------------------------------- Precipitation of TAP-tagged proteins were performed using immunoglobulin G-Sepharose beads (GE Healthcare, Piscataway, NJ) as previously described [@pone.0013379-Ansbach1]. For precipitation of FLAG-tagged proteins, cells expressing FLAG-fusion proteins were cultured in YES medium and collected when an optical density of 1.2 at 600 nm was reached. Cells were then lysed with glass beads in lysis buffer A {50 mM Tris-HCl (ph 8.0), 150 mM NaCl, 0.1% NP-40, 10% glycerol, 50 mM NaF, 1 mM Na~3~VO~4~, 5 mM EDTA, 5 mM *N*-methylmaleimide, 1 µM microcyctin, 0.1 µM okadaic acid, 0.2 mM *p*-4-amidoinophenyl-methane sulfonyl fluoride hydrochloride monohydrate (*p*-APMSF) and Roche complete EDTA-free protease inhibitor cocktail} using a FastPrep cell disrupter (Qbiogene, Irvine, CA) for two cycles of 20 seconds each at speed 6, with a one-minute interval on ice between the two cycles. Protein extracts were clarified by centrifugation at 13,000 rpm in an Eppendorf microcentrifuge 5415D for 10 min at 4°C, mixed with anti-FLAG M2 agarose (Sigma-Aldrich) and incubated for 2 hr at 4°C. The agarose beads were collected and washed three times in lysis buffer A. Proteins associated with the beads were analyzed by Western blotting. Pulsed-field gel electrophoresis (PFGE) {#s4f} --------------------------------------- Exponentially growing cells were treated with the indicated amount of camptothecin (CPT) or hydroxyurea (HU) for 3 h at 30°C, and then they were washed and released into fresh YES medium. Cells were collected at the indicated times, and chromosomal DNA samples were prepared in agarose plug and analyzed with CHEF-DRII system (Bio-Rad) as previously described [@pone.0013379-Ansbach1], [@pone.0013379-Noguchi3]. Detection of Rad22-YFP DNA repair foci {#s4g} -------------------------------------- Cells expressing Rad22-YFP foci from its own promoter were grown at 25°C in YES liquid medium until mid-log phase, and then Rad22-YFP localization was analyzed as previously described [@pone.0013379-Ansbach1], [@pone.0013379-Noguchi3]. At least 200 cells were counted for each strain in each experiment. Chromosome cohesion assay {#s4h} ------------------------- Chromosome cohesion assay was performed as described previously [@pone.0013379-Ansbach1]. We used a cold-sensitive *nda3-K311* strain harboring bacterial *LacO* tandem repeat sequences inserted in the vicinity of the centromere on chromosome 1 [@pone.0013379-Ding1]. This strain is engineered to express the LacI repressor fused to GFP-nuclear localization signal (NLS), which is recruited to *LacO* repeat sequences, allowing to visualize the centromere 1 [@pone.0013379-Ding1]. The *nda3-K311* cells were grown to mid-log phase at 30°C and shifted to a restrictive temperature, 20°C. At the indicated time, GFP foci were monitored and imaged. Quantification of GFP foci has been performed at least three times, and at least 200 cells were counted for each strain in each experiment. Supporting Information {#s5} ====================== ###### *S. pombe* strains used in this study. (0.05 MB DOC) ###### Click here for additional data file. We thank Adam Leman, Laura Roseaulin, Robert Skibbens and Sonya Vengrova for their critical reading and helpful comments. We also thank Keith Gull and Teresa Wang for providing the TAT-1 and anti-Cds1 antibodies, respectively. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by funds from Leukemia Research Foundation (to E.N.) and National Institutes of Health (GM077604, to E.N.). This work was also supported by The American Recovery and Reinvestment Act of 2009 (R01GM077604-S1, to E.N.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: JBR CN MMD EN. Performed the experiments: JBR CN MMD LKW EN. Analyzed the data: JBR CN MMD AMH BA EN. Contributed reagents/materials/analysis tools: JBR CN MMD LKW ABA AMH BA EN. Wrote the paper: JBR EN.

Introduction {#s1} ============ The human body consists of numerous cell types that are highly organized into functional units constituting tissues and organs. Expression patterns of genes have been under selection for eons and, as a result, cell types and tissues differ from each other both morphologically and functionally. The mechanisms leading to the development, differentiation, and maintenance of tissues have been under intensive investigation by generations of scientists. A generally accepted view of gene expression programs divides genes in two main categories: i) housekeeping genes that are virtually always expressed in every tissue and work to maintain basic cellular functions; and ii) genes whose expression is preferential in one or a few tissues and which provide specialized functions that have a strong effect on the physiology of the whole organism. Compared to the housekeeping genes, tissue-specific genes have been described as longer [@pone.0001880-Vinogradov1], with longer introns [@pone.0001880-CastilloDavis1], a lower GC content [@pone.0001880-Vinogradov2], and lower substitution rates at non synonymous sites [@pone.0001880-Duret1]. Moreover, tissue-specific genes seem to evolve faster and they are more likely to be mutated in genetic diseases with Mendelian inheritance [@pone.0001880-Winter1]. In terms of gene expression, tissue specificity can be addressed in strict terms of genes that are exclusively transcribed in only one particular tissue type, but there is evidence indicating that most tissues with similar function share many expression patterns. Therefore, the concept of tissue-selectivity, which considers those genes whose expression is enriched in one or more similar tissues [@pone.0001880-Liang1], might be more useful. Affymetrix high-density oligonucleotide arrays [@pone.0001880-Lipshutz1] have been already used for investigating tissue-specific expression patterns [@pone.0001880-Liang1], [@pone.0001880-Su1]. However, there are several problematic aspects in the GeneChip technology, related especially to the mis-annotation of many probes. Dai and collaborators [@pone.0001880-Dai1] have observed that updating the probe annotation for all the Affymetrix chipsets affects a large number of the probe sets. More recently, it has been shown that updating the definitions of the Affymetrix probes leads to more precise and accurate results as compared with the original annotations provided by the manufacturer [@pone.0001880-Sandberg1]. Re-annotation of the Affymetrix probes has been also shown to improve the cross-platform reproducibility and meta-analysis of independent microarray experiments [@pone.0001880-Carter1]. The aim of this study was to investigate tissue-selective expression patterns, integrating publicly available gene expression data. A total of 195 images of Affymetrix GeneChips were collected from the GEO database (<http://www.ncbi.nlm.nih.gov/geo/>). All probes present on the chipset were re-annotated according to the latest release of the Entrez Gene database (<http://www.ncbi.nlm.nih.gov/sites/entrezdbgene>). After extended quality control and preprocessing, we explored the tissue-selective expression patterns. Results {#s2} ======= Identification of tissue-selective genes {#s2a} ---------------------------------------- We searched for genes expressed in a tissue-selective manner. A tissue-selectivity score was computed for each gene and used as a weight for the expression values. After permutation test we could identify 1601 genes selectively expressed in one or more tissues. About 35% of 1601 genes were selectively expressed only in one tissue, 20% were shared by two, and 13% by three tissues. Ten percent of the tissue-selective genes were shared by six or more tissues. The majority of tissue-selective genes shared by ten or more tissues were expressed in neural system tissues. The majority of the tissue selective genes were found in the immune system (32% of 1601), followed by central and peripheral nervous systems (17%), muscles (15%) and reproductive organs (9%). Altogether, the other categories accounted 27% of selective genes. Functions of tissue-selective genes {#s2b} ----------------------------------- The 1601 tissue-selective genes covered a wide range of cellular and molecular functions as they could be annotated into 1694 distinct Biological Process, 1094 Molecular Function and 290 Cellular Component functional families from the three gene ontology classifications ([File S1](#pone.0001880.s004){ref-type="supplementary-material"}, tables 0.2, 0.3 and 0.4). The gene ontology classification revealed a suggestive distribution of the 1601 tissue-selective genes into functional families: 19% of them were classified in the Molecular Function family "signal transducer activity", and about 8% in the group of "receptor binding" proteins. Moreover, when the same genes were grouped according to the Cellular Component ontology, about 18% were annotated under the family "extracellular region". The classification of the 1601 tissue-selective genes according to the Biological Process ontology also showed that about 16% were associated with the term "development", and almost 14% to the term "immune response". The gene ontology annotation was also used to characterize the genes identified in each of the 78 tissues separately (details in [File S1](#pone.0001880.s004){ref-type="supplementary-material"}). For example, "blood coagulation", "iron ion homeostasis", "lipid metabolism", and "gluconeogenesis" were found over-represented in liver with p\<1.0E-4 ([File S1](#pone.0001880.s004){ref-type="supplementary-material"}, table 44.2). Similarly, "male gamete generation" and "spermatogenesis" were found in testis with p\<1.0E-18 ([File S1](#pone.0001880.s004){ref-type="supplementary-material"}, table 55.2). Generally, the selective genes in all the analyzed tissues showed excellent correlation with the known physiological functions. Tissue-selective genes and human diseases {#s2c} ----------------------------------------- It has been suggested that slow-evolving housekeeping genes are underrepresented among disease genes, due to a higher chance of embryonic lethality when mutated [@pone.0001880-Winter1]. The 1601 tissue-selective genes were enriched in disease-genes as they were linked to 361 diseases described in the OMIM database (<http://www.ncbi.nlm.nih.gov/sites/entrezdbOMIM>). We also observed that, in most of the cases, the 1601 tissue-selective genes are associated to pathological phenotype in the tissues or organs where they are found selectively expressed. For instance, among the genes we found selectively expressed in fetal heart, GATA4 and NKX2.5 have been associated with Atrial Septal Defect 2 [@pone.0001880-Garg1], [@pone.0001880-HirayamaYamada1]. In addition, mutations of the gene NKX2.5 have been described in Tetralogy of Fallot [@pone.0001880-Benson1], [@pone.0001880-Goldmuntz1] and in Atrial Septal Defect with Atrioventricular Conduction Defect [@pone.0001880-Schott1], [@pone.0001880-Watanabe1]. The placenta-selective gene RASA1 is reported in two diseases characterized by aberrations of blood vessels: the Parkes Weber Syndrome and Capillary Malformation-Arteriovenous Malformation [@pone.0001880-Eerola1]. Several muscle-selective genes have been described in a number of myopathies, as well as several gland-selective genes are associated with syndromes of the endocrine system and metabolic diseases. More extensive listing of diseases is available in the supplemental results ([File S1](#pone.0001880.s004){ref-type="supplementary-material"}, table 0.5). Tissue connectome {#s2d} ----------------- We wanted to investigate the hypothesis that tissues and organs sharing tissue-selective genes might present some degree of relatedness. For this propose, we built a network of tissues that we named connectome. In the connectome, each node represented a tissue and the genes selectively expressed in two or more tissues formed the edges between the nodes. Each edge was thus associated with the number of shared genes. The number of edges in the network was computed as a function of the number of shared genes between the tissues and three cutoff values (30 or more genes, 20 or more genes, and 5 or more genes) were chosen as representative of different degrees of connectivity ([Figure 1d](#pone-0001880-g001){ref-type="fig"}). {ref-type="table"}.](pone.0001880.g001){#pone-0001880-g001} ### The 30 sharing genes connectome {#s2d1} It was possible to observe separate networks of seven central nervous system (CNS) tissues, five testis tissues, twelve immune cells, and six muscles ([Figure 1a](#pone-0001880-g001){ref-type="fig"}). Moreover, a chain-like connection was found between fetal lung -- fetal liver -- liver -- kidney. The forebrain structures, consisting of telencephalic and diencephalic structures, clustered together. The central tissue in the CNS cluster was amygdala, which is consistent with the view that amygdala, rather than being a structural and developmental unit, is a collection of adjacent cell groups within the forebrain [@pone.0001880-Swanson1], [@pone.0001880-Swanson2]. Whole blood had extended connections with several circulating cells, such as T cells, B cells, natural killer cells, monocytes, and BDCA4-dendritic cells. On the other hand, tight connections were also formed between the cells resident in the bone marrow, such as early erythroid cells, B lymphoblasts, endothelial cells, and CD34 clones. ### The 20 sharing genes connectome {#s2d2} Two larger sub-networks emerged, constituted respectively of CNS tissues, and immune cells and tissues, muscles, excretory organs, and thyroid ([Figure 1b](#pone-0001880-g001){ref-type="fig"}). Tonsil bridged the connection between immune cells and muscles. This makes sense as in the tonsil there are both myoepithelial and immune cells. Tonsil presented also connections with B cells, BDCA4-dendritic cells, and lymph node. Histologic studies of the tonsil showed that they are lymphoid structures consisting mainly of B-lymphocytes, but they are occupied also by T-lymphocytes, activated B-lymphocytes and other cells of the immune system. Tonsil shares histological features with lymph nodes as its cells are supported by a fine network of reticular fibers and high-endothelial venules function in the "homing" of circulating lymphocytes [@pone.0001880-CA1], [@pone.0001880-AK1]. During the fetal life massive erythropoiesis happens in several tissues such as the liver [@pone.0001880-Timens1]. Fetal liver, in fact, connected early erythroid cells and CD105-endothelial cells that are precursors of circulating cells and which reside in the bone marrow after the birth. The testis tissues still appeared unconnected to any other tissue. The cluster of 15 CNS tissues included structures spanning from the spinal cord to telencephalic structures, forming the center of this network. ### The 5 sharing genes connectome {#s2d3} All CNS tissues, including the olfactory bulb clustered together ([Figure 1c](#pone-0001880-g001){ref-type="fig"}). In addition, some peripheral nervous system structures joined this cluster. This sub-network was no longer distinct but shared links with other structures, particularly fetal tissues such as heart and lung. In addition, muscles shared links with the CNS cluster, which may be due to the innervation of the muscle samples as well as to the genes involved in ion homeostasis, which are expressed both in neurons and muscular cells. Fetal brain, hippocampus and olfactory bulb bridged the CNS cluster with other tissues. Neurogenesis, production of new neurons, continues in adult hippocampus and olfactory bulb and the genes expressed in these newborn neurons may give them immature characteristics, which are shared by fetal brain and other fetal tissues [@pone.0001880-Gage1]. Hypothalamus and pituitary gland, which are anatomically connected, linked together. The hippocampus regulatory gene network {#s2e} --------------------------------------- One of the goals of our analysis was to find clear correspondence between the tissue-selective gene expression programs and specific functions of tissues. Within the CNS, neurogenesis during adult life takes place in the hippocampus during normal and pathological conditions. The analysis of the connectome showed interesting links of the hippocampus with other anatomical structures where cell duplication and differentiation are known to happen. We wanted to test the hypothesis that genes selectively expressed in hippocampus would form a transcriptional network underlying this specific function. For this, we built a network of the hippocampus-selective genes based on their co-citation into the Pubmed database as well as the presence of specific transcription factor binding sites (TFBS) in their promoter regions ( [Figure S1](#pone.0001880.s002){ref-type="supplementary-material"}). The transcription factor NF-kappaB, which was not among the hippocampus-selective genes, presented an interesting topological position as it had connections with a number of hippocampus-selective genes ([Figure S1](#pone.0001880.s002){ref-type="supplementary-material"}). Detailed analysis of the promoters of the NF-kappaB interactors revealed the presence of a significantly conserved binding sequence for E2F and, 92--115 bp downstream, one for NF-kappaB ([Figure S2](#pone.0001880.s003){ref-type="supplementary-material"}). Screening the whole set of known human promoters, we found the E2F-NF-kappaB module in 1901 regulatory sequences, suggesting a common mechanism of transcriptional regulation. The gene ontology clustering of these genes showed significant over-representation of the families "nervous system development", "cell adhesion", "transmembrane receptor protein tyrosine kinase signaling pathway", and "retinoic acid receptor activity" (details in [File S2](#pone.0001880.s005){ref-type="supplementary-material"}). In addition to the hippocampus-selective genes, also some fetal brain-selective, amygdala-selective, and prefrontal cortex-selective genes could be regulated by the E2F-NF-kappaB module. All these areas have been extensively investigated for neurogenesis [@pone.0001880-Gould1]. Discussion {#s3} ========== We have integrated microarray data produced in several laboratories for exploring the tissue-selective expression patterns in 78 normal human tissues. One of the interests about the tissue-specific genes concerns their functional role in normal and pathological conditions. We observed that the group of tissue-selective genes is enriched in "signal transducer activity", "receptor binding", and "extracellular region", as well as "development" and "immune response" functional families. Our results are largely concordant with the findings of Winter and collaborators, who have reported that genes encoding secreted proteins highly correlate with tissue specificity [@pone.0001880-Winter1]. Freilich and collaborators also reported that genes with tissue-specific expression patterns mainly encode for regulatory proteins involved in signal transduction activity [@pone.0001880-Freilich1]. In addition, they observed that the tissue-specific group is possibly enriched in transcription factors encoding genes [@pone.0001880-Freilich1]. We found only 59 transcription factor genes expressed in a tissue-selective manner. Accordingly, Yu and collaborators have reported that ubiquitously expressed transcription factors can combine with other factors contributing to tissue-specificity. In addition, they observed that individual transcription factors can participate in tissue-specific gene regulation by interacting with distinct partners in different tissues [@pone.0001880-Yu1]. We propose that many transcription factors are evenly expressed in many tissues and that several stimuli mediated by tissue-specific signal transduction machineries mediate the functional activation of specific combinations of them at the protein level only in certain tissues and in defined temporal windows. The connectome shows a novel and intriguing way to investigate the relatedness of human tissues and organs. While interpreting these results, it should be also taken into account that many human tissues have a complex architecture, as they consist of a certain number of specialized cells with variant transcriptional profiles. Microarrays can reliably detect cRNA species at the concentration of a few pico-molars, but it can be problematic to assign a certain gene expression event to the correct cell subpopulation of complex tissues. Nevertheless, we believe that the tissues should be always studied as functional entities and their global gene expression should be target of interest. When thinking of the liver, for instance, it should be considered that the identity of this organ is given by the combination of expression programs in several kinds of cells, more than just hepatocytes. Moreover, it is reasonable to think that in samples of many tissues some amount of blood cells are also present. This can easily explain the wide connections that circulating cells form with other kinds of anatomical structures in terms of shared gene expression. However, we observed a very tight intra-connectivity within the groups of nervous tissues, blood cells, testis tissues, and muscles, suggesting that the identity of these anatomical structures is determined by the differential expression of many genes. This is also evident from several clustering analyses we preformed on the data (details in [File S3](#pone.0001880.s006){ref-type="supplementary-material"}). The connectome of tissues also shows interesting interactions largely explained by the functional and morphological similarity of some tissues. It is the case, for instance, of the connection between the tonsil and lymph node. In other cases, the topological features of certain tissues in the network are suggestive of developmental mechanisms, as for the central position of the amygdala within the central nervous system connectome. Increasing attention is being oriented to the inference of transcriptional regulatory networks from high throughput gene expression screenings. These approaches aim to link gene expression data to the activity of transcription factors in cause-effect models. Some effort has been already put also into the investigation of regulatory gene networks of tissue-specific genes having a central role in the physiology and development of specific anatomical structures [@pone.0001880-Wasserman1], [@pone.0001880-Krivan1], [@pone.0001880-Olson1]. We investigated in detail the expression patterns that might play a role in determining some functional aspect of the hippocampus. We computationally predicted a promoter module formed by conserved consensus motifs for the transcription factors E2F and NF-kappaB present on 1901 human known promoters. The gene ontology classification suggests that these genes are directly involved in neurogenesis and central nervous system development. The E2F family of transcription factors, by interaction with several partners such as pRb, p107 or p130, are thought to regulate the cell cycle [@pone.0001880-Attwooll1], [@pone.0001880-Trimarchi1] and trigger signals that also either promote cellular growth, cell cycle exit, or terminal differentiation in neurons [@pone.0001880-Trimarchi1], [@pone.0001880-Liu1]. NF-kappaB has been described as playing an important role in synaptic activity and plasticity, neuroprotection, and in behavioral aspects of learning and memory [@pone.0001880-Meffert1]. Moreover, members of NF-kappaB family have been found to be expressed in areas of active neurogenesis in post-natal and adult mouse brains [@pone.0001880-DenisDonini1]. Altogether, these results expand our understanding of how gene expression programs determine the functional identity of human tissues. Methods {#s4} ======= Data collection {#s4a} --------------- Total number of 195 Affymetrix HG-U133A CEL files was collected from the GEO database (<http://www.ncbi.nlm.nih.gov/geo/>) from 6 different data sets. All the gathered arrays had been hybridized to normal adult or fetal human tissues or cell types for a total of 78 different classes ([Table 1](#pone-0001880-t001){ref-type="table"}). Data were selected according to the following criteria: i) all the experiments had been documented according to the MIAME standard; ii) all the arrays had been hybridized with samples isolated from human tissues and experiments were not carried out with cell lines; iii) all the samples came from healthy control subjects or from reference RNA samples; iv) the raw array images (CEL files) were available for download; and v) the Affymetrix chipset used for the hybridization was the human HGU-133A. A quality check of the data was performed using the recommendations of the manufacturer. Altogether, 195 sets of individual array data were used for further analysis. 10.1371/journal.pone.0001880.t001 ###### Summary of the data set analyzed. {#pone-0001880-t001-1} tissue category Internal tissue code -------------------------- -------------- ---------------------- Hippocampus neural 1 BronchialEpitelia respiratory 2 LimbMuscle muscle 3 ExtraocularMuscle muscle 4 Kidney excretory 5 SubthalamicNucleus neural 6 Skin connective 7 GlobusPallidus neural 8 CiliaryGanglion neural 9 AtrioVentricularNode neural 10 DRG neural 11 Placenta reproductive 12 721-BLymphoblasts immune 13 PB-CD8TCells immune 14 PB-CD4TCells immune 15 BM-CD71EarlyErythroid immune 16 PB-CD14Monocytes immune 17 PB-CD56NKCells immune 18 PB-CD19BCells immune 19 BM-CD33Myeloid immune 20 BM-CD105Endothelial immune 21 BM-CD34 immune 22 PB-BDCA4DentriticCells immune 23 SuperiorCervicalGanglion neural 24 MedullaOblongata neural 25 Pons neural 26 Appendix immune 27 TrigeminalGanglion neural 28 TemporalLobe neural 29 Tongue muscle 30 UterusCorpus reproductive 31 PsoasMuscle muscle 32 FetalLung respiratory 33 ParietalLobe neural 34 Tonsil immune 35 WholeBlood circulatory 36 HBEC circulatory 37 SalivaryGland gland 38 AdrenalCortex gland 39 Ovary reproductive 40 Thyroid gland 41 Lung respiratory 42 FetalBrain neural 43 Liver excretory 44 LymphNode immune 45 Amygdala neural 46 Heart muscle 47 Uterus reproductive 48 Prostate gland 49 Pancreas gland 50 PrefrontalCortex neural 51 CingulateCortex neural 52 Thymus immune 53 FetalLiver excretory 54 Testis reproductive 55 Trachea respiratory 56 AdrenalGland gland 57 SpinalCord neural 58 Cerebellum neural 59 PituitaryGland gland 60 Thalamus neural 61 BoneMarrow immune 62 CardiacMyocytes muscle 63 FetalThyroid gland 64 OlfactoryBulb neural 65 TestisGermCell reproductive 66 TestisIntersitial reproductive 67 TestisLeydigCell reproductive 68 Hypothalamus neural 69 OccipitalLobe neural 70 CerebellumPeduncles neural 71 SmoothMuscle muscle 72 CaudateNucleus neural 73 WholeBrain neural 74 Islet gland 75 Adipocytes connective 76 TestiSeminiferousTubule reproductive 77 FetalHeart muscle 78 Data Pre-processing {#s4b} ------------------- Sequence-based re-annotation of the Affymetrix probes on an HGU-133A chipset [@pone.0001880-Dai1] according to the latest release of the Entrez Gene database was used (<http://www.ncbi.nlm.nih.gov/sites/entrezdbgene>). The expression values for each gene were calculated using the RMA algorithm [@pone.0001880-Irizarry1]. Tissue-selectivity analysis {#s4c} --------------------------- A tissue-selectivity score was computed for each tissue-gene pair from the expression data matrix. Permutation test was performed to define a significance threshold. Details in [Table S1](#pone.0001880.s001){ref-type="supplementary-material"}. Gene ontology analysis {#s4d} ---------------------- Fisher\'s exact test was performed in order to select over-represented gene ontology classes in the tissue-selective genes. The functional families presenting p\<0.01 were considered as significantly represented. Gene network and promoter analysis {#s4e} ---------------------------------- The hippocampus-selective genes were processed in the software Bibiosphere to build up gene networks based on their co-citation in the literature as well as the presence of TFBS for known transcription factors in their promoter regions (<http://www.genomatix.de/products/BiblioSphere/>). Because of the extensive connectivity of NF-KappaB within the network, the genes presenting a significant TFBS for NF-KappaB were selected for further analysis. The promoter sequences of these genes were retrieved using the Gene2Promoter software (<http://www.genomatix.de/online> help/help eldorado/Gene2Promoter Intro.html) and analyzed with FrameWorker (<http://www.genomatix.de/online> help/help gems/FrameWorker.html) to search for common models containing at least two TFBS. Finally, the significant model constituted by E2F and NF-KappaB was screened for in the whole set of known human promoters using ModelInspector (<http://www.genomatix.de/online> help/help fastm/modelinspector help.html). Supporting Information {#s5} ====================== ###### (0.06 MB DOC) ###### Click here for additional data file. ###### (0.12 MB PDF) ###### Click here for additional data file. ###### (0.04 MB PDF) ###### Click here for additional data file. ###### (7.50 MB PDF) ###### Click here for additional data file. ###### (0.82 MB PDF) ###### Click here for additional data file. ###### (0.17 MB PDF) ###### Click here for additional data file. The authors would like to thank Dr. John R. Walker (Novartis Research Foundation, USA) for kindly providing the raw data for the GEO dataset GDS596, Dr. Alan Schulman (University of Helsinki, Finland) and Dr. Roy Siddall (University of Helsinki, Finland) for comments on the manuscript. The authors are also grateful to the Finnish IT Centre for Science for providing the supercomputing facilities, Dr. Martin Seifert (Genomatix Software GmbH, Germany) for helpful advice concerning the Genomatix software. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This research was funded by the Institute of Biotechnology, University of Helsinki, and by the Academy of Finland (grant 117968). [^1]: Conceived and designed the experiments: DG. Analyzed the data: EC DG PS AD TR LN PA. Wrote the paper: EC DG PS AD PA.

1. Introduction {#sec1} =============== Practicing physicians leave medicine for multiple reasons including illness, family responsibilities, career dissatisfaction or change, substance use, and early retirement \[[@B1], [@B2]\]. Returning to practice can be challenging for many reasons. State licensing boards and hospital credentialing processes have various re-entry policies, including demonstration of baseline knowledge, ongoing continuing medical education (CME), and completion of a formal refresher program. Some recruiters require re-entry re-education with a six-month clinical gap. The American Medical Association recommends that a physician absent from practice for 2 or more years participates in a formal program capable of assessing essential clinical competencies and refreshing relevant skills for workforce re-entry \[[@B3]\]. The American Board of Anesthesiology written policy on reattaining certification status dictates that consideration is an individualized, case-by-case basis. Registrants may be required to "undertake continuing medical education, complete additional training, and complete other activities deemed necessary" \[[@B4]\]. In addition to professional and institutional challenges, an anesthesiologist re-entering practice after a period of inactivity may face personal obstacles as well. Low self-confidence, limited resources in acquiring up-to-date knowledge, and a lack of professional network may limit success. Few programs address the professional and educational needs of inactive physicians, especially in anesthesiology \[[@B5]\]. In general, programs evaluate the re-entering or remediating physician through assessment for competence using multisource examinations \[[@B6]\]. The American Society of Anesthesiology lists two re-entry programs that provide education for the inactive physician seeking return to practice in New York and Philadelphia. Both the aforementioned programs provide re-education during observation alongside other learners; both provide varying program duration based on clinical time away. A new program in Texas (KSTAR/UTMB) has just started. We report on the refresher program at Drexel College of Medicine that facilitates anesthesiologists\' return to clinical practice. 2. Methods {#sec2} ========== 2.1. Program Overview {#sec2.1} --------------------- In 1968, the Medical College of Pennsylvania was among the first universities to initiate a re-entry program for nonpracticing physicians returning to clinical practice \[[@B7], [@B8]\]. Successful with over 400 trainees, the program temporarily ceased in 1993 when the hospital closed. In 2006, it was reinstituted and redesigned based on six core ACGME competencies to parallel current medical education, with adult learning principles. While the program maintains a basic structure of formative assessment, education, and then reassessment in a university-teaching environment with individual and group learning and a dedicated preceptor, the details are individualized (for the adult learner) based on the physicians\' goals. A steering committee was convened with assistant deans of education, including CME and assessment, to guide the redesign of the program. The committee meets regularly in an advisory role, including review of applicants. While the director oversees the entire Medicine Physician Refresher/Re-entry Program broadly, each department has dedicated preceptors to provide specialty-specific one-to-one education and mentoring. At the same time, the trainee learns with the faculty in their specialty-specific department. The program began with internal medicine and pediatrics; because of increasing demand, we developed the anesthesiology track in 2012. In addition to a clinical and didactic refresher, the individualized, structured program provides administrative support (through the continuing medical education (CME) department), career counseling, and postprogram follow-up to address requests from credentialing groups. Alumni, randomly and on their own volition, often stay in touch for years to share their accomplishments. Upon initial contact, the program director and returning anesthesiologist discuss relevant career background, anticipated future practice scope, reasons for leaving and returning, organizational requirements, and specific goals or expectations. Then, if the physician desires, the program director communicates with referring groups to further identify parameters for the physician\'s individualized re-education. All initial application review is by the director, steering committee, and then specialty-specific preceptor. The director and preceptor further conduct phone interviews with trainees to assure that goals can be met. Accepted trainees are enrolled in 6-week blocks (originally selected to match the medical school block duration) for 6, 12, or more weeks depending upon the individualized goals and recommendations. To optimize focused attention, in the anesthesiology department, one re-entry physician is accepted at any one time. The distinct outcome of this re-entry program is the refresher and update of skills are necessary to practice anesthesia as the American Board of Anesthesiology would deem fit. This includes sufficient knowledge base, application of knowledge to develop sound plan, and reinsertion into CME cycle requisites. Re-entering physicians, each with their own specific obstacles and needs seek a refresher; we ask them to identify their primary goals. Ascertaining said goals helps to further personalize the program, address each learner\'s individual needs, and fine-tune interactions with preceptors. We believe that going beyond the textbook and organizational needs to address the varied backgrounds of participants is crucial to fostering a constructive, successful environment. The program director, physician trainee, preceptor, and program coordinator develop an individualized curriculum. Several components of the program are mandatory; however, within the mandatory components, the individual cases and days can be adapted based on the physician\'s career goals. This process is informed by an initial needs assessment, during week 1, over a 2-day period, including Post Licensure Assessment System (PLAS) multiple-choice examinations, oral discussions of clinical scenarios, three standardized patient evaluations and their documentations. The information from the needs assessment is reviewed with the student over a two-day long feedback didactic session in a group setting. Concepts are reinforced, or new concepts introduces, as trainees learn from one another as well as the faculty. This allows trainees to contribute to each other\'s experiences and expand learning from everyone\'s topics and issues. The re-entry physician also completes a clinical interest survey to communicate their focus of interest for program scheduling. While trainees all follow the overall structure of the preceptorship, there is flexibility based on needs and goals within each component. For instance, all trainees receive formative feedback on their standardized patients\' evaluations. If a trainee needs more practice with communication skills versus medical content, then cases for practice are selected to meet their individual needs. Another example is in the self-selection of the anesthesia cases among all operating rooms, based on the physicians\' individual goals. Trainees keep a daily patient/case log to assure a wide range of clinical exposures that meet their specific future work goals. Trainees also write pre/postop notes for practice, on encountered patients in a word document (not in the patient\'s actual chart), which they review weekly with their preceptor for summative feedback. Preceptors together observe the trainee\'s recorded standardized patient scenarios (done as pretest, practice, and posttest) using a communication skill and medical content checklist to identify strengths and potentials for improvement. Updating computer, presentation, and peer teaching skills are other aspects of training. 2.2. Re-Entry Program Components {#sec2.2} -------------------------------- ### 2.2.1. Clinical Setting {#sec2.2.1} Similar to the program at Mt. Sinai, this program does not provide a hands-on experience for inactive returning anesthesiologists \[[@B9]\]. Our program seeks to gradually immerse the physician back into a clinical setting through didactics, discussions, observation in the OR alongside other learners for 8 hours/day, and standardized patient testing with feedback. These are all controlled exposures to clinical scenarios in an environment with the anesthesia team present, accounting for the limitation that inactive physicians do not often have an active license and thus are not able to obtain malpractice insurance \[[@B10]\]. Furthermore, similar to Mt. Sinai \[[@B9]\], we have found that pursuing relicensure, hospital credentialing, and current malpractice tremendously adds to the cost and time to start a re-entry program for physicians, rendering the re-entry program not feasible. In a tertiary academic center in an urban setting, physicians are exposed to an all-encompassing variety of real patient cases highly beneficial to a re-entering physician, who are seeking to update their skills and regain their confidence. Re-entering physicians choose from a myriad of available cases to suit their individual needs based on their self-assessed and practical future goals, similar to adult learning models. Candidates do not perform hands-on procedures but have direct interaction with the patients in the perioperative setting. Day-to-day discussions of comorbidity and subsequent anesthetic plans are reinforced by preceptors at the bedside. Re-entry physicians follow these patients into the operating room and witness their entire induction, maintenance, and emergence. They witness the assessment and recovery process in the postoperative care unit. Throughout this process, they are expected to actively interact, question, and learn from the perioperative team. We have learned that after applying didactic learning to such experiences, physicians are confident and academically up-to-date to enter a clinical setting where their employers often then provide an initial short supervised period. ### 2.2.2. Standardized Patient Evaluations {#sec2.2.2} Two to three times during each 6-week period (pretest, posttest, and practice based on the individual need), physicians evaluate standardized patients for formative and summative feedback. Cases have been written by Drexel physicians (subspecialist and primary care physician-educators, and reviewed by a committee of peers); there is access to a large variety of scenarios, both for communication and clinical content. The cases and checklists are standardized, identical to ones used for Drexel medical students, residents, and physician re-entry trainees. The level of the re-entry anesthesiologist is different from that of the other learners, but the content still proves useful. After the evaluations, as a group in an interactive setting where all contribute, learners, and faculty review the recorded evaluations, provide feedback, and are scored based on standardized checklists. The checklists measure interviewing skills as well as content medical knowledge. The content medical knowledge lists questions a physician should ask to elaborate a focused history of the specific chief complaint and elicit/narrow its differential diagnoses. The interviewing skills measure data gathering, interpersonal, information giving, and organizational skills as well as standardized patient satisfaction. As such, physicians first practice, learn, and improve in a safe and controlled environment. ### 2.2.3. Web-Based Modules {#sec2.2.3} To assure reentering physicians\' re-education parallels current up-to-date medical education, which may have been different from the physician\'s early medical education, we incorporate technology as a learning tool. In addition to in-person didactics and discussions in anesthesia conferences, the curriculum includes web-based interactive programs for focused learning and assessment on medical knowledge, communication skills, and clinical reasoning. Three distinct sources provide these online annotated video and virtual patients (DxR <https://dxrgroup.com/healthcare-education-products/dxr-clinician/>, [Doc.com](http://Doc.com)<https://doc.com/>, and Aquafer <https://www.aquifer.org/>). Each program has its own learning objectives. Participants are expected to complete 1-2 modules (self-selected based on goals) per week in each program and encouraged to redo modules as reinforcement, for practice in a virtual and safe setting. Didactic learning is assessed via multiple-choice or short-answer exam questions with feedback. Trainees report that these resources are invaluable for their learning and often voluntarily complete more modules than their weekly assignments. ### 2.2.4. Medical Documentation {#sec2.2.4} Trainees write medical notes regarding patients they had observed in the electronic health record (EHR) format to reinforce documentation skills, which they review through case logs with preceptor weekly. The templates are from the EHR, copied as pdf file where trainees may write their notes, but they are neither online nor part of the patient\'s permanent records; the documentations provide room for practice and learning. Trainees could participate in further EHR training with administrators, based on their individual needs. ### 2.2.5. Preceptorship {#sec2.2.5} Trainees participate in grand rounds, clinical cases, and conferences. Trainees learn within a multidisciplinary team with the entire anesthesiology staff (faculty, nurses, and residents) and self-select to work closely with faculty who share their interest. Each physician regularly meets one-on-one with a designated anesthesiology mentor for two hours for further individualized learning. These meetings include oral presentations, discussion of daily OR cases, review of notes, and summative performance evaluation. ### 2.2.6. Simulations {#sec2.2.6} Trainees participate in simulated scenarios in a multidisciplinary setting alongside nurses and other healthcare providers using the same simulations used to teach anesthesiology residents. These sessions include trained individuals as simulated patients to create realistic patient encounters, including but not limited too medical history obtain, consenting, conflict resolution, and ultrasound training. Robotic models that display changeable vital signs, can be "administered" medication, and be subjected to procedures which mimic acute scenarios. Sessions take place in the simulation center and are designed to benefit all learners involved; anesthesia trainees gain personal and interdisciplinary experience in this fashion. Learners are able to participate as the primary clinical provider and observe others. Each trainee has at least one simulation session during the re-entry period. These sessions are followed by debriefing sessions where the scenarios are discussed in detail for learning opportunities. ### 2.2.7. Computer Searches and Critical Appraisal Training {#sec2.2.7} During the orientation week, students meet with the science librarian for information technology skill sessions to learn evidence-based medicine research skills for continued self-improvement after the program. Trainees are expected to use these skills during their presentations; they are required to make one formal presentation on any self-chosen topic to their peers as well as volunteer for informal self-selected clinical presentations. ### 2.2.8. Advocacy and Support {#sec2.2.8} Most re-entry physicians confront emotional, personal, career, and financial barriers when returning to practice. The continuing medical education department provides hours of career and emotional support before, during, and after the program. We advocate on trainees\' behalf by writing letters, brainstorming career options, providing networking and employment introductions, and speaking to committees/employers. ### 2.2.9. Performance Evaluation {#sec2.2.9} Weekly evaluations monitor achieved learning goals. We regularly seek trainees\' written and verbal feedback, review gaps, and adjust schedules. Faculty provides summative feedback weekly on trainees\' clinical knowledge, communication skills, professionalism, history taking and documentation skills, and formative feedback using evaluation forms also used for medical students. Since all faculties participate in teaching medical students and residents, they use these evaluation forms regularly. Each trainee receives a minimum of 1 evaluation per week of the course. Upon completion, each trainee receives a detailed evaluation letter and certificate explaining their accomplishments and assessments; individual faculty feedback is listed, as well. The letters do not endorse competency for practice; rather they report on physicians\' accomplishments. 3. Results {#sec3} ========== Results reflect 6-year data from anesthesiology department re-entry physician trainees, from August 2012 to February 2018. Twelve physicians began and completed the anesthesiology re-entry program. Trainee demographics are in [Table 1](#tab1){ref-type="table"}. Each application was reviewed; none were rejected. Seven out of twelve were US medical graduates, five were international medical graduates, four were board-certified, and five had active licenses. Range of clinical inactivity was 0--10 years. Those who had no gaps had been immediately practicing in a global health setting but relocated to the US (one while in anesthesia residency in the United Kingdom, the other while practicing in South Africa). Even though they did not have a gap in their careers, each career setting provides its own needs and skills, and international physicians must complete US residencies. These physicians sought to update their skills before seeking residencies. Only one was practicing as an anesthesiologist; six were not working; the others worked in various capacities: insurance, high school teacher, postdoctoral student, and physician in another capacity (emergency, general practitioner, and chronic pain). Reasons for leaving and returning to clinical practice were: 1 relocation, 4 substance abuses, 1 family issue, 1 medical disability, 1 burnout, and 1 other. Those with substance abuse history had successfully completed programs with the Physicians\' Health Program. None had left practice for medical negligence. Each physician noted their specific goals after program completion ([Table 2](#tab2){ref-type="table"}). Nine of twelve trainees achieved their individual goal (75%); 7 (78%) achieved their goal within 1 year of course completion. Six pursued residency training. Five sought employment. One sought license renewal prior to seeking employment. Of these six who sought employment and five (82%) immediately began work. Four (67%) of six seeking residencies obtained spots (one of the two has not tried yet). Three did not achieve their goals; two were seeking residencies (one had left her residency to raise a family, another was IMG moving to US); and one just finished the program. 4. Discussion {#sec4} ============= Returning physicians bring unique backgrounds, needs, skills, and knowledge from their nonclinical experiences. Inactive physicians face difficulty returning to work for various reasons. Increasingly, they are asked to gain re-education. A re-entry program should provide flexible individualized re-education to meet physicians\' varying needs. Assessment is important for ongoing improvement, and a re-entry program is an effective way by which physicians can seek assessment and improve their professional skills \[[@B2], [@B10], [@B11]\]. This program meets the AMA Guiding Principles \[[@B3]\]; physicians seek the program from all geographic regions and note its national and international reputation. Although twelve physicians of differing backgrounds do not represent all returning physicians\' needs, our approach and practical experience alongside work done by other programs can provide an initial guide for the growing field of anesthesiology re-entry \[[@B9], [@B11]--[@B14]\]. We seek physician feedback for constant improvement; thus, we have learned important lessons. A multidisciplinary team of faculty, resident, and nursing support is critical to trainee re-education. Clinician-educators are best suited for re-entry physicians, as they are at the forefront of education. Re-education is best among a community of learners alongside fellows, residents, and students. The university provides undergraduate and graduate medical education; anesthesiology re-entry physicians learn alongside a variety of learners. This variety of knowledge distribution strengthens clinical skills and promotes the learning process. Interactions within this structure help to boost confidence and teaching skills. The re-entering anesthesiologist contributes to the learning of others by bringing nonclinical experiences. We explain choices available to trainees and obstacles they may face during or after the course. After hours of career and personal counseling, some may opt not to pursue re-entry, as with one of the trainees. Randomly, some trainees maintain close contact with us years later, reaching out for news or even guidance. We have expanded our career advisory roles, as trainees return for networking and advice long after program completion. We have an alumni association, where current trainees communicate (online or in person) with alumni for questions and support. Despite challenges of re-entry, participants are eager to regain proficiency and as evident in our program, the majority has succeeded in achieving their anticipated goals. The re-entry program has been a part of the College of Medicine since 1968, yet challenges still persist. While the re-entering physicians may learn in one department, resource utilization in other departments (CME, library sciences, and standardized patient program) requires timely coordination. While an academic setting is ideal for education and patient scenario variety, at the same time there are many learners (students, residents, and fellows) which necessitate clever and planned space coordination. Costs are significant in time and personnel due to the need for dedicated staff and faculty \[[@B11]\]. Two program coordinators help organize details: one is full-time in CME who coordinates all re-entry physicians\' schedules with their individual departments\' corresponding coordinator. This person has other duties too. Follow-up with re-entry physicians to ascertain goal completion in short term is achieved via direct contact with participants. These individuals often keep informal contact with the program to share their success and positivity. As years progress, it becomes difficult to maintain contact and ensure long-term achievement. Further program improvement includes developing dedicated means to contact destination sites to ascertain trainee status and provide long-term support. This paper reflects 12 physicians\' experiences; there is clear need for continued program evaluation. Our general re-entry program originally began in 1968; the anesthesiology track was developed in 2012. While this physician cohort, with representation from all regions and backgrounds, has had a 75% success in reaching their individual goals, we reflect short-term data upon program completion. Although our results are encouraging, there are more questions that need further investigation. Of specific note, areas of interest for future study could include which learning activities were of most use, if coaching support was beneficial, or if supervision upon reinstitution to practice helped support reintegration. For instance, it would be interesting to note if re-entering physicians more readily leave medicine again; or if they left for medical reasons, is relapse common; or do they enter advocacy fields to help colleagues maintain practice. Even as such, this anesthesiology physician re-entry program can provide a useful service to the community. 5. Lessons for Practice {#sec5} ======================= A multidisciplinary team of faculty, resident, and nursing support is critical to trainee re-education.Clinician-educators are best suited for re-entry physicians, as they are at the forefront of education, and this is best accomplished among a community of learners alongside fellows, residents, and students.The re-entering anesthesiologist contributes to the multidisciplinary variety of knowledge and strengthens clinical skill that accelerates the learning process along with a boost of confidence. We would like to acknowledge Adithya Joolukuntla, MD, for presenting this project in abstract form at The Anesthesiology Annual Meeting 2018. Data Availability ================= The data used in this case series are presented in the article. Data not presented in the manuscript are not to protect subject privacy. Ethical Approval ================ The study was approved by the Institutional Review Board of Drexel University \#1710005735. Conflicts of Interest ===================== The authors declare that they have no conflicts of interest. ###### Anesthesia trainee demographics. Anesthesia trainee demographics   --------------------------------- ----- Gender *n*  Male 7  Female 5 Location *n*  US graduate 7  International graduate 5 Board certification *n*  Maintained 4  Not maintained 8 Licensure *n*  Active 5  Not active 7 Reason for leaving *n*  Relocation 4  Substance abuse 4  Family issue 1  Medical disability 1  Burnout 1  Other 1 ###### Main goals and outcomes for trainees. Main goals and outcomes for trainees who completed the physician refresher/re-entry course       -------------------------------------------------------------------------------------------- ----- --------------------------------- ----- Goal *n* Outcome *n* License 1 License successfully reinstated 1 Employment 5 Refreshed and practicing 4 Residency 6 Residency training 4 \(i\) Decided not to seek residency 1 \(ii\) Just completed the program 1 [^1]: Academic Editor: Michael Frass

Significance StatementCurrent good manufacturing practice (cGMP) compliant human pluripotent stem cells (hPSC) have been recently established for clinical application. To validate the capability of such cGMP lines for liver therapy, this article describes their proficiency on advanced hepatocyte production using both conventional culture plate and three‐dimensional hydrogel systems. The subsequent study on hepatocyte transplantation into immune competent mice using alginate encapsulation model further demonstrates the suitability of cGMP hPSC‐derived hepatocytes for cell‐based therapy. To the authors\' knowledge, this is the first report demonstrating that cGMP‐compliant hPSCs can generate cells with advanced hepatic function potentially suitable for future therapeutic applications. Introduction {#sct312419-sec-0001} ============ For patient\'s presenting with end‐stage chronic liver disease or acute liver failure (ALF), orthotopic transplantation of a donor liver remains the only curative treatment. Due to lengthening waiting lists and a severe scarcity of donors, mortality rates are increasing annually [1](#sct312419-bib-0001){ref-type="ref"}. As a result, an alternative strategy for treating these patients is urgently required. Allogenic transplantation of primary adult hepatocytes, the major functional cell‐type of the liver, is considered a viable solution in certain clinical indications [2](#sct312419-bib-0002){ref-type="ref"}, [3](#sct312419-bib-0003){ref-type="ref"}, [4](#sct312419-bib-0004){ref-type="ref"}. Lack of sufficient numbers of high‐quality hepatocytes; a result of isolating cells from tissue deemed unsuitable for transplantation, has limited the success of this programme [5](#sct312419-bib-0005){ref-type="ref"}, [6](#sct312419-bib-0006){ref-type="ref"}. Derivation of human embryonic stem cells (hESCs) [7](#sct312419-bib-0007){ref-type="ref"} and their related induced pluripotent stem cells (hiPSCs) [8](#sct312419-bib-0008){ref-type="ref"}, [9](#sct312419-bib-0009){ref-type="ref"}, however, has generated growing optimism that the development of cellular therapies, such as would be suitable for liver disease, is finally an obtainable goal [10](#sct312419-bib-0010){ref-type="ref"}. Unlike primary hepatocytes, which cannot be cultured or expanded in vitro [11](#sct312419-bib-0011){ref-type="ref"}, [12](#sct312419-bib-0012){ref-type="ref"}, human pluripotent stem cells (hPSCs) possess an unlimited ability for self‐renewal [13](#sct312419-bib-0013){ref-type="ref"}. This capability to produce large batches of cells is of clinical significance as hepatocyte numbers approaching 8 billion may be required for transplant when correcting metabolic liver function in pediatric patients [14](#sct312419-bib-0014){ref-type="ref"}, or 15 billion to support liver failure in adults [15](#sct312419-bib-0015){ref-type="ref"}. Studies have shown that both hESCs [16](#sct312419-bib-0016){ref-type="ref"}, [17](#sct312419-bib-0017){ref-type="ref"} and hiPSCs [18](#sct312419-bib-0018){ref-type="ref"}, [19](#sct312419-bib-0019){ref-type="ref"}, [20](#sct312419-bib-0020){ref-type="ref"} can be differentiated into hepatocytes (hPSC‐Heps), sharing functional attributes of their in vivo equivalents, including albumin/α‐1 antitrypsin (A1AT) protein secretion, cytochrome P450 activity and glycogen storage. As research tools, hPSCs have delivered novel insights into human hepatic development [21](#sct312419-bib-0021){ref-type="ref"}, the creation of liver disease models [22](#sct312419-bib-0022){ref-type="ref"}, [23](#sct312419-bib-0023){ref-type="ref"}, and provided new platforms for pharmacological testing [24](#sct312419-bib-0024){ref-type="ref"}. Furthermore, the successful, albeit limited, repopulation of rodent livers following transplantation of both hESC, and hiPSC‐derived hepatocytes has been reported by several groups [25](#sct312419-bib-0025){ref-type="ref"}, [26](#sct312419-bib-0026){ref-type="ref"}, [27](#sct312419-bib-0027){ref-type="ref"}, suggesting hPSC‐Heps may be a viable treatment option for patients with liver disease. The stem cell field has developed at an exceptional rate, which has resulted in the first human trials using cells derived from hESCs/hiPSCs being undertaken [28](#sct312419-bib-0028){ref-type="ref"}, [29](#sct312419-bib-0029){ref-type="ref"}, [30](#sct312419-bib-0030){ref-type="ref"}, [31](#sct312419-bib-0031){ref-type="ref"}, [32](#sct312419-bib-0032){ref-type="ref"}. Although approved for clinical use, these lines were in fact developed for research purposes and not produced under current good manufacturing practice (cGMP) guidelines. These manufacturing regulations for stem cell therapy products are described by the Food and Drug Administration in the U.S., and since the 2004 European Union Tissues and Cells Directive, the European Medicines Agency within the European Union. Generating cells under cGMP conditions ensures their clinical safety, and for cellular therapies should be differentiated in fully defined, xeno‐free conditions [33](#sct312419-bib-0033){ref-type="ref"} to ensure reproducibility, and prevent xeno‐mediated infection or immune rejection [34](#sct312419-bib-0034){ref-type="ref"}. In 2011, scientists from King\'s College London submitted the first xeno‐free clinical grade hESCs [35](#sct312419-bib-0035){ref-type="ref"} to the U.K. Stem Cell Bank [36](#sct312419-bib-0036){ref-type="ref"}. More recently, cGMP‐compliant hiPSCs have been generated by teams in People\'s Republic of China [37](#sct312419-bib-0037){ref-type="ref"}, Japan [38](#sct312419-bib-0038){ref-type="ref"}, the U.S. [39](#sct312419-bib-0039){ref-type="ref"} and the U.K. [40](#sct312419-bib-0040){ref-type="ref"}. These clinically compliant lines have been extensively characterized as pluripotent with profiles comparable to previously validated hPSCs derived outside of these manufacturing guidelines [41](#sct312419-bib-0041){ref-type="ref"}. However, given previous reports describing the varied potential for hPSC lines to differentiate into target cell types [42](#sct312419-bib-0042){ref-type="ref"}, [43](#sct312419-bib-0043){ref-type="ref"}, [44](#sct312419-bib-0044){ref-type="ref"}, evaluation of their hepatic differentiation potential represents an essential prerequisite to further clinical development work. The objective of this study was therefore to undertake this evaluation and then test the potential of differentiated cells to be engineered into three‐dimensional (3D) constructs suitable for clinical application. Materials and Methods {#sct312419-sec-0002} ===================== Cell Lines and Cell Culture {#sct312419-sec-0003} --------------------------- Two cGMP hiPSC lines (CGT‐RCiB‐10 \[line 1; Cell & Gene Therapy Catapult, London, U.K.\] and LiPSC‐GR1.1 \[line 2; Lonza, Walkersville, MD\]) and one cGMP hESC line (KCL037 \[line 3; Gifted from D. Ilic, King\'s College London\]) were used in this study. All three lines were recovered into the culture conditions recommended by their respective suppliers. After two passages, each of the lines was subsequently maintained on Vitronectin XF (STEMCELL Technologies, Vancouver, BC, Canada) coated Corning Costar TC‐treated 6‐well plates (Sigma--Aldrich, St. Louis, MO) in TeSR‐E8 (STEMCELL Technologies, Vancouver, BC, Canada) and passaged every 4 days using Gentle Cell Dissociation Reagent (STEMCELL Technologies). Line 3 was passaged in TeSR‐8 supplemented with 10 μM Y‐27632 dihydrochloride (R&D Systems, Minneapolis, MN) to ensure cell survival. Hepatocyte differentiation was carried out in Essential 6 Medium (Thermo Fisher Scientific, Waltham, MA; days 1--2), RPMI‐1640 Medium (Sigma--Aldrich; days 3--8) and HepatoZYME‐SFM (Thermo Fisher Scientific; day 9 onward) within Corning Falcon 100 × 20 mm style tissue culture dishes (Sigma--Aldrich). The following growth factors and small molecules were supplemented into the media for hepatocyte differentiation as shown in Figure [1](#sct312419-fig-0001){ref-type="fig"}A: 3 μM CHIR9901 (Sigma--Aldrich), 10 ng/ml BMP4 (R&D Systems), 10 μM LY29004 (Promega, Madison, WI), 80 ng/ml FGF2 (R&D Systems), 100 ng/ml and 50 ng/ml day 4 onward Activin A (Qkine, Cambridge, U.K.), 10 ng/ml OSM (R&D Systems) and 50 ng/ml HGF (PeproTech, Rocky Hill, NJ). Day 21 hPSC‐Heps were dissociated into a single‐cell suspension using TrypLE Express Enzyme (1×), no phenol red (Thermo Fisher Scientific). {#sct312419-fig-0001} Brightfield and Immunofluorescence Imaging {#sct312419-sec-0004} ------------------------------------------ Brightfield microscopy was performed on a Leica DMIL LED inverted microscope and imaged using the Leica DFC3000 G camera (Leica Microsystems, Wetzlar, Germany). For immunofluorescence staining, samples were fixed for 10 minutes with 4% wt/vol paraformaldehyde, and then blocked and permeabilized in 1% wt/vol bovine serum albumin (Sigma--Aldrich), 3% donkey serum (Thermo Fisher Scientific) and 0.1% Triton X‐100 (Sigma--Aldrich). An additional 10 minutes of permeabilisation was performed using 0.5% Triton X‐100 for detection of nuclear antigens. Primary antibodies were applied for 1 hour and after wash steps Alexa Fluor‐555/488 conjugated secondary antibodies (Thermo Fisher Scientific) were incubated for 40 minutes. NucBlue Fixed Cell ReadyProdes Reagent (Thermo Fisher Scientific) was applied for visualization of cell nuclei. Imaging for two‐dimensional (2D) culture was performed on an Operetta High Content Screening System (PerkinElmer, Waltham, MA), and for 3D culture, a Nikon Eclipse Ti inverted spinning disk confocal microscope (Nikon, Minto, Japan). Real‐Time PCR {#sct312419-sec-0005} ------------- Total RNA was isolated using the RNeasy Mini Kit (QIAGEN, Hilden, Germany) according to manufacturer\'s protocol. RNA was quantified spectrophotometrically using the NanoDrop 2000 (Thermo Fisher Scientific). Three hundred and fifty nanograms of total RNA was used to produce first‐strand cDNA using the SuperScript VILO cDNA synthesis kit (Thermo Fisher Scientific). Quantitative real‐time PCR (RT‐PCR) was performed in a 10 μl reaction mixture consisting of cDNA, custom designed oligonucleotide primers (Sigma--Aldrich) and Fast SYBR Green PCR Master Mix (Thermo Fisher Scientific), on a CFX384 Touch Real‐Time PCR Detection System (Bio‐Rad, Hercules, CA). ACTB mRNA was used for housekeeping normalization. Flow Cytometric Analysis {#sct312419-sec-0006} ------------------------ Adherent cells were dissociated into a single‐cell suspension using TrypLE Express Enzyme (1×), no phenol red, and subsequently treated for 30 minutes with LIVE/DEAD Fixable Cell Stain (Thermo Fisher Scientific) and then fixed using 4% wt/vol paraformaladehyde (PFA) for 10 minutes. Cells were incubated with fluorophore‐conjugated antibodies for 30 minutes in the dark, and then washed twice with phosphate buffered solution (PBS). Immunophenotyping was carried out using the BD FACSCanto II system (Becton Dickinson, Franklin Lakes, NJ) and analyzed using FlowJo software (Becton Dickinson). Assessment of Hepatic Function {#sct312419-sec-0007} ------------------------------ Albumin production of hPSC‐Heps was measured using the Human Albumin Quantification Set (Bethyl Laboratories, Montgomery, AL). Culture medium supernatants were collected after 48 hours and stored at −20°C. Enzyme‐linked immunosorbent assay (ELISA) was carried out according to manufacturer\'s instructions. Absorbance was measured at 450 nm on a Promega GloMax Multi+ Detection System plate reader (Promega). Native cytochrome P450 *CYP3A4* activity was assessed using the *CYP3A4* P450‐Glo Assay with Luciferin‐IPA (Promega). The bioluminescent substrate was incubated on hPSC‐Heps for 1 hour before being collected for analysis. Luminescence was measured using a Promega GloMax Discover multimode microplate reader (Promega). Fabrication of ICC PEG‐DA Scaffolds {#sct312419-sec-0008} ----------------------------------- Thermo Scientific 4,000 Series monosized polystyrene beads of 100 ± 1.5 μm diameter (Thermo Fisher Scientific) were suspended in 70% EtOH and agitated using an ultrasonic bath. The dispersed bead suspension was seeded into hexagonal polypropylene molds and left to dry overnight on an orbital shaker. A self‐standing colloidal crystal lattice was produced through annealing the beads at 120°C for 4 hours. Poly(ethylene glycol)‐diacrylate (PEG‐DA; Thermo Fisher Scientific) acrylate‐PEGN‐hydroxysuccinimide (Laysan Bio Inc., Arab) and Irgacure 2,959 photoinitiator (BASF, London, U.K.) were mixed together in dH~2~0 at a concentration of 50%, 10%, and 1% wt/vol, respectively. The bead lattices were placed within this precursor solution, and centrifugation (500*g*, 5 minutes) was performed to ensure complete infiltration. Hydrogel fabrication was completed through UV light induced gelation of the precursor solution and the polystyrene crystal lattice was removed from the scaffold through tetrahydrofuran (Sigma--Aldrich) soaking for 4 hours. Generation of hPSC‐Derived Hepatocyte Spheroids {#sct312419-sec-0009} ----------------------------------------------- A single cell suspension of day 21 hPSC‐Heps was prepared and 0.3 × 10^6^ cells were seeded per well of a 24‐well Aggrewell‐400 (STEMCELL Technologies). Aggrewell plates were prepared as recommended by the supplier. Centrifugation at 200*g* for 3 minutes was carried out to deposit cells into the microwells of the plate. Alginate Encapsulation of hPSC‐Derived Hepatocyte Spheroids {#sct312419-sec-0010} ----------------------------------------------------------- Encapsulation was performed as previously published [45](#sct312419-bib-0045){ref-type="ref"}, [46](#sct312419-bib-0046){ref-type="ref"}. In brief, spheroids were washed in saline before being resuspended into a final 1.8% ultra‐pure low‐viscosity, high‐glucuronic acid (≥60%), sodium alginate (FMC BioPolymer, Drammen, Norway) solution, which was then delivered by syringe pump through a 0.2 mm diameter nozzle, from which droplets were electrostatically deposited into a divalent cationic solution (1 mM BaCl~2~ + 50 mM CaCl~2~) to cause gelation. Live/Dead Staining {#sct312419-sec-0011} ------------------ Fluorescine diacete (FDA; Sigma--Aldrich) and cell‐impermeant ethidium homodimer‐1 (EthD‐1; Thermo Fisher Scientific) were used as recommended by the supplier for staining of viable and dead cells. Spheroids and alginate encapsulated cells were incubated in 4 μM EthD‐1 for 35 minutes, washed with Hank\'s Balanced Salt Solution (HBSS) containing calcium (Thermo Fisher Scientific), then incubated in 50 μg/ml FDA for 90 seconds, and finally washed five times with HBSS before imaging on a Leica TCS SP8 Confocal laser scanning microscope (Leica Microsystems, Wetzlar, Germany). Transplantation of hPSC‐Derived Hepatocyte Spheroids {#sct312419-sec-0012} ---------------------------------------------------- Alginate microencapsulated hepatocyte spheroids were intraperitoneally xenotransplanted into immune competent (C57BL/6 and Crl:CD1 \[CD‐1\]) and immune deficient (Rag2γ) mice. Spheroids were cultured in vitro for 3 days (CD‐1) or 7 days (C57BL/6 and Rag2γ) prior to encapsulation, and incubated within RPMI‐1640 medium for 2 hours before transplantation. Empty cell‐free microspheres were transplanted as a control. Surgical procedures were carried out under isoflurane anesthesia (1%--5% isoflurane, 95% oxygen, 1 l/min), with 30 μg/kg buprenorphine being administered immediately postsurgery. To create a sterile site of surgery, the mouse abdomen was shaved and cleaned with both antiseptic iodopovidone and isopropyl alcohol. A small incision through the skin, and a subsequent through the linea alba of the peritoneum allowed saline suspended alginate microspheres, containing approximately 2 × 10^3^ hepatocyte spheroids, to be delivered into the peritoneal cavity using a sterile pipette. Recovery of hPSC‐Derived Hepatocyte Spheroid Containing Microspheres {#sct312419-sec-0013} -------------------------------------------------------------------- The mice were sacrificed by subcutaneous pentobarbital euthanasia 72 hours after transplantation. Blood samples were collected through cardiac puncture, and serum was diluted 1:10 for the detection of human albumin by ELISA. Injection of 5 ml saline into the peritoneal cavity was performed so that microspheres could be collected by peritoneal lavage. Microspheres were washed in saline and then maintained on ice, in RPMI‐1640 medium, until further analyses could be performed. Immunohistochemical Staining {#sct312419-sec-0014} ---------------------------- Recovered microspheres were first fixed with 4% paraformaldehyde for 15 minutes, washed four times using PBS and transferred into 70% ethanol. The dehydrated samples were then paraffin infiltrated using Excelsior AS Tissue Processor (Thermo Fisher Scientific) and paraffin embedded using HistoStar Embedding Workstation (Thermo Fisher Scientific). Five micrometres thickness slides were then sectioned ready for immunohistochemical staining with a mouse and rabbit specific horseradish peroxidase/3‐amino‐9‐ethylcarbazole (HRP/AEC) detection immunohistochemistry (IHC) kit (Abcam, Cambridge, U.K.). Results {#sct312419-sec-0015} ======= We firstly recovered two lines of hiPSCs, as well as one line of hESCs, each of which having been derived independently using cGMP‐compliant protocols. We maintained all lines in identical culture conditions comprising of xeno‐free cell culture matrix, Vitronectin, and chemically defined pluripotency culture medium, TeSR‐E8. After several passages within these culture conditions, each of the lines had fully reconditioned, with comparable cell morphologies and colony sizes (Supporting Information [S1](#sct312419-supitem-0001){ref-type="supplementary-material"}); each line producing characteristic rounded colonies, with small densely packed cells. We then began investigating the capability of each line in producing high quality hPSC‐Heps using an adapted four‐stage hepatocyte differentiation protocol based on Hannan et al. [47](#sct312419-bib-0047){ref-type="ref"}, and their potential application in downstream clinical applications by achieving advanced phenotypes in macroporous hydrogels and ensuring viability within transplantation ready cell encapsulation models for ALF therapy (Fig. [1](#sct312419-fig-0001){ref-type="fig"}). Differentiation of cGMP‐Compliant Stem Cells Toward hPSC‐Derived Hepatocytes {#sct312419-sec-0016} ---------------------------------------------------------------------------- To prepare for hepatic differentiation, we seeded fragmented hPSC colonies obtained through enzyme‐free dissociation with minimal trituration onto gelatin coated tissue culture dishes. We treated hESCs with the small molecule Y‐27632 dihydrochloride, an inhibitor of the RHO/ROCK pathway, overnight to ensure sufficient cell attachment onto the gelatin. We found that the hiPSC lines did not require the addition of this small molecule to be passaged, or adhere to gelatin. We then initiated hepatic differentiation 2 days postseeding; having allowed enough time for the hPSCs to reestablish rounded colonies, with the outer cells tending to have a larger, more spread morphology. We closely monitored the morphology of the cells throughout the differentiation and cross‐compared against that of our previous publication, that used non‐cGMP‐compliant hiPSCs, to ensure appropriate differentiation was achieved (Fig. [2](#sct312419-fig-0002){ref-type="fig"}A). Upon differentiation, the clear borders signifying stem cell colonies dissipate gradually with the peripheral cells starting to spread and migrate out sporadically by day 1 postdifferentiation. These protruding cells expand in size and proliferate to close the space between neighboring colonies until a confluent monolayer is formed by day 4. At this stage, cells exhibit definitive endoderm (DE)‐like morphology which persists until the media condition is changed to contain oncostatin‐M and hepatocyte growth factor at day 9. After this alteration, the morphology becomes more dynamic as the cells continue to differentiate. By day 14 the cells start to transform from their elongated morphology, observed at day 11, into a more cuboidal shape. The signature, well defined, polyhedral morphology of hepatocytes is observed across the whole culture by day 17. As demonstrated by the dynamic morphological transformation throughout the course of differentiation, all three of cGMP‐compliant lines appear capable of generating populations of DE, hepatic endoderm, and subsequently hPSC‐Heps (Fig. [2](#sct312419-fig-0002){ref-type="fig"}B). {#sct312419-fig-0002} Characterization of cGMP‐Compliant hPSC‐Derived Hepatocytes {#sct312419-sec-0017} ----------------------------------------------------------- We next validated that the cGMP‐hPSCs were differentiating through the correct developmental lineage trajectory by collecting the cells and quantifying their mRNA expression at four distinct stages of the protocol (Fig. [3](#sct312419-fig-0003){ref-type="fig"}A). These time‐points represent undifferentiated hPSCs, DE (day 7), hepatic endoderm (HE, day 14) and finally hPSC‐Heps (day 21). {#sct312419-fig-0003} Importantly the expression of *NANOG*, a key pluripotency gene, is significantly downregulated upon differentiation, and is obsolete at day 21, the stage in which hPSC‐Heps have been produced. Furthermore, the expression of both *CXCR4* and *SOX17*, well established markers for both DE and primitive endoderm, are found to peak and be significantly upregulated at day 7 of differentiation when compared with undifferentiated hPSCs; after this time‐point the expression of both genes dissipates. To assess the correct differentiation into hPSC‐Heps, we selected six major hepatic genes for investigation. By day 21 a significant increase in the relative expression of *AFP*, can be measured. *AFP* is the gene encoding α‐fetoprotein, a major plasma protein produced by the developing liver and considered to be the fetal version of albumin [48](#sct312419-bib-0048){ref-type="ref"}. Incidentally, expression of *ALB*, albeit not significant, is detected at days 14 and 21 of differentiation, with the expression being greatest at the later time‐point. This provides evidence that at this stage the cell type produced is one that resembles a maturing hepatocyte. In addition, the relative gene expression for *HNF4A*, a hepatic transcription factor, as well as *ASGR2* (day 21 only), which encodes an asialoglycoprotein receptor isoform primarily found on liver cells, are significantly higher than that of undifferentiated or DE cells. In the case of *HNF4A*, the relative gene expression is peaked at day 14, indicative that the hepatocyte‐fate determination has occurred. Moreover, the expression for *SERPINF2*, the gene encoding the serpin α‐2 antiplasmin which is secreted in plasma by hepatocytes, is significantly elevated after 21 days of differentiation. Finally, nonsignificant, but elevated expression of *CYP3A7* is detected in day 21 hPSC‐Heps than the earlier time‐points. This cytochrome P450 3A family isoform is predominately expressed in the developing liver, with the translated enzyme involved in the metabolism of drugs, together with the synthesis of cholesterol, and various other lipids. After validating the hepatic specific morphology and gene expression, we next sought to characterize the population profiles of the cells generated using our adapted protocol by performing flow cytometry analysis on day 21 hPSC‐Heps (Fig. [3](#sct312419-fig-0003){ref-type="fig"}B). The hepatic progenitor markers EpCAM (98.1%) and cytokeratin‐19 (58.8%) were expressed on most cells analyzed. In addition to these hepatic progenitor markers, expression of the asialoglycoprotein receptor 1 (ASGPR1), an endocytotic cell surface receptor specific to adult hepatocytes, was detected on 41.3% of day 21 hPSC‐Heps. The presence of hepatocyte progenitor markers and nonsignificant gene expression of *ALB* and *CYP3A7* displays the need for further maturation culture to produce a cell more closely resembling an adult hepatocyte. 2D Maturation of hPSC‐Heps {#sct312419-sec-0018} -------------------------- Having validated that cGMP‐compliant hPSCs were able to generate cells resembling immature hepatocytes when cultured in our differentiation conditions, we next aimed to advance their hepatic maturation through culturing the cells within different culture model systems to challenge the clinical relevance of our protocols. We first seeded the day 21 hPSC‐Heps onto collagen‐1 coated tissue culture plastic, as it is an extracellular protein that has been shown capable of supporting the long‐term culture, and liver‐specific functions, of isolated adult hepatocytes [49](#sct312419-bib-0049){ref-type="ref"}. Upon seeding, the hPSC‐Heps recovered their polyhedral morphology within 2 days. It should be noted that if not seeded to confluence, then many cells do not remain viable in 2D culture on collagen‐1, and those remaining do not proliferate, or go on to develop a mature phenotype. After 3 weeks of maturation culture on collagen‐1 coated tissue culture plastic, we performed immunofluorescent staining to assess the hepatic maturity and heterogeneity of the hPSC‐Heps (Fig. [4](#sct312419-fig-0004){ref-type="fig"}A). Firstly, none of the pluripotency or endodermal markers, such as OCT4 and CXCR4 are detected in the cultures, which is a good indicator that the conversion of stem cell to hepatic lineage cell was completed in our protocol. It is encouraging to observe hepatocyte‐specific markers, for example the nuclear transcription factor, HNF4α and protease inhibitor, A1AT, to be abundantly expressed throughout the cultures. To further delineate the maturity of hPSC‐Heps cultured on collagen‐I coated 2D surfaces, we assessed the presence of hepatoblast specification (*AFP*, KRT19, and EpCAM) and hepatocyte specification (*ALB* and ASGPR1) markers, respectively. We found that even though the expression of mature markers, such as *ALB* and ASGPR1, are prevalent across the culture, we still observed substantial regions of cells that expressed *AFP*, KRT19, and EpCAM. The perseverance of these progenitor markers reveals a potential limitation of 2D culture in terms of differentiating a fully adult‐like hepatocyte. Of note, zona occluding 2 (ZO‐2), a component of tight junction proteins was highly expressed in 2D culture, highlighting the polyhedral morphology of the cells. ![Characterization of human pluripotent stem cells (hPSC)‐derived hepatocyte maturation in two‐dimensional (2D) model. **(A):** Immunofluorescent images revealing the transition of pluripotency (OCT4), endodermal (CXCR4), and hepatic specification (*HNF4A*, A1AT, *AFP*, KRT19, and EPCAM) and mature hepatic (*ALB*, ASGPR1, and ZO‐2) expression in hPSC‐Heps after 20 days postseeding on 2D model. Representative images selected from each of the three lines. Scale bar: 100 μm. **(B):** Differential gene expression showing the relative expression of four key hepatic genes (*AFP*, *ALB*, *CYP3A7*, and *CYP3A4*) in preseeding (pre) and 20 days postseeding (20 dps) into 2D model. Statistical significance determined by Student\'s *t* test (two‐tailed); *n* = 3 experiments. Data shown for cell line 1. Data are mean ± SEM, \*, *p* \< .05; \*\*, *p* \< .005; ns: nonsignificant. **(C):** Albumin production rate of hPSC‐derived hepatic endoderm (HE), and hPSC‐Heps 8 and 14 days postseeding (8 and 14 dps) into 2D model; n.d.: not detected; *n* = 6 experiments per cell line. Data are mean ± SD, \*\*\*\*, *p* \< .0001. **(D):** Cytochrome P450 3A4 enzyme activity of hPSC‐derived HE, and hPSC‐Heps 8 and 20 days postseeding into 2D maturation culture, n.d.: not detected; *n* = 3 (mean luminescence value \[*n* = 6\] of 3 independent experiments). Data are mean ± SD, \*\*\*\*, *p* \< .0001.](SCT3-8-124-g004){#sct312419-fig-0004} For a broader evaluation of the maturation of hPSC‐Heps in 2D, we compared the mRNA expression of cells 20 days postseeding to that of the preseeding population (Fig. [4](#sct312419-fig-0004){ref-type="fig"}B). A significant reduction in the level of *AFP* is achieved, while conversely, a significant enhancement in *ALB* expression occurs. A similar significant 15‐fold increase is found in the relative expression of *CYP3A7* after the additional 20 days of culture. Interestingly, although enzyme activity was measurable in the previous assay, the gene expression of *CYP3A4* is elevated, but not significantly higher than that of the preseeding population. These mRNA expression results corroborate with our immunostaining observation that our cells are yet to achieve advanced maturity in this 2D model system. To characterize the liver specific functions of the cells, we monitored the albumin production rate (Fig. [4](#sct312419-fig-0004){ref-type="fig"}C and Supporting Information [S2](#sct312419-supitem-0001){ref-type="supplementary-material"}) and the metabolic activities of the cells over an extended period (Fig. [4](#sct312419-fig-0004){ref-type="fig"}D). As a 2D monolayer, hPSC‐Heps secreted albumin at levels detectable by ELISA after 6 days postseeding. Once cultured for a further week, the amount of albumin protein measured was significantly higher and an indicator of a maturing phenotype. Moreover, the albumin production is comparable between the three lines assessed, with no significant differences measured after 8 or 14 days of maturation. Furthermore, when cultured for an additional week (20 days postseeding), noninduced enzyme activity of cytochrome P450 3A4 can be detected (Fig. [4](#sct312419-fig-0004){ref-type="fig"}D). For each of the three lines, the enzyme activity detected is significantly greater at 20 days postseeding when compared with that of those after just 8 days. Again, no significant difference between the different cGMP‐compliant lines was measured at either time‐point, nor was activity detected in hPSC‐derived cells prior to 2D maturation culture. The presence of *CYP3A4* activity, which is initially absent from the liver of new‐borns, and responsible for \>50% of medicinal drug metabolism, is a clear indicator that the extended culture period results in matured hepatocytes [50](#sct312419-bib-0050){ref-type="ref"}, [51](#sct312419-bib-0051){ref-type="ref"}. 3D Maturation of hPSC‐Heps Within a PEG‐DA‐Based Scaffold Suitable for Biomedical Applications {#sct312419-sec-0019} ---------------------------------------------------------------------------------------------- Having achieved considerable maturation on 2D collagen‐1 coated tissue culture plastic, we proceeded to load day 21 hPSC‐Heps into a more physiologically relevant model system for advanced hepatic maturation. We used a 3D macroporous PEG‐DA hydrogel, known as an inverse colloidal crystal (ICC) scaffold, that aims to mimic the anatomy of native liver tissue. The ICC scaffold has uniform sized pores, interconnected in a hexagonal pattern, and we have previously demonstrated the generation of liver organoids using both primary human fetal liver cells [52](#sct312419-bib-0052){ref-type="ref"} and non‐cGMP‐compliant hPSCs [53](#sct312419-bib-0053){ref-type="ref"} within this model. To facilitate the characterization of the morphogenic transformation of hPSC‐Heps within this scaffold, we performed a series of immunofluorescence staining and constructed the 3D images using confocal microscopy. Firstly, we observed that upon seeding into the ICC hydrogel, hPSC‐Heps establish cell‐matrix interactions with the coated ECM protein and achieve confluence over the concave surface of the internal pores of the scaffold within 3--5 days postseeding (Fig. [5](#sct312419-fig-0005){ref-type="fig"}A). Within 7 days postseeding, the cells self‐assemble into mechanically stable interconnected clusters that resemble organoid structures for up to at least 3 weeks in culture. These distinct phases of morphogenesis were further illustrated with immunofluorescence imaging of beta‐catenin (CTNNB1) and keratin 18 (KRT18) staining. Keratin 18 is the major intermediate filament protein in the liver, with a role in regulating glucose metabolism and modulating insulin signaling [54](#sct312419-bib-0054){ref-type="ref"}, whereas beta‐catenin is the central constituent of canonical Wnt signaling and is implemented as a fundamental regulator in hepatic physiology and development [55](#sct312419-bib-0055){ref-type="ref"}. {#sct312419-fig-0005} To further evaluate the phenotype of hPSC‐Heps within this ICC culture model, we selected additional hepatic specific markers, along with markers of hepatocyte polarity, for broader evaluation by immunofluorescence imaging. Closer inspection on confocal micrograph revealed that the self‐assembled interconnected clusters inside the ICC scaffold consisted of a heterogeneous population, with *AFP* positive liver progenitor cells occupying the periphery of the cluster and surrounding the *ALB* positive mature hepatocytes at the core (Fig. [5](#sct312419-fig-0005){ref-type="fig"}B). As most cells of the organoid‐like clusters stained positive for cytoplasmic albumin, amid only few positive for *AFP*, this staining served as confirmation that the seeded hPSC‐Heps go on to develop a more mature phenotype. Moreover, correctly localized expression of the tight junction proteins ZO‐1, ZO‐2, occludin and claudin‐1, bile‐salt efflux pump (BSEP) and dipeptidyl peptidase‐4 (CD26) were observed. These proteins are key components of bile canaliculi that form between the lateral faces of hepatocytes, and merge into bile ductules. The performed immunofluorescence staining provides validation of the liver‐specific signature of the organoid‐like structures generated. We then performed a series of characterization to interrogate whether seeding hPSC‐Heps into this 3D environment resulted in the acquisition of greater maturity compared with the 2D culture model. RT‐PCR revealed that the expression of mature hepatic genes for proteins involved in biosynthesis (*ALB*), glycoprotein homeostasis (*ASGR2*) and metabolic functions (*CYP3A7* and *CYP3A4*) were significantly upregulated in the cells cultured within ICC scaffolds compared with those in 2D (Fig. [5](#sct312419-fig-0005){ref-type="fig"}C). Notably, the expression of fetal hepatocyte associated *AFP* was significantly higher in the 3D culture, however, this level is still significantly lower than in the day 21 cells initially seeded into the scaffold (Supporting Information [S3](#sct312419-supitem-0001){ref-type="supplementary-material"}). Likewise, functional assays for albumin secretion (Fig. [5](#sct312419-fig-0005){ref-type="fig"}D) and *CYP3A4* enzyme activity (Fig. [5](#sct312419-fig-0005){ref-type="fig"}E) showed significant improvements for hPSC‐Heps matured within the ICC scaffold when compared with their 2D counterparts. The attainment and enhancement of these cellular functions are both major features of advanced hepatic differentiation, and their presence would be an unconditional necessity for any potential stem cell‐derived therapy. Cumulatively, these data confirm that cGMP‐compliant hPSC‐derived hepatocytes can successfully be matured into a functional hepatic phenotype within a 3D, readily up‐scalable, system. This combination of cGMP‐compliant cells and a biocompatible scaffold not only provides a platform conducive to the further study of hPSC‐Heps, but could also hold potential for future drug development and safety studies, and for assisting as a vehicle for cell transplantation. Generation of Alginate Encapsulated Hepatocyte Spheroids {#sct312419-sec-0020} -------------------------------------------------------- To assess the broader translational potential of cGMP‐compliant hPSC‐derived hepatocytes for cell‐based therapies aimed at ALF, we carried out microencapsulation of hPSC‐Heps within alginate; a methodology that allows transplanted cells to be isolated from the recipient\'s immune responses [56](#sct312419-bib-0056){ref-type="ref"}. Successful encapsulation of hepatocytes within alginate hydrogels has been reported [57](#sct312419-bib-0057){ref-type="ref"}. However, to reduce the possibility of cell death during the encapsulation process [58](#sct312419-bib-0058){ref-type="ref"}, we used an additional established culture model, spheroid culture [59](#sct312419-bib-0059){ref-type="ref"}; known to both prolong viability [60](#sct312419-bib-0060){ref-type="ref"} and phenotypes of hPSC‐Heps [61](#sct312419-bib-0061){ref-type="ref"} (Supporting Information [S4](#sct312419-supitem-0001){ref-type="supplementary-material"}). To generate spheroids for encapsulation (Fig. [6](#sct312419-fig-0006){ref-type="fig"}A), we used nonadherent microwell containing plates, called AggreWell, which have previously been used to generate spheroids from hiPSCs [62](#sct312419-bib-0062){ref-type="ref"}. We centrifuged single cell suspensions of hPSC‐Heps into microwells to facilitate cell--cell interactions for spheroid formation in large readily scalable quantities (Fig. [6](#sct312419-fig-0006){ref-type="fig"}B). The spheroids were left in culture for 7 days prior to encapsulation within 1.8% alginate microspheres. The 3D aggregates maintain their structure and uniformity as they are pumped through the microcapsule generator\'s 0.22 mm diameter nozzle at a rate of 10 ml/hour. The crosslinking of alginate occurs in under 5 minutes within barium chloride/calcium chloride solution, with most microspheres containing a single spheroid. We then washed the collected microspheres in saline and subsequently placed back into HepatoZYME‐SFM. To ensure that the viability of these hPSC‐Heps was preserved throughout this process, we carried out live/dead immunofluorescence staining using FDA and ethidium homodimer‐1 (EthD‐1) 6 hours after the encapsulation (Fig. [6](#sct312419-fig-0006){ref-type="fig"}C). Confocal microscopy confirmed that both alginate‐ and noncapsuled spheroids contain minimal to no dead cells (EthD‐1 stained nuclei), with close to all cells having hydrolysed FDA into fluorescent fluorescein. Furthermore, encapsulation does not impact the viability of hepatocyte spheroids placed back into further culture (Supporting Information [S5](#sct312419-supitem-0001){ref-type="supplementary-material"}). {#sct312419-fig-0006} To test the clinical suitability of alginate encapsulated hPSC‐Heps as a bridging therapy for ALF, we xenotransplanted spheroid containing microspheres into the peritoneal cavity of immune competent C57BL/6 and CD‐1 mice, as well as immune deficient Rag2γ mice. We monitored transplanted mice over a period of 12 days, and no signs of postoperative complication or mortality were observed. Blood sera collected from mice 3 days postprocedure revealed that the cells remain functional upon transplantation as the presence of human albumin (Fig. [6](#sct312419-fig-0006){ref-type="fig"}D) was detected by ELISA in all animals having received microspheres containing cGMP‐hPSC‐derived hepatocyte spheroids. Additionally, we recovered transplanted microspheres at this time‐point to evaluate in vivo cell survival and hepatic phenotype preservation. Microspheres were found dispersed throughout the peritoneal cavity with no sign of bruising, inflammation, or fibrosis (Supporting Information [S6A](#sct312419-supitem-0001){ref-type="supplementary-material"}). Upon closer inspection, we observed that microspheres recovered from each mouse strain had remained mechanically intact in vivo*,* and detected no sign of host cell invasion into the alginate (Supporting Information [S6B](#sct312419-supitem-0001){ref-type="supplementary-material"}); demonstrating that the encapsulation had served as an effective barrier. Most importantly, the viability and hepatic phenotype of transplanted spheroids was preserved in vivo throughout the critical therapeutic window for an ALF bridging therapy. Live/dead immunofluorescence staining of recovered spheroids revealed a minimal number of nonviable cells (Fig. [6](#sct312419-fig-0006){ref-type="fig"}E), comparable to that of encapsulated cells maintained in vitro (Supporting Information [S7](#sct312419-supitem-0001){ref-type="supplementary-material"}). We performed immunohistochemical staining on recovered microspheres that confirmed that the encapsulated hepatocyte spheroids had remained positive for human KRT18 and *ALB.* Staining with STEM121 that reacts specifically to a human cytoplasmic protein, known to be expressed in various tissues including the liver, further confirmed that cells present within the microspheres were of human origin. Moreover, antibodies against murine CD45 revealed no host cell infiltration within the alginate (Fig. [6](#sct312419-fig-0006){ref-type="fig"}F and Supporting Information [S8](#sct312419-supitem-0001){ref-type="supplementary-material"}). Overall, this proof‐of‐concept study together with our recent publications on in vitro [63](#sct312419-bib-0063){ref-type="ref"} and in vivo [64](#sct312419-bib-0064){ref-type="ref"} alloimmune studies suggests that microspheres encapsulating cGMP‐compliant hPSC‐Heps are both safe and effective for cell‐based therapies aimed at ALF. In summary, we report that a library of cGMP‐compliant hPSCs can be differentiated into hepatocytes by a chemically defined protocol, which is suitable for clinical implementation. We confirm that this development goes via an endoderm‐like stage, before immature hepatocytes are obtained after 3 weeks. When matured by further culture in either 2D, or 3D models, these hPSC‐Heps express proteins known to be present on adult hepatocytes, have improved hepatic gene expression, and go on to demonstrate key liver functions. Discussion {#sct312419-sec-0021} ========== The manufacture of cGMP‐compliant hPSC lines and optimization of translationally relevant differentiation technologies, is essential for clinical application of hPSCs [65](#sct312419-bib-0065){ref-type="ref"}. We have, to our knowledge for the first time in this report, validated that a library of clinical grade cGMP‐hPSC can successfully be differentiated into hepatocytes in a chemically defined protocol. Cells generated in this way demonstrated genes, proteins, and hallmark functional characteristics of hepatocytes, but as shown by us and others previously [66](#sct312419-bib-0066){ref-type="ref"}, fell short of the benchmarks set by primary adult cells. By subsequently seeding these immature hepatocytes into bio‐engineered 3D scaffolds fabricated from FDA approved material, we were able to drive the cells into liver tissue functionally approximated to the standard needed for clinical efficacy. Following maturation within our 3D scaffold, elevated expression of key hepatic genes, such as *ALB*, *CYP3A7*, and *CYP3A4* for example were all found to occur. *CYP3A7* was first deemed as being exclusively expressed in the developing fetal liver [67](#sct312419-bib-0067){ref-type="ref"}, but it is now known to be present in up to 88% of adult livers [68](#sct312419-bib-0068){ref-type="ref"}, [69](#sct312419-bib-0069){ref-type="ref"}. This elevation in expression is of importance because cytochrome P450 enzymes are essential for the metabolism of numerous endogenous compounds and drugs. The CYP3A sub family specifically makes up 30% of the adult liver\'s cytochrome P450 constituency [70](#sct312419-bib-0070){ref-type="ref"} and metabolize half of marketed drugs [71](#sct312419-bib-0071){ref-type="ref"}. Importantly, the level of *CYP3A4* enzyme activity is also elevated in hPSC‐Heps within 3D ICC culture when compared with 2D; as is the albumin production rate of the cells. It is also important to highlight that the upregulation in hepatic functions, and the organoid‐like morphological transformation observed in cGMP‐compliant hPSCs, are well aligned with our previous 3D study using a more well‐established non‐cGMP‐compliant hiPSC line [53](#sct312419-bib-0053){ref-type="ref"}. The unique fabrication technique of our scaffold permits the scale‐up of this cell‐scaffold complex---up to containing the billions of cells required for human translation---and can be functionalized with different recombinant proteins, molecules or mechanical parameters. Cumulatively, these design features suggest such a scaffold could be the ideal carrier for delivering cGMP‐compliant hPSC‐Heps into patients. Assuming delivery of suitable numbers of functionally optimized cGMP cells can be achieved as above, a further challenge for clinical application will be to deal with the potential allogenic immune rejection of the host [72](#sct312419-bib-0072){ref-type="ref"}. Alginate hydrogel microencapsulation provides hPSC‐derivatives with a physical barrier from the recipient\'s immune system, through enclosure within a naturally occurring anionic polymer. Alginate, typically obtained from brown algae, is considered ideal for biomedical applications due to its biocompatibility, low cost, and ease of gelation [73](#sct312419-bib-0073){ref-type="ref"}. Numerous cell types have successfully been encapsulated within alginate, including mesenchymal stromal cells [74](#sct312419-bib-0074){ref-type="ref"}, pancreatic islets [75](#sct312419-bib-0075){ref-type="ref"}, and human hepatocytes [76](#sct312419-bib-0076){ref-type="ref"}, while optimized GMP grade alginate encapsulation protocols have already been established for the transplantation of human hepatocytes to provide metabolic function in patients with ALF [63](#sct312419-bib-0063){ref-type="ref"}. Status 1 ALF failure patients are assessed as having a life expectancy of hours, to a few days, without a liver transplantation [77](#sct312419-bib-0077){ref-type="ref"}. The median number of days from status 1 listing to death is just 5.5 with over a fifth of adults dying on the waiting list [78](#sct312419-bib-0078){ref-type="ref"}. The condition of patients with ALF can change significantly, even within 72 hours of being listed for transplantation [79](#sct312419-bib-0079){ref-type="ref"} with 13.9% of patients requiring "superurgent liver transplantation" due to ALF having a spontaneous recovery [80](#sct312419-bib-0080){ref-type="ref"}. These clinical reports support the hypothesis that ALF is potentially reversible without the need for transplantation if the host liver can be given enough time to recover. Our study demonstrated cGMP‐derived hepatic constructs remained viable, and more importantly, functional within the peritoneal cavity of fully immune competent C57BL/6 and CD‐1 mice for a time period long enough to result in recovery from ALF if used in patients. This in turn strongly advocates for further development of hPSC‐hepatocytes as a clinical therapeutic [63](#sct312419-bib-0063){ref-type="ref"}, [81](#sct312419-bib-0081){ref-type="ref"}, [82](#sct312419-bib-0082){ref-type="ref"}, [83](#sct312419-bib-0083){ref-type="ref"}. Conclusion {#sct312419-sec-0022} ========== We report here a library of clinical grade hPSCs manufactured under cGMP conditions amenable to reproducible hepatic differentiation, 3D culture, and alginate encapsulation that is potentially suitable for human application. Author Contributions {#sct312419-sec-0024} ==================== S.J.I.B., S.S.N.: concept and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing; J.M.S., A.J.F.K., A.L.A., D.K.: collection and/or assembly of data; J.M., M.S., D.I.: provision of study material or patients; A.D.: concept and design, manuscript writing; R.R.M.: concept and design, collection and/or assembly of data, manuscript writing; S.T.R.: concept and design, financial support, manuscript writing, final approval of manuscript. Disclosure of Potential Conflicts of Interest {#sct312419-sec-0025} ============================================= S.T.R. is a scientific founder, shareholder, and consultant for DefiniGen, Ltd. The other authors indicated no potential conflicts of interest. Supporting information ====================== ###### Appendix S1: Supplementary Data ###### Click here for additional data file. Generation of the GMP line LiPSC‐GR1.1 was supported by the NIH Common Fund Regenerative Medicine Program, and reported in Stem Cell Reports. The NIH Common Fund and the National Center for Advancing Translational Sciences (NCATS) are joint stewards of the LiPSC‐GR1.1 resource. We acknowledge Cell and Gene Therapy Catapult (London, U.K.) and Dr. Ricardo Baptista for the generation and provision of the CGT‐RCiB‐10 hiPSC line. We thank the Nikon Imaging Centre at Kings College London for help with spinning disk confocal microscopy. We are grateful to the KCH NHS Foundation Trust, BRC Flow Cytometry Facility for advice and technical assistance, and Dr. Nicholas Powell for provision of Rag2γ mice. S.J.I.B. was supported by a GSTT BRC Ph.D. award. S.T.R. was supported by an MRC Clinician Scientist Award (MGSBACR). [^1]: Co‐first authors.

Introduction {#s1} ============ Internal fertilization without copulation or another form of prolonged physical contact is a rare reproductive mode among vertebrates. In many newts (Salamandridae), the male deposits a spermatophore (a gelatinous mass topped with a packet of sperm) on the substrate in the water, which the female subsequently takes up with her cloaca. Because successful insemination in these animals requires intense coordination of both sexes, males have evolved a courtship display with several variations of tail-waving (henceforth used for different types of tail movements, but essentially consisting of tail-fanning, i.e. tail folded against the male\'s flank while undulating fast) towards the female\'s snout [@pone.0056538-Halliday1]--[@pone.0056538-Houck1] (Movie S1). A female reacts by closely following the courting male [@pone.0056538-Halliday2]--[@pone.0056538-Denol2] (Movie S1). The male then turns away from the female and she touches his tail, thereby confirming her following and prompting spermatophore deposition [@pone.0056538-Halliday1] (Movie S1). The whole courtship sequence ends with the female being lead over the spermatophore, which adheres to her cloaca and results in insemination. Although it is obvious that male tail-waving plays a central role in the courting process, and despite this behavior was initially described already two centuries ago [@pone.0056538-Spallanzani1], [@pone.0056538-Rusconi1], the exact effects of male tail-waving on female behavior remain unknown. The involvement of both visual and olfactory stimuli during courtship has been proposed. First, males develop species-specific epigamic characters during the breeding season, such as skin extensions (e.g. on legs, tail and crest) and intense coloration patterns [@pone.0056538-Griffiths1]. The impact of these visual stimuli during courtship is poorly understood [@pone.0056538-Halliday1], [@pone.0056538-Halliday3], but some studies suggested a direct or indirect female sexual preference for some of these male visual traits [@pone.0056538-Green1]--[@pone.0056538-Secondi1]. Second, males have sexually dimorphic glands in their cloaca [@pone.0056538-Verrell1], [@pone.0056538-Sever1], and it is generally acknowledged that male tail-fanning helps to send pheromones from the male\'s cloaca towards the snout of the female [@pone.0056538-Kikuyama1], [@pone.0056538-Noble1]--[@pone.0056538-Belvedere1]. This is also obvious from the fact that the male opens his cloaca during tail-waving (personal observation). Known sex pheromones in salamanders can be categorized in two main groups. Attractant pheromones facilitate the location of potential mates and serve to bring males and females together [@pone.0056538-Woodley1]. Conversely, courtship pheromones are used only after initial contact with a potential mate, and operate by altering female behavior during courtship [@pone.0056538-Arnold2], [@pone.0056538-Vaccaro1]. Until now, only one population-specific decapeptide [@pone.0056538-Nakada1] and its two species-specific variants [@pone.0056538-Iwata1] in the genus *Cynops* have been described in salamandrids. Using experiments with sponges [@pone.0056538-Kikuyama1], [@pone.0056538-Nakada1]--[@pone.0056538-Toyoda1], these molecules were shown to work as sex attractants. In other salamandrids, experiments with linear olfactometers [@pone.0056538-Secondi2], Y-mazes [@pone.0056538-Belvedere1], [@pone.0056538-Malacarne1] and two-choice aquaria [@pone.0056538-Poschadel1], [@pone.0056538-Osikowski1] also suggest the presence of attractant pheromones. However, since a male only starts tail-waving in front of the female after an initial investigation [@pone.0056538-Halliday1], [@pone.0056538-Griffiths1], [@pone.0056538-Houck1], [@pone.0056538-Halliday2], [@pone.0056538-Wells1], the pheromones emitted during this behavior are expected to be courtship rather than attractant pheromones [@pone.0056538-Arnold2], [@pone.0056538-Vaccaro1], and a straightforward function for these chemical cues therefore remains unknown. The demonstration and functional interpretation of courtship pheromones preferentially requires a behavioral experiment that, in addition to control over pheromone presence, includes the visual and tactile stimuli of a second animal mimicking the presence of a mating partner in a natural courtship. Here we developed such experiments with two females. Using a second female instead of a male allowed us to separate the male\'s visual secondary sexual characters and his courtship display from olfactory cues. To test for a pheromone function, we used water in which a male had been courting another male as a source of olfactory cues. We show that the male chemical cues emitted during tail-waving are courtship pheromones that induce all typical features of natural female responses culminating in spermatophore pick-up from the substrate. Because this induced female behavior is essential in getting her cloaca positioned for spermatophore pick-up, these pheromones are a key aspect of insemination without physical contact in salamandrids. Materials and Methods {#s2} ===================== Ethics statement {#s2a} ---------------- The research was done with permission and according to the guidelines of Agentschap voor Natuur en Bos (permit ANB/BL-FF/V12-00050). All experiments complied with EU and Belgian regulations concerning animal welfare. Animals were released back to the pond of their origin after the experiments were finished. Animals {#s2b} ------- For both alpine newts (*Ichthyosaura alpestris*) and palmate newts (*Lissotriton helveticus*), 40 adult females and 40 adult males were collected in February/March 2012 from a pond near Hasselt, Belgium, using a modified Ortmann\'s funnel trap [@pone.0056538-Drechsler1]. They were housed in single sex aquaria (60×35×35 cm) filled up with 30 liters of aged tap water and containing vegetation from the pond of the newts\' origin. Up to 15 newts were kept together per aquarium. The temperature and light regime were artificially regulated (15--18°C, 13 h/11 h light/dark). Newts were fed *ad lib* with maggots or earthworms every other day. Receptivity test {#s2c} ---------------- Behavioral experiments and collection of stimuli were preceded by a receptivity test on the day before the experiment, and only receptive animals were selected. Receptivity was tested by putting a male and a female together in a plastic container (25×16×14 cm) filled with 800 ml of aged tap water. Males that courted successfully were considered receptive, and were used for collection of chemical stimuli. Only females that followed a courting male were selected for the behavioral experiments. During the receptivity test, females were not allowed to pick up spermatophores. Chemical stimuli and control {#s2d} ---------------------------- Collection of male courtship water was done by putting two receptive males in a plastic container (25×16×14 cm) filled with 800 ml of aged tap water. To ensure similar concentrations in all experiments, we measured ten minutes of active male tail-fanning. Approximately one on eight males was tail-fanning to another male for this amount of time. The males were then removed and the male courtship water was used immediately in the behavioral experiment. The water of two non-courting males was collected in the same way and for the same time as above, but for each male separately to ensure that there was no courtship display. We also performed experiments using this water to show that the observed courtship behavior is induced by molecules that are released during male courtship display only. Control behavioral experiments were done in aged tap water. For all experiments, we used aged tap water that had been kept in the room where the animals were housed, ensuring similar temperature conditions. Experiment design {#s2e} ----------------- Experiments were conducted in a plastic container (25×16×14 cm) covered with brown cardboard from the outside to prevent outside visual stimuli. The container was filled with 800 ml of either chemical stimuli or control water. Two randomly picked receptive females were taken with a glove from their housing container, briefly rinsed in aged tap water and placed in the experiment container. Their behavior was recorded for 10′ using a digital camera connected directly to a computer, where the video recordings were stored for later analyses. Females were used in one experiment per day only and only once in each experimental design. To reduce potential variation in receptivity to a minimum, all experiments were performed on consecutive days. They were done during the day, at the same time of the day, and under the same light and temperature conditions. We performed both intraspecific (using two alpine newt females) and interspecific experiments (using one alpine newt and one palmate newt female). Data analysis and statistics {#s2f} ---------------------------- The video recordings of the experiments were analyzed for female responses similar to those under natural conditions with a male. For statistical comparison, we classified behavior in three quantifiable components ([Fig. 1](#pone-0056538-g001){ref-type="fig"}): First, females showed the typical following behavior (*f*), i.e. they closely followed the movements of the other female. Second, the following female frequently touched the tail (*tt*) of the leading female, thereby mimicking the natural behavior of tail-touching before spermatophore deposition. Third, male courtship water eventually induced female tail-waving (*w*) behavior (resembling that of male tail-fanning), similar to that shown after prolonged courtship with a male. These behaviors can be shown by one of the two females, but sometimes both females tried to follow each other simultaneously, resulting in a wheel-shaped movement, or reciprocal tail-waving. Quantification was done as follows: {ref-type="sec"}). (**B**) *Equivalent female behavior in a two-female experiment:* after addition of male courtship water, one female follows the other one, or both females try to follow each other; the following female regularly touches the tail of the other one; a female uses tail-waving in trying to encourage the other female.](pone.0056538.g001){#pone-0056538-g001} 1. *Following*: the amount of time in which a female incessantly shows interest towards the other female, including turning towards another female and following her. Because this could be considered objective, we also measured pointing, i.e. the amount of time that an imaginary straight line, perpendicular to the line connecting the eyes of the following female, intersects the other female\'s body. The results of these analyses (not shown) were very similar and equally statistically significant. 2. *Tail-touching*: the duration of a female touching another female\'s tail with her snout. To avoid interpreting accidental touching (e.g., as the result of the first female moving her tail against the following female) as a positive response, only the female\'s snout actively pressed against the other female\'s tail at an angle of 45°--90° was measured. 3. *Tail-waving*: the duration of one female waving her tail towards the other female (see [results](#s3){ref-type="sec"}). It was measured from the time the female starts bending her tail to start waving until the tail was bent back more than 90° away from the female\'s body. The cumulative duration of each behavior was measured in seconds. The differences in behavior between different stimuli were tested with the Kruskal-Wallis test followed by *post hoc* two-tailed Mann-Whitney U test [@pone.0056538-Dytham1] using SPSS [@pone.0056538-SPSS1]. Results {#s3} ======= Female tail-waving to a male {#s3a} ---------------------------- Although tail-waving is often regarded as a male-specific behavior, we observed this behavior in females during receptivity tests. Females were not allowed to pick-up spermatophores during these tests, which resulted in prolonged courtship and eventually reduced male courtship display. In many instances, the female then started tail-waving to the male (Movie S1), but stopped this behavior and continued following as soon as the male resumed his courtship display. This female tail-waving was only observed after an initial period of male courtship, i.e., after a female had already been following a male, indicating that this behavior is performed only after contact with pheromones. Although the function of male pheromones is not to induce female tail-waving, this behavior occurs in male courtship water, but not in non-courting male water and control water, and therefore was used in our experiments as indirect evidence of active courtship pheromones. Intraspecific two-female experiments in alpine newts {#s3b} ---------------------------------------------------- Our two-female experiment with two alpine newts showed that male courtship water induces following, tail-touching and tail-waving in females (Movie S2), while these behaviors are practically absent in control water (two-tailed Mann-Whitney U-test, *P~f~*\<0.001, *P~tt~*\<0.001 and *P~w~*\<0.05, respectively) ([Fig. 2](#pone-0056538-g002){ref-type="fig"}). All typical features of a female\'s mating responses can thus be evoked without the male\'s visual secondary sexual and behavioral characteristics, such as the crest and coloration, and tail-waving, respectively. Tests with water in which non-courting males had been kept were not significantly different from control water (*P~f~* = 0.152; *P~tt~* = 0.130, *P~w~* = 0.317), but were significantly different from male courtship water (*P~f~*\<0.001; *P~tt~*\<0.05, *P~w~*\<0.01) ([Table 1](#pone-0056538-t001){ref-type="table"}). Altogether, these experiments indicate that the female behavioral responses in alpine newts are caused by pheromones emitted during the male\'s courtship display, and do not require male-specific visual stimuli. {#pone-0056538-g002} 10.1371/journal.pone.0056538.t001 ###### Results of the statistical tests. {#pone-0056538-t001-1} stimulus 1 *N* stimulus 2 *N* focal female behavior Mann-Whitney U *U* *Z* ----------------------------------------------------- ----- ----------------------- ----- ----------------- ---------- ---------------- ----------- ---------- Intraspecific test (*I. alpestris*+*I. alpestris*) Ia 20 MW 20 *I. alpestris* f \<0.001\* *42.000* *−4.621* tt \<0.05\* *123.000* *−2.207* w \<0.01\* *126.500* *−2.719* Ia 20 W 20 *I. alpestris* f \<0.001\* *30.000* *−5.109* tt \<0.001\* *93.000* *−3.269* w \<0.05\* *120.000* *−3.097* MW 20 W 20 *I. alpestris* f 0.152 *180.000* *−1.432* tt 0.130 *156.000* *−1.514* w 0.317 *190.000* *−1.000* Interspecific test (*I. alpestris*+*L. helveticus*) Ia 15 W 15 *I. alpestris* f \<0.001\* *90.000* *−3.782* tt \<0.001\* *22.500* *−4.215* w \<0.05\* *75.000* *−2.396* Lh 15 W 15 *I. alpestris* f 1.000 *112.500* *0.000* tt 0.150 *97.500* *−1.438* w 1.000 *112.500* *0.000* Interspecific test (*I. alpestris*+*L. helveticus*) Ia 15 W 15 *L. helveticus* f 1.000 *112.500* *0.000* tt 0.524 *104.500* *−0.637* w 1.000 *112.500* *0.000* Lh 15 W 15 *L. helveticus* f \<0.001\* *37.500* *−3.707* tt \<0.05\* *69.000* *−2.225* w 0.317 *105.000* *−1.000* Intraspecific vs. interspecific female preference Ia (+*I. alpestris*) 20 Ia (+*L. helveticus*) 15 *I. alpestris* f 0.340 *121.500* *−0.954* tt 0.183 *110.500* *−1.333* w 0.398 *128.000* *−0.846* Ia = alpine newt male courtship water, MW = water in which non-courting alpine newt males had been kept, W = control water, Lh = palmate newt male courtship water, N = number of different females used in the experimental design. *P*-values smaller than 0.05 are considered significant and are indicated with an asterisk. The values for Kruskal-Wallis tests were as follows: intraspecific tests (*P~f~*\<0.05, *P~tt~*\<0.05 and *P~w~*\<0.05), interspecific tests for alpine newt females (*P~f~*\<0.05, *P~tt~*\<0.05 and *P~w~*\<0.05), interspecific tests for palmate newt females (*P~f~*\<0.05, *P~tt~*\<0.05 and *P~w~* = 0.096). Interspecific experiments {#s3c} ------------------------- To further test the requirement of conspecific (but not necessarily male-specific) visual cues and species-specificity of male courtship chemical stimuli, we repeated our two-female experiment using one alpine newt and one palmate newt. The latter species is much smaller and lacks the orange belly and dark dorsal skin (species-specific visual stimuli in alpine newts [@pone.0056538-Himstedt1]), but shows similar courtship behavior [@pone.0056538-Wambreuse1]. Our experiments indicate that alpine newt male courtship water elicited the three typical courtship behaviors in alpine newt females (*P~f~*\<0.001; *P~tt~*\<0.001, *P~w~*\<0.05) but not in palmate newt females (*P~f~* = 1; *P~tt~* = 0.524; *P~w~* = 1), while the opposite was true in palmate newt male courtship water (palmate newt females: *P~f~*\<0.001; *P~tt~*\<0.05, *P~w~* = 0.317; alpine newt females *P~f~* = 1; *P~tt~* = 0.150; *P~w~* = 1) ([Fig. 2](#pone-0056538-g002){ref-type="fig"}) (Movie S3). These experiments indicate that male courtship chemical stimuli are species-specific, because they only stimulate conspecific females while the other species\' behavior remains unaffected and similar to that in control water. Furthermore, affected females of alpine newts found females of both species equally seductive in alpine newt male courtship water (*P~f~* = 0.340; *P~tt~* = 0.183; *P~w~* = 0.398) ([Table 1](#pone-0056538-t001){ref-type="table"}), indicating that chemical stimuli can equally induce female courtship behavior towards another newt, i.e. in the absence of species-specific and gender-specific visual cues. Discussion {#s4} ========== Behavioral studies searching for pheromones in Salamandridae until now had focused on an attractant function of the chemical stimuli, i.e. showing that olfactory communication brings males and females together. However, although such experiments indicated olfactory recognition in salamandrids, they did not provide a satisfactory explanation for one of the most obvious behaviors indicating pheromone use, i.e. male tail-fanning. Our two-female experiments here show that pheromones emitted during male courtship are able to induce all female responses that precede spermatophore pick-up. This is evidenced by three behaviors, following, tail-touching and tail-waving, all of which can be observed under influence of male pheromones released during tail-fanning, but are basically non-existent in the absence of these molecules. The olfactory cues emitted during male tail-waving therefore resort under the group of courtship pheromones, i.e. molecules aimed at changing behavior in females. Male newts that perform courtship more energetically are known to have a higher mating success [@pone.0056538-Teyssedre1], which in light of our experiments can be explained as a more efficient transfer of olfactory stimuli. Our interspecific experiments suggest that visual cues are largely irrelevant when the pheromone has reached a certain threshold in the female. This helps to explain the occurrence of sexual interference in newts, i.e., males interfering in courtship of another male. During courtship, a second male sometimes intervenes between both sexes and deposits a spermatophore in front of the already following female [@pone.0056538-Denol1], [@pone.0056538-Verrell2], [@pone.0056538-Sparreboom1]. We hypothesize that such newts exploit olfactory stimuli of another male and the indiscriminate behavior of the following female to enhance their own reproductive success. Male visual stimuli would then mainly serve as sexual defense against such sexual interference [@pone.0056538-Arnold3]. Being more visually attractive would insure a courting male that he remains the main focus of the female that is stimulated by olfactory stimuli [@pone.0056538-Halliday1], [@pone.0056538-Verrell3]. A higher importance of olfactory stimuli over visual stimuli had already been shown in experiments with high light and low light conditions in alpine newts [@pone.0056538-Denol3]: Although both sexes spent significantly more time in orientation (without courtship displays) in low light conditions, males were performing more tail-fanning and obtained more positive response from females, resulting in similar mating success as in high light conditions. Our experiments surprisingly revealed that females under influence of male courtship pheromones can show tail-waving to a potential mating partner, a behavior that is suggestive of the presence of female olfactory cues. This kind of male-like behavior has been termed 'pseudo-male' or 'heterotypical' female behavior [@pone.0056538-Halliday4]--[@pone.0056538-Robalo1] and might be caused by sexual frustration, when females are deprived of males and are highly receptive [@pone.0056538-Halliday4], [@pone.0056538-Denol4]. We therefore hypothesize that, just as in males, tail-waving helps the female to deliver her olfactory stimuli [@pone.0056538-Verrell4], [@pone.0056538-Verrell5] more efficiently to the male, thereby stimulating him to continue courtship. Olfactory investigation of females by males is common prior and during courtship [@pone.0056538-Halliday1], [@pone.0056538-Griffiths1], [@pone.0056538-Houck1], [@pone.0056538-Halliday2], [@pone.0056538-Arnold2], [@pone.0056538-Wells1]. For example, female olfactory cues are known to induce male tail-fanning in the smooth newt [@pone.0056538-Zippelius1]. Additionally, crested newt males do not show sexual behavior after their nostrils are plugged and female olfactory cues cannot reach them [@pone.0056538-Malacarne2]. The inferred fundamental role of pheromone use in newts has important implications for their reproduction. If the female does not follow the courting male, he will rarely deposit a spermatophore, and even if he would, chances that it would be picked up by the female are low [@pone.0056538-Halliday1], [@pone.0056538-Halliday2], [@pone.0056538-Denol2], [@pone.0056538-Halliday3]. Therefore, male courtship pheromones in salamandrids are a necessary prerequisite for newt reproduction. Such a strong dependence on a pheromone is unusual in amphibians. In lungless salamanders (Plethodontidae), it has been shown that two pheromones (SPF and PRF), alone or in combination with another pheromone (PMF), reduce courtship time (i.e., the time to spermatophore deposition) [@pone.0056538-Rollmann1]--[@pone.0056538-Houck3]. However, spermatophore deposition in this family equally occurs in salamanders after ablation of their courtship mental gland, i.e. in the absence of these pheromones. Since this is not the case in salamandrids [@pone.0056538-Arnold1], [@pone.0056538-Toyoda2], such a strong olfactory dependence would become critical for newt survival if pollution would interact directly (e.g., chemical interaction with the pheromones in the water) or indirectly (e.g., by altering hormone levels involved in the pathway of pheromone production) with the pheromone system [@pone.0056538-Park1], [@pone.0056538-OrtizSantaliestra1]. Our behavioral experiments indicate that the pheromone function of tail-waving is evolutionary conserved in at least two lineages, *Ichthyosaura* and *Lissotriton*. Variations aside, a courtship that basically consists of tail-waving and the male leading the female over the deposited spermatophore is present in multiple genera and species of newts [@pone.0056538-Halliday1], [@pone.0056538-Griffiths1], [@pone.0056538-Houck1], [@pone.0056538-Halliday2], [@pone.0056538-Wells1]. It is therefore likely that this pheromone function originated early in newt evolution and has been retained in several genera. Similar to pheromones in Plethodontidae, salamandrid pheromones are likely to constitute fast evolving molecules [@pone.0056538-Iwata2]--[@pone.0056538-Wilburn1]. Our two-female experiments indicate that the olfactory cues produced during tail-waving are species-specific (hence the term pheromone can be used, [@pone.0056538-Wyatt1]) only inducing the typical behavior in females of the same species. These molecules therefore have the potential to function as a significant mechanism of reproductive isolation [@pone.0056538-Iwata1], [@pone.0056538-Osikowski1], [@pone.0056538-Malacarne3]. Multiple species of newts co-inhabit the same ponds during the breeding season [@pone.0056538-Macgregor1], and their crepuscular or nocturnal lifestyle [@pone.0056538-Griffiths2], [@pone.0056538-Rafinski1] in often turbid water [@pone.0056538-Secondi3] can make visual mate recognition difficult. The recognition of conspecific male courtship pheromones is probably an important mechanism in maintaining reproductive isolation and reducing hybridization in salamandrids. Conclusions {#s5} =========== Interpretation of the function of a pheromone often depends on the behavioral experiments being performed. Although it could be argued that pheromones produced during tail-waving are attractants because they keep the female close to the male due to following, they essentially are courtship pheromones, i.e. they alter female courtship responses [@pone.0056538-Vaccaro1]. Our two-females experiments reveal that male courtship pheromones are extremely potent, rendering even females of other genera irresistible to exposed females. The essence of male courtship pheromones therefore goes beyond the traditionally tested attractant function, and is an essential element for eliciting the mating behavior in females leading to successful insemination without physical contact. Our new behavioral experiment, that seems to work over a broad range of species, opens possibilities for the identification and study of pheromones across the whole clade of newts. Supporting Information {#s6} ====================== ###### **Male-female natural mating behavior in alpine newts.** (MP4) ###### Click here for additional data file. ###### **Two-female experiments (intraspecific).** (MP4) ###### Click here for additional data file. ###### **Two-female experiments (interspecific).** (MP4) ###### Click here for additional data file. We thank the staff of the botanical garden of Ghent University and Marc Fourier (Natuurpunt) for collaboration in obtaining newts for the research, Kim Roelants for making alpine newt drawings, Frank Pasmans, Margo Maex and two anonymous reviewers for constructive comments on an earlier draft. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: DT IVB FB. Performed the experiments: DT IVB SM DDF SJ BW FB. Analyzed the data: DT IVB. Contributed reagents/materials/analysis tools: BW. Wrote the paper: DT IVB SM DDF SJ BW FB.

Background {#Sec1} ========== Autochthonous malaria in the Atlantic Forest occurs mainly in the Southern and Southeastern regions of Brazil, where it is also called "bromeliad malaria" because the immature forms of the main mosquito vectors develop in the water that collects in bromeliads (Bromeliaceae), an abundant group of plants in this biome \[[@CR1]--[@CR4]\]. In recent decades there has been a low incidence of outbreaks of autochthonous malaria in the Atlantic Forest, and most cases have been asymptomatic with low circulating parasite loads \[[@CR5], [@CR6]\]. The main parasite involved in the transmission of malaria in the Atlantic Forest is *Plasmodium vivax* and, less frequently, *Plasmodium malariae* and *Plasmodium falciparum* \[[@CR4]\]. The species known to be vectors of bromeliad malaria belong to the subgenus *Kerteszia* of the genus *Anopheles,* and are distributed along the Atlantic coast of Brazil although some can also be found in the Amazon region \[[@CR7]\]. *Anopheles* (*Kerteszia*) *cruzii* is considered the species most associated with the transmission of human and simian *Plasmodium* in the Atlantic Forest \[[@CR8]--[@CR12]\]. Originally it had a widespread distribution in Brazil, extending from the Southern to Northeastern regions of the country \[[@CR7]\]. However, with increasing deforestation in the Atlantic Forest its distribution has decreased \[[@CR2], [@CR4], [@CR13]\]. Immature forms of *An. cruzii* usually develop in epiphytic or terrestrial bromeliads in shady places under the forest canopy and are very abundant in humid forests on coastal slopes \[[@CR14], [@CR15]\]. The species tends to be more abundant in hot, rainy periods but can be observed frequently in the forest even in drier, cooler seasons \[[@CR16]--[@CR18]\]. *Anopheles cruzii* females exhibit increased biting activity in the first few hours after evening twilight but can be found actively biting throughout the night or even in daylight hours in the forest, in peridomestic areas or even inside houses \[[@CR16]\]. The term "acrodendrophily" refers to the tendency of certain wild species to live and preferentially feed in the canopy. The term was proposed by Garnham et al. \[[@CR19]\], when they observed that some mosquitoes were collected more frequently in the upper stratum of the Kaimosi Forest in Kenya. Many studies have demonstrated that *An. cruzii* has acrodendrophilic behaviour although in several studies and locations this species has been observed in high density at ground level in the forest \[[@CR20]--[@CR24]\]. In locations where *An. cruzii* is more acrodendrophilic, human cases of malaria do not seem to occur or are rarer, even in the presence of monkeys infected with simian *Plasmodium* species. During surveys in Southern and Southeastern Brazil, Deane et al. \[[@CR25], [@CR26]\] noted that in the Serra da Cantareira Forest, in the state of São Paulo, more than 99% of *An. cruzii* individuals were collected in the canopy. Although 62% of the howler monkeys (*Alouatta guariba clamitans*) tested in the study area were infected by *Plasmodium*, only one natural, accidental human infection due to *Plasmodium simium* was observed. The same occurred in the municipality of Campo Alegre, in the State of Santa Catarina, where approximately 90% of the *An. cruzii* collected were collected in the canopies and 43% of the howler monkeys tested were infected with *Plasmodium* spp. but no human cases were observed \[[@CR26]\]. A similar pattern was observed in studies in the municipalities of Guaíba, in the state of Rio Grande do Sul, and Santa Leopoldina, in the state of Espírito Santo \[[@CR27], [@CR28]\]. However, near the municipality of Joinville, Santa Catarina, where there have been cases of simian and human malaria, 42% of the *An. cruzii* collected were collected at ground level \[[@CR26]\]. Deane et al. \[[@CR21]\] tested the hypothesis of reproductive isolation among populations of *An. cruzii* collected in traps placed in the canopy and at ground level in a forest near the municipality of São Francisco do Sul, Santa Catarina. Using mark-recapture techniques with fluorescent dust, the authors noticed that there was a vertical movement of *An. cruzii* between the canopy and ground level in the forest as mosquitoes previously collected and marked in the canopy were later collected in ground traps and vice versa. This observation reinforced the hypothesis that under certain circumstances *An. cruzii* may carry *Plasmodium* species between humans and non-human primates, establishing a cycle of zoonotic malaria transmission in these regions. Recently, it has been shown that human infection by *P. simium* or other variant forms of *P. vivax* that circulate among monkeys in the Atlantic Forest may be more common than previously thought \[[@CR29]\], reinforcing the need for further investigations of ecological aspects of zoonotic transmission of bromeliad malaria, especially in areas where *An. cruzii* occurs in high abundance. Variations in the acrodendrophilic behaviour of *An. cruzii* could be due to the increased human presence in the Atlantic Forest and consequent changes in the environment. Although these changes have led to a degradation in the habitats that favour the development of this species, they can lead to increased contact between humans and females of the local populations of this vector as the increased human presence represents a greater supply of blood meals at ground level in the forest and, potentially, a reduced supply of other vertebrate hosts in the canopy. Landscape changes due to increased human presence can, therefore, be expected to lead to a reduction in *An. cruzii* abundance, but to favour an increase in the relative proportion of this mosquito searching for blood sources at ground level. To test this hypothesis, entomological surveys were performed and landscape metrics were measured in a remnant of Atlantic Forest on the outskirts of the megacity of São Paulo where several cases of human malaria have been reported in the last decades \[[@CR11], [@CR30], [@CR31]\]. Methods {#Sec2} ======= Study area {#Sec3} ---------- The Capivari-Monos EPA (Environmental Protection Area) is approximately 40 km from the center of the city of São Paulo, Brazil, and extends over the first hills near the crest of the Serra do Mar mountain range at altitudes varying from 740 to 850 m above sea level. The climate is a super-humid, oceanic, tropical climate with average annual temperatures of around 19 °C and rainfall of between 1600 and 2200 mm \[[@CR32]\]. The original vegetation is dense montane ombrophilous forest found in remnants in varying degrees of conservation, from areas in regeneration dating from the 1950s, when logging was stopped, to areas recently degraded as a result of urban and rural expansion. The district of Engenheiro Marsilac lies within the EPA and has a population of around 10 thousand inhabitants and a population density of approximately 41 inhabitants per km^2^. The population, most of whom have low incomes, live mainly in rural settlements \[[@CR32]\]. Cases of autochthonous malaria have been recorded in the district in the last decades as well as in neighbouring municipalities such as Embu-Guaçu, São Bernardo do Campo, Itanhaém and Juquitiba and the Serra do Mar State Park \[[@CR11], [@CR30], [@CR31]\]. To select the sampling sites, landscape variations in the region were considered and were included areas with different degrees of anthropogenic influence (Fig. [1](#Fig1){ref-type="fig"}). The areas where specimens were collected were (1) Embura---a village surrounded by small farms and the EPA forest (23° 53.036′ S/46° 44.485′ W); (2) Marsilac---a village surrounded by the EPA forest and near a railway line (23° 54.395′ S/46° 42.486′ W); (3) transition zone---private property near Marsilac village constituting a transitional area between a rural environment and the EPA forest (23° 54.556′ S/46° 42.167′ W); (4) Cachoeira do Marsilac---private property in the EPA forest next to a waterfall with a visitation area (23° 56.378′ S/46° 41.659′ W); and (5) Evangelista de Souza---well-preserved EPA forest near a railway station (23° 56.140′ S/46° 38.090′ W).Fig. 1Study sites in the Capivari-Monos EPA, São Paulo, Brazil: (1) Embura village, (2) Marsilac village, (3) Transition zone, (4) Cachoeira do Marsilac and (5) Evangelista de Souza. The areas were classified according to the map of Atlantic Forest biome remnants in the municipality of São Paulo (available at <http://geosampa.prefeitura.sp.gov.br/PaginasPublicas/_SBC.aspx>). Green represents dense ombrophilous forest; blue represents heterogeneous forest; pink, natural fields; and white, areas where there is human activity (roads, rural properties or villages). Crosses inside the circles indicate collection points. The inner circle represents a 500 m buffer and the dashed circle a 1 km buffer around the collection points. The map was created using QGIS v2.18.9 (<http://www.qgis.org>) Field collections {#Sec4} ----------------- Culicidae collections were carried out monthly from March 2015 to April 2017. To determine the frequency of *An. cruzii* in the canopy and at ground level, automatic CDC traps with CO~2~ bait were used at each collection point, one trap in the tree canopy in heights between 10 and 13 m, depending on the selected branch of the tree (measured according to the length of the ropes used to hoist the traps), and one trap about 1 m from the ground. Traps were installed early in the afternoon and removed the following morning after approximately 18 h of exposure (Fig. [2](#Fig2){ref-type="fig"}a, b). Considering that between the years of 2013 to 2015 in the region of study the sunset ranged from 5:28 p.m. (May--June) to 6:59 p.m. (January) and the sunrise ranged from 5:12 a.m. (November--December) to 6:50 a.m. (June--July) \[[@CR33]\], the data obtained by CDC traps were not corrected according to seasonality changes in day-length \[[@CR34]\], since the 18-h exposure period of the CDC traps ensured that all traps were operating during the hours of higher biting activity of *An. cruzii* \[[@CR16], [@CR17]\].Fig. 2Collection techniques used in the study: **a** CDC trap installed near the ground; **b** CDC trap installed in the canopy of a tree; and **c** Shannon trap. Images: Laboratory for Entomology in Public Health, University of São Paulo (LESP/FSP/USP) Shannon traps were set up near the CDC traps to collect mosquitoes during the first 2 h after evening twilight (when many mosquito species, including *An. cruzii*, exhibit increased biting activity) \[[@CR16], [@CR17]\]. In each Shannon trap, two individuals with manual battery-powered aspirators collected mosquitoes on the outer and inner surface of the tent (Fig. [2](#Fig2){ref-type="fig"}c). For the Evangelista de Souza site, data on *An. cruzii* abundance and frequency at ground level and in the canopy were obtained from Ribeiro et al. \[[@CR35]\] and Duarte et al. \[[@CR11]\], who performed field collections from May 2009 to June 2010. These authors used the same collection methods and traps as those used in the present study and very kindly provided all the information on their collections in exhaustive detail to the present authors. For Embura village, the information from the studies cited above and the collections in the present study were included. Specimens were identified morphologically in the Laboratory for Entomology in Public Health, School of Public Health, University of São Paulo (LESP/FSP/USP), with taxonomic keys for Culicidae \[[@CR2], [@CR15], [@CR36]\]. Landscape analysis {#Sec5} ------------------ To investigate whether landscape changes influence variations in *An. cruzii* abundance and acrodendrophilic behaviour, landscape composition and configuration metrics were calculated and used as explanatory variables. The locations of the Shannon traps were georeferenced in the field and plotted with QGIS 2.18 (<http://www.qgis.org>) on the map of Atlantic Forest biome remnants in the municipality of São Paulo (a 1:5000-scale orthophoto mosaic) available at <http://geosampa.prefeitura.sp.gov.br/PaginasPublicas/_SBC.aspx>. In order to investigate whether variations in the abundance and acrodendrophily of *An. cruzii* may be better predicted by a more local or wider landscape scale, buffers extending 500 m and 1 km were created around each collection point to define the surrounding landscape. These scales were chosen based on observed flight radius of *An. cruzii*, which accordingly to Ferreira et al. \[[@CR37]\] was about 1000 m. Five classes of vegetation or land use were observed: dense ombrophilous forest, heterogeneous forest, natural fields, anthropogenic areas and villages. The classes "dense ombrophilous forest" and "heterogeneous forest", environments that favour the development of *An. cruzii*, were grouped into a single class called "forest cover". Similarly, the classes "anthropogenic areas" and "villages" were grouped into a class called "anthropogenic use". Landscape composition was measured by considering the relative abundance of each landscape class within each buffer (except for the class "natural fields", which was considered of minor importance). Landscape configuration, considered here as the degree of fragmentation of the class "forest cover" in each landscape matrix, was measured as the number of "forest cover" fragments and their total edge length in kilometres (Table [2](#Tab2){ref-type="table"}). Landscape configuration metrics were determined with Fragstats v4.2 \[[@CR38]\]. Data analysis {#Sec6} ------------- The abundance and frequency of *An. cruzii* collected at ground level were considered response variables. Mosquitoes collected in Shannon traps were used to calculate abundance because this technique allows the number of mosquitoes per human host/hour during peak blood-feeding periods (the first 2 h after evening twilight) to be measured. Therefore, the response variable "abundance" used in the statistical models refers to the average number of *An. cruzii* specimens per human/hour in the Shannon trap calculated for each sample and site. To investigate variations in *An. cruzii* acrodendrophily, the data from the CDC traps were used. Mosquito frequency at ground level was calculated by dividing the number of *An. cruzii* in the trap at ground level by the total number of individuals of this species collected at the site in the same sample (number of specimens collected in the CDC trap at ground level plus the number of specimens collected in the CDC traps in the canopy). Since the samples collected at each site represent pseudo-replicates (repeated measures over time at the same site), generalized linear mixed-effect models were used \[[@CR39]\]. Thus, the fixed effect was represented by the landscape composition and configuration metrics, and the random effect by the different years and months when collections were carried out. The variables year and month were considered as different random factors, since monthly variations reflect the effect of seasonality, while annual variations are less predictable, and may reflect atypical climatic or environmental conditions \[[@CR40]\]. Because of the low number of sampling sites, it was decided to test models with only one predictive variable in the fixed effect. For the abundance models it was opted for Poisson errors (log link), and for the ground/canopy frequency binomial errors (logit link) was used. An information-theoretical approach based on the Akaike Information Criterion corrected for small samples (AICc) was applied to select the most plausible statistical models \[[@CR40]\]. The models with the smallest AICc were considered the best fit, and ∆AICc ≤ 2 was adopted as the cutoff to select models with more empirical support. The strength of evidence in favour of each model was evaluated using Akaike weights \[[@CR41]\]. The selected models were checked for independence of residuals, over dispersion and presence of zero-inflated data \[[@CR42]\]. In all cases the models were an adequate fit for the expected behaviour. All analyses were performed with the R computational environment \[[@CR43]\] and the lme4 \[[@CR44]\], bbmle \[[@CR45]\], DHARMa \[[@CR46]\] and ggplot2 \[[@CR47]\] packages. Results {#Sec7} ======= A total of 15,764 mosquitoes belonging to 80 species/taxa in 15 genera were collected between March 2015 and April 2017 (Additional file [1](#MOESM1){ref-type="media"}). Among these, 6823 specimens of *An. cruzii* were identified, of which 781 individuals were collected in CDC traps (11.4% of the total) and the remainder in Shannon traps. Based on these data and the data from collections in Embura village and Evangelista de Souza between 2009 and 2010, *An. cruzii* represents 48% of all the mosquitoes collected in the Capivari-Monos EPA in these studies. However, the relative abundance of this species varied between the sites, ranging from approximately 5% in Embura village to more than 74% of all the mosquitoes collected in Evangelista de Souza. Average abundance of *An. cruzii* per human/hour in the Shannon traps varied from 0.5 in Embura to 57.5 in Evangelista de Souza. Average frequency of this species at ground level (based on CDC traps) varied from 0.18 (18%) in Evangelista de Souza to 0.58 (58%) in Embura village (Table [1](#Tab1){ref-type="table"}).Table 1Results of mosquito collections at each site: total number of specimens, total number of *Anopheles cruzii*, relative abundance of *An. cruzii*, mean abundance of *An. cruzii* per human/hour in Shannon traps and mean proportion of *An. cruzii* collected at ground level in CDC trapsSiteTotal no. individualsTotal no. of *An. cruzii*Relative abundance of *An. cruzii*Mean abundance of *An. cruzii* per human/hour (Shannon traps)Mean proportion of *An. cruzii* collected at ground level (CDC traps)Embura village49592700.050.5 (0.3--0.8)0.58 (0.43--0.72)Marsilac village28088410.307.25 (2--12)0.24 (0.18--0.30)Transition zone457720490.4522 (9--34)0.47 (0.40--0.53)Cachoeira do Marsilac563839130.6942.5 (8--77)0.22 (0.19--0.26)Evangelista de Souza589343770.7457.5 (34--137)0.18 (0.13--0.24)Numbers in parentheses correspond to the 95% confidence interval for the mean. Collections in Marsilac village, the transition zone and Cachoeira do Marsilac were performed from March 2015 to April 2017, while collections in Embura village were performed between May 2009 and June 2010 and between March 2015 and April of 2017. Collections in Evangelista de Souza were carried out from May 2009 to June 2010 Turning to the landscape metrics, Evangelista de Souza and Cachoeira do Marsilac had higher values for forest cover and lower values for areas that had undergone anthropogenic changes, forest fragments and total edge length than the other sites. Table [2](#Tab2){ref-type="table"} shows the landscape metrics for the areas in the vicinity of each collection point.Table 2Landscape metrics for each of the five sites in the Capivari-Monos EPASiteProportion of forest coverProportion of anthropogenic useNumber of forest fragmentsTotal edge length (km)1 km bufferEmbura village0.6170.375726.627Marsilac village0.6340.3661325.516Transition zone0.7090.2911227.948Cachoeira do Marsilac0.9200.064117.375Evangelista de Souza0.9150.080720.545500 m bufferEmbura village0.5960.40447.515Marsilac village0.5320.46867.644Transition zone0.5310.46955.940Cachoeira do Marsilac0.8830.11716.208Evangelista de Souza0.9880.01134.534The data were obtained for 500 m and 1 km buffers around the collection sites Among the mixed-effect models proposed to evaluate the relationship between landscape and *An. cruzii* abundance, the best fit was observed for the model that included the variable "forest cover 1 km" (∆AICc = 0, weight = 0.961), in which a positive predictive relationship was observed and an increase in forest cover led to an increase in *An. cruzii* abundance (Fig. [3](#Fig3){ref-type="fig"}). The second-best fit was for the model with the variable "total edge 500 m" (ΔAICc = 6.6, weight = 0.036), which had a negative predictive relationship with *An. cruzii* abundance. The null model showed the worst fit of all the proposed models (ΔAICc = 663.7, weight = \< 0.001) (Table [3](#Tab3){ref-type="table"}).Fig. 3Observed values (points) and predicted values (lines) for abundance of *Anopheles cruzii* per human/hour as a function of the proportion of forest cover within a 1 km buffer of each site. The black line represents the average prediction of the model for the fixed effect. The colored lines represent the variations in the intercept due to random effects (collection month and year) Table 3Models proposed to predict the effect of landscape variations on *Anopheles cruzii* abundance (mean number of mosquitoes collected per human/hour) and acrodendrophily (measured as the proportion of mosquitoes collected at ground level)Response variableExplanatory variableFixed effectRandom effect (standard deviation of the random-effect intercepts)AICc∆AICcWeightInterceptSlopeMonthYearMean *Anopheles cruzii* abundance per human/hour*Forest cover 1* *km−* *1.935 (0.507)5.508 (0.244)1.0400.7771391.500.961*Total edge 500 m7.786 (0.583)− 0.899 (0.046)1.1820.8791398.16.60.036Anthropogenic use 1 km3.487 (0.465)− 5.225 (0.232)1.0370.7851403.211.70.003Anthropogenic use 500 m3.395 (0.433)− 3.774 (0.173)1.0220.6991465.373.8\< 0.001Forest cover 500 m− 0.379 (0.457)3.775 (0173)1.0220.7001465.473.9\< 0.001Total edge 1 km5.695 (0.499)− 0.139 (0.007)0.9950.8471599.9208.5\< 0.001Forest fragments 500 m3.589 (0.486)− 0.302 (0.016)0.9940.8661627.9236.4\< 0.001Forest fragments 1 km3.326 (0.482)− 0.101 (0.006)0.9790.8631749.2357.8\< 0.001Null model2.645 (0.497)--1.0020.8982055.2663.7\< 0.001Proportion of *Anopheles cruzii* at ground level*Total edge 500* *m− 6.618 (1.429)0.795 (0.142)2.2872.050325.100.886*Forest cover 1 km2.552 (1.389)− 5.371 (1.045)2.3221.873331.05.90.047Anthropogenic use 500 m− 2.651 (1.138)3.710 (0.740)2.3081.985332.47.30.024Forest cover 500 m1.057 (1.269)− 3.709 (0.741)2.3081.984332.47.30.023Anthropogenic use 1 km− 2.721 (1.102)5.033 (1.008)2.3211.866332.77.60.020Total edge 1 km− 4.412 (1.242)0.115 (0.029)2.2661.801343.618.5\< 0.001Forest fragments 500 m− 2.398 (1.069)0.161 (0.073)2.2361.794354.128.9\< 0.001Null model− 1.940 (1.037)--2.1731.795356.631.5\< 0.001Forest fragments 1 km− 2.139 (1.053)0.034 (0.028)2.2001.787357.432.3\< 0.001The values of the models with the best fit are shown in italicsLandscape variables were measured for 500 m and 1 km buffers around the collection point. For each model the intercept and slope for the fixed effect, the standard deviation of the random-effect intercepts, the Akaike information criterion for small samples (AICc and ΔAICc) and Akaike weight are shown. The standard error of estimates is shown in brackets In terms of the relationship between landscape and *An. cruzii* acrodendrophily, the model with the best fit was the one with the variable "total edge 500 m" (ΔAICc = 0, weight = 0.886), in which there was a positive predictive relationship between edge length and *An. cruzii* frequency at ground level (Fig. [4](#Fig4){ref-type="fig"}). The variable "forest cover 1 km" also showed a good fit (ΔAICc = 5.9, weight = 0.047) and a negative predictive relationship with *An. cruzii* frequency at ground level. This model had the second-best fit of those proposed. Again, the null model had one of the worst fits (ΔAICc = 31.5, weight = \< 0.001); however, the difference in AICc between this model and the best model was not as high as the corresponding difference for the abundance models (Table [3](#Tab3){ref-type="table"}).Fig. 4Observed values (points) and predicted values (lines) for proportion of *Anopheles cruzii* in CDC traps at ground level as a function of the total edge length in kilometers within a 500 m buffer around each site. The black line represents the average prediction of the model for the fixed effect. The colored lines represent the variations in the intercept due to random effects (collection month and year) Discussion {#Sec8} ========== It was hypothesized in this study that factors associated with anthropogenic landscape changes may be responsible for the observed variation in *An. cruzii* abundance and acrodendrophily. The models with the best fits suggest that a loss of natural vegetation leads to a reduction in *An. cruzii* abundance, while an increase in the edge effect due to fragmentation and suppression of forest areas may favour greater activity by this species near ground level, which in turn may increase the contact rate between these mosquitoes and humans living on the edges of forest fragments where this species occurs. As observed in many studies, *An. cruzii* tends to be the dominant, or at least one of the most abundant, species of Culicidae in the humid forests of the coastal slopes of Southeastern and Southern Brazil, and its population density is directly related to the abundance of bromeliads \[[@CR16], [@CR20], [@CR23], [@CR48], [@CR49]\]. Recently, Chaves et al. \[[@CR50]\] studied the influence of landscape on the distribution and abundance of *An. cruzii* and *Anopheles bellator*. They found that landscape configuration and composition seem to play a significant role in the abundance of both species, the former being more abundant in dense forest areas and the latter more prevalent in *restinga* (sandy, salty soil close to the sea covered with characteristic herbaceous plants). They also found that neither of these species occurs in rural environments, where bromeliads tend to be sparse or absent. Dorvillé \[[@CR51]\] analysed several ecological studies on Culicidae in Southeastern Brazil and concluded that mosquitoes of the subgenus *Kerteszia* can be used as bioindicator species as they are highly susceptible to environmental degradation and their presence reflects a high degree of preservation. In studies between the years 2010 and 2013 in parks in built-up areas of the city of São Paulo, none of the specimens collected were from the *Kerteszia* subgenus although there were a significant number of bromeliads in isolated, preserved forest fragments in the study area \[[@CR52], [@CR53]\], another indicator of the susceptibility of this group to environmental changes. Previously, Ribeiro et al. \[[@CR35]\] had also observed this difference in *An. cruzii* abundance between preserved and degraded areas of the Capivari-Monos EPA (Evangelista de Souza and Embura village, respectively). There is a consensus among the various studies regarding the preference of *An. cruzii* for blood feeding in the tree canopy \[[@CR2], [@CR8], [@CR14]\]. Nevertheless, variations in the biting activity of this mosquito in the canopy and at ground level have been observed in different locations, and few studies have sought to understand the factors that lead to these variations \[[@CR20], [@CR22]--[@CR24], [@CR26], [@CR54]\]. Investigating the vertical stratification of mosquitoes in forest areas in the northeast of the state of São Paulo, Forattini et al. \[[@CR20]\] observed differences in *An. cruzii* acrodendrophily between the study sites and an increase in acrodendrophily in nighttime collections. The authors suggested that the proximity of one of the sites to areas inhabited by humans may have led to a small reduction in acrodendrophily and that the increase in acrodendrophily during the night may have been related to the greater number of animals, particularly birds, at rest in the tree canopy. In a study conducted between 1981 and 1982 in Serra dos Órgãos National Park in the state of Rio de Janeiro, Guimarães et al. \[[@CR22]\] also observed an increase in acrodendrophily in this species at night and a more equitable distribution of individuals between the canopy and ground level during daylight hours. In the same study, the authors observed that throughout the year the average relative humidity of the upper stratum of the forest was lower than in the lower stratum, which could partly explain the preference of *An. cruzii* for flying and feeding in the treetops as the condensation of water vapor on the body of these insects may lead to a reduction in flight activity. In contrast, Deane \[[@CR8]\] observed that in forest fragments in plateau regions of the Atlantic Forest farther inland, such as Cantareira State Park in São Paulo, *An. cruzii* showed marked acrodendrophilic behaviour, while in coastal mountains where the relative humidity is high, populations of this species seemed to feed more frequently near the ground. These observations suggest that local variations in *An. cruzii* acrodendrophily are influenced by environmental and microclimatic factors, genetic variations between populations in different regions and the possibility that this mosquito actually represents a cryptic species complex \[[@CR55]--[@CR57]\]. The observations of the present study suggest that the acrodendrophilic behaviour of *An. cruzii* varies with changes in the composition and configuration of the surrounding landscape and that there is an increase in the blood-feeding activity of this mosquito in the lower stratum of the forest in response to an increase in edge effect in areas with greater human activity following deforestation and fragmentation of the native forest cover in these places. The results of the present study should be viewed with some caution as the number of sampling points was small and the models for the frequency of this mosquito at ground level do not have a much better fit than the null model (unlike the models used to explain the abundance). However, two factors lend support to our findings: the fact that the increased presence of human hosts appears to increase mosquito activity near ground level, and the fact that the microclimatic variations caused by the edge effect may also lead to greater number of mosquitoes near the ground. The increase in human activity in areas on the edges of forest fragments or even inside the forest in itself represents an increase in the number of hosts available to the vector and to other mosquitoes in search of blood meal sources. It would, therefore, not be surprising to find that *An. cruzii* moves from the canopy to the ground more often in situations where humans and their domestic animals represent the nearest and most abundant sources of blood meals. Guimarães et al. \[[@CR58]\] studied the feeding habits of mosquitoes collected in the canopy and at ground level in the Serra dos Órgãos National Park and found that most species, including *An. cruzii*, exhibited opportunistic behaviour, feeding on humans and other animals used as bait. Chaves et al. \[[@CR59]\] reviewed several studies on mosquito feeding habits and used a null model of species co-occurrence to test whether mosquito feeding behaviour tended to be random (no preference for a particular species), segregated (certain mosquitoes feed only on a particular host) or aggregate (mosquitoes tend to feed primarily on a particular host). They found that in studies where data were collected at multiple sites there was a trend toward aggregate behaviour and concluded that contact between mosquitoes and hosts depends more on the availability of a given host than on an innate preference of the mosquito species. As for the effect of microclimatic variations, factors such as luminosity, wind speed, temperature and humidity are known to vary from the edge to the interior of forest fragments \[[@CR60]--[@CR62]\]. In a study in the Atlantic Forest, Magnago et al. \[[@CR63]\] showed that low humidity, high winds and higher temperatures are more common in edge environments than inside the forest. Based on this, a second hypothesis is that the reduction in humidity due to the edge effect in more fragmented areas may favour increased *An. cruzii* activity in the lower stratum, which agrees with the observations of Veloso et al. \[[@CR64]\] and Guimarães et al. \[[@CR65]\] regarding the effect of air humidity on the flight activity of *Kerteszia* mosquitoes. The main limitations of the present study include the small number of sites investigated and the fact that these are all in the same area, the Capivari-Monos EPA. In addition, it was not possible to observe the number of mosquitoes collected per human/hour in the tree canopy because Shannon traps could not be set up on suspended platforms with human collectors for logistical reasons. Clearly, it is not known whether the results that would have been obtained had it been possible to set Shannon traps up with collectors would differ from those actually obtained; however, there can be no doubt that the presence of human collectors at ground level and in the canopy would be more appropriate for the present study as CDC traps do not fully mimic the presence of humans or simian hosts. Although competing models were tested for two different landscape scales (500 m and 1 km), there was no prior evidence to believe that a larger landscape scale could better predict the variations in the abundance and acrodendrophily of *An. cruzii* than smaller, more local scale, or vice versa. Interestingly, the results suggest that a larger forest cover scale (1 km buffer) better predicts variations in the mosquito population abundance; while a smaller landscape scale (500 m buffer) better predicts variations in the acrodendrophily behaviour. However, it was not investigated how these variables would respond to larger or smaller landscape scales besides those measured. It has recently been shown that environmental variables measured at more local scales can have a great influence on spatial heterogeneity of the abundance of mosquitoes in forested urban environments \[[@CR66]\]. In a more local scale the presence and quantity of bromeliads is a factor directly related to the abundance of *An. cruzii* \[[@CR64]\]. Although this type of plant is quite common throughout the study area, the distribution and abundance of bromeliads at the sampling sites was not investigated, which may be considered one of the limitations of the present study. Nevertheless, such variations may be indirectly reflected in the abundance of adult mosquitoes that was lower in the more modified areas where bromeliads tend to be less abundant. In a study about the infectivity of *Anopheles* mosquitoes at the Capivari-Monos EPA, Duarte et al. \[[@CR11]\] found *An. cruzii* specimens naturally infected by both *P. vivax* and *P. malariae*. One of the areas where these were found was Embura village, where *An. cruzii* abundance was lower. Interestingly, species found in low abundance, such as *Anopheles triannulatus*, *Anopheles strodei* and *Anopheles lutzi*, were also found naturally infected by *P. vivax* and *P. malariae* in sites that had been subjected to anthropogenic change. As the role of these species of mosquitoes in malaria transmission in the Atlantic Forest is not yet known, other studies should be conducted in these areas to understand the vector--host transmission dynamics. Information on the prevalence of *Plasmodium* species in the *An. cruzii* populations collected in Cachoeira do Marsilac, the transition zone and Marsilac village was obtained from studies carried out in parallel with the present study. However, some of the laboratory work for these studies is still being carried out at the time of writing and data are therefore not yet available to compile with the data on abundance and acrodendrophily from the present study. This analysis should be done as part of a future study. In the past, bromeliad malaria was an endemic disease of great epidemiological importance in the Southeastern and Southern regions of Brazil. Although it is now under control as a result of the substantial effort made, it has not been totally eliminated as there are still transmission foci in various places. Nevertheless, many of these foci probably do not come to the attention of the health authorities because of the high proportion of asymptomatic and oligosymptomatic cases. Furthermore, the symptoms can be confused with those of other diseases \[[@CR29], [@CR67]--[@CR69]\]. Anthropogenic changes in the landscape and the consequent reduction in biodiversity are important factors in the emergence of malaria outbreaks in the Amazon region, where the phenomenon known as frontier malaria is now acknowledged to exist \[[@CR70]--[@CR72]\]. However, little is known about the effect of such changes on the dynamics of malaria transmission in the Atlantic Forest. Conclusion {#Sec9} ========== The data obtained in this study indicate that anthropogenic changes in the landscape lead to a reduction in the abundance of *An. cruzii* but can increase the contact rate between these mosquitoes and humans living at the edges of forest fragments where this species is found. Future studies should, therefore, seek to elucidate the effects of these landscape changes on the dynamics of *Plasmodium* transmission in the Atlantic Forest, which according to some studies includes the participation of simian hosts. The development of predictive models that seek to improve the understanding of how malaria vectors respond to changes in landscape composition and configuration can provide important information to assist planning and targeting of prevention and control actions. Additional file =============== {#Sec10} **Additional file 1.** Species and number of individuals collected in the Capivari-Monos Environmental Protection Area by collection. Collections made from March 2015 to April 2017. EPA : Environmental Protection Area CDC : Centers for Disease Control and Prevention AICc : Akaike Information Criterion corrected for small samples km : kilometre m : metre mm : millimetre °C : degrees celsius CO~2~ : carbon dioxide Study design: ARM-S, MTM, AMRCD, LFM. Field collections: ARM-S, WC-J, ROC. Data analysis and scientific writing: ARM-S. Scientific review: MTM, AMRCD, LFM, WC-J, ROC. All authors read and approved the final manuscript. Acknowledgements {#FPar1} ================ We would like to express our gratitude to the field and laboratory teams at the Superintendency for the Control of Endemic Diseases, São Paulo Zoonosis Control Center, and the School of Public Health, São Paulo University: João Carlos do Nascimento, Paulo Frugoli dos Santos, Luis Milton Bonafé, Antônio Waldomiro de Oliveira, Laércio Molinari, Gabriel Marcelino Neto, Luiz Sposito Jr, Renildo Souza Teixeira, Aristides Fernandez, Daniel Pagotto Vendrami, Gabriela Cristina de Carvalho, Ramon Wilk da Silva, Eduardo Evangelista de Souza and Amanda Alves Camargo. Competing interests {#FPar2} =================== The authors declare that they have no competing interests. Availability of data and materials {#FPar3} ================================== The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Consent for publication {#FPar4} ======================= Not applicable. Ethics approval and consent to participate {#FPar5} ========================================== Not applicable. Funding {#FPar6} ======= This work was funded by the São Paulo Research Foundation (FAPESP) Grants Nos. 2014/50444-5 and 2014/10919-4. ARM-S was supported by the São Paulo Research Foundation (FAPESP) Ref. No. 2015/18630-6. Publisher's Note {#FPar7} ================ Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

{#sp1 .369} {#sp2 .370}

Introduction ============ Climate change is altering the environments in which all organisms develop. Plant species can respond to these novel conditions through phenotypic plasticity, adapt through natural selection, or migrate to follow conditions to which they are adapted. The amount of variation in natural populations for traits that will be critical in future climates is generally unknown (Davis and Shaw [@b12]; Parmesan [@b35]) and understanding trait response, both phenotypic and ultimately genetic, will be critical for predicting how organisms will respond to novel abiotic conditions (Parmesan [@b35]). Predicting these responses to environmental changes will require an understanding of the environmentally induced variation in the phenotype of individuals (phenotypic plasticity) and the molecular mechanisms underlying those responses (Nicotra et al. [@b34]). Some portion of plastic variation may be adaptive, and some is likely to be neutral or even maladaptive (van Kleunen and Fischer [@b27]). Of particular interest when predicting the response of species to climate change is the portion of the reaction norm that reflects active, adaptive plasticity that leads to an increase in mean global fitness for the genotype (see Nicotra et al. [@b34] and references therein). Selection analyses, in which fitness is regressed against trait plasticity, provide a tool to assess the adaptive value of a plastic response (van Kleunen et al. [@b28]; Stinchcombe et al. [@b47]). Adaptive plasticity is predicted to evolve when a species is subjected to fine-scale environmental heterogeneity relevant within the life span of the organism and when conditions can be predicted based on environmental cues (Sultan et al. [@b48]; Valladares et al. [@b49], [@b50]; Herman et al. [@b16]). Thus, species that are distributed across strong environmental gradients present an ideal system in which to examine drivers and consequences of phenotypic variation, particularly in a climate change context. Considering predicted shifts in temperature and precipitation (IPCC [@b21]), alpine systems provide a unique opportunity to explore the importance of plasticity in response to global climate change. Alpine systems have steep temperature and water availability gradients associated with elevation, local topography, and aspect. Within-species variation in trait means has been observed in several alpine species. For example, specific leaf area (SLA) and leaf length decrease (Korner et al. [@b29]; Byars et al. [@b9]; Garibaldi et al. [@b15]; Zhang et al. [@b54]), height decreases (Wang et al. [@b53]; Hoffmann et al. [@b19]) and seed mass increases (see references in Segal [@b44]) with increasing elevation. Exploring the mechanisms that underlie such phenotypic patterns will provide a better understanding of the capacity of alpine plants to respond to future climates. Recent studies suggest that phenotypic plasticity can be mediated through epigenetic effects (Richards et al. [@b38]; Bossdorf et al. [@b4]; Scoville et al. [@b43]; Herrera and Bazaga [@b18]; Zhang et al. [@b55]; Herman et al. [@b16]). The most studied epigenetic effect is DNA methylation which has been shown to increase in variance in response to stressful conditions (Verhoeven et al. [@b52]; Dowen et al. [@b13]) and has known effects on ecologically important phenotypes (Johannes et al. [@b23]; Bossdorf et al. [@b4]; Zhang et al. [@b55]; Cortijo et al. [@b10]). Because epigenetic states can be thus altered, epigenetic effects could provide a rapid source of phenotypic variation without any change in genetic variation (Rapp and Wendel [@b37]; Bossdorf et al. [@b3]), thereby affecting the ability of populations to persist in the face of changing climate. Attempts to probe this connection between epigenetic mechanisms and phenotypic responses have thus far been limited. We examined the extent and correlates of adaptive phenotypic plasticity in response to growth temperature in half-sib lines collected across an elevation gradient for the alpine herb, *Wahlenbergia ceracea* Lothian (Campanulaceae). We asked whether seed that had developed at low, intermediate, and high elevation within the species\' natural range differed in genetic structure, mean trait values, or in plasticity in response to temperature. In addition, we assessed variation in DNA methylation between maternal plants and offspring and between offspring from different elevations grown under different temperature regimes to assess the presence of heritable or induced variation in epigenetic markers. The results provide striking evidence of differentiation in trait means and plasticity over small geographic distances and suggest that adaptive plasticity is associated with increased variation in DNA methylation. Material and Methods ==================== Study species and seed collection --------------------------------- *Wahlenbergia ceracea* Lothian (Campanulaceae) is an alpine short-lived perennial herb found in moist sites in high montane and alpine elevations in Australia (∽1600--2200 m). We collected mature seed capsules from five plants at each of three elevation ranges: low (1610--1625 m asl), medium (1750--1830 m asl), and high (1940--1975 m asl) in Kosciuszko National Park, NSW, Australia (−36.432, 148.338), relative to the natural distribution of the species in the park. Higher elevation sites are on average colder, have longer periods under snow and experience fewer extreme heat and freezing events (Briceño Rodriguez [@b7]). Plants were a minimum of 10 m apart and the total range between any two plants was 14 km (Appendix [A1](#a1){ref-type="app"}). A related study showed that plants were significantly shorter at higher elevations but showed few other morphological differences under field conditions at maturity (Segal et al. unpubl results). Seed were collected at the point of natural dispersal in 2011 to yield 15 half-sib maternal family lines. We found no correlation between elevation and seed mass (Segal [@b44]). Seedling growth conditions -------------------------- We planted three seeds from each of the 15 maternal lines into each of ten 50 × 123 mm (210 mL) tubes using commercial seed-raising mix (Debco Pty Ltd, Victoria, Australia) with micronutrients (December 20--21, 2011). Tubes were randomly assigned to one of five blocks so that each block contained two replicates from each line, one per temperature treatment. We grew five blocks each in two controlled temperature glasshouses to test the effect of a cool (mean ± SD of 20.6 ± 0.02°C/11.5 ± 0.01°C day/night temperature on a 12 h cycle) and a warm (mean ± SD of 29.8 ± 0.02°C/19.2 ± 0.02°C day/night) temperature regime on seedling growth and development. We arranged tubes randomly within blocks on the glasshouse bench and matched block locations between the adjacent warm and cool temperature glasshouse chambers to minimize differences in light and air circulation. Overall soil temperatures (measured with iButton (Maxim Integrated, San Jose, CA) loggers in soil trays) were ∽4°C higher in the warm than the cool glasshouse both day and night (Hoyle et al. [@b20]). The cool glasshouse roughly approximated mid-day early growing season soil temperatures in situ and the warm glasshouse temperature regime represented current summer high temperatures (Hoyle et al. [@b20]), although without the variation present in situ (see also Hoyle et al. [@b20]; Briceño et al. [@b8]). We recorded date of seedling emergence and pinched out all but the largest seedling in each tube before seedlings had begun to shade one another. We measured height (mm, to top of shoot apical meristem or base of bolt on flowering plants), rosette diameter (length across two opposite leaves at widest point), and number of true leaves 8, 11, and 14 weeks after sowing as well as 90 days after germination. Date of first flowering was recorded for each seedling that flowered in the first 20 weeks of the experiment after which time data were recorded monthly. We monitored total capsule production until plants naturally senesced. Hand pollination trials revealed no significant effect of hand pollination on either seed number per capsule or individual seed mass. Thus, we conclude that *W. ceracea* is autogamous and that capsule and seed production in the glasshouse is not likely to be limited by pollen (A2). AFLP genotyping and MS-AFLP epi-genotyping ------------------------------------------ We collected leaf tissue from the high-elevation and low-elevation maternal plants in the field, dried it in a plant press and froze at −20°C. From each of three or four offspring per maternal line grown under each of the controlled conditions, we collected and immediately froze leaf tissue (−20°C). For both generations, tissue was collected at the peak of the growing season and before senescence. We screened a total of five high-elevation and five low-elevation maternal plants and 66 offspring individuals for genetic variation using AFLP (*N* = 76). We used the Qiagen DNeasy Plant Mini kit (Qiagen, Valencia, CA) to perform duplicate DNA extractions from each sample and ran the duplicates through the AFLP protocol to ensure reliable scoring of fragments. The standard protocol suggested by Qiagen was used, eluting the DNA with water instead of TE in the final step. The AFLP protocol was based on standard methods with some modifications (Richards et al. [@b38], [@b39]). In the selective amplification, we multiplexed two primer pairs using 4 pmol of the 6-carboxy-fluorescein (6-FAM) fluorescently labeled *EcoRI*+AGC primer (/56-FAM/TACTGCGTACCAATTCAGC) with 4 pmol of the 4,7,2′,4′,5′,7′-hexachloro-6-carboxyfluorescein (HEX) fluorescently labeled EcoRI+ACG primer (/5HEX/TACTGCGTACCAATTCACG) and 25 pmol of the MseI+CAA in standard selective amplification reaction mixture and PCR conditions (Richards et al. [@b38]). We visually inspected the AFLP fragments using the open source program PEAKSCANNER v 1.0 (Applied Biosystems) and manually scored approximately 200 total loci for the HEX and 6FAM primer sets combined with a binary code, zero for band absent, one for band present. Of the 200 loci identified, nine were polymorphic. The repeatability of banding patterns across duplicate samples was assessed to determine whether banding patterns were consistent, only positions that could be reliably scored were included in the analysis. Throughout, we use "locus" to indicate a specific fragment size in the AFLP and MS-AFLP results. We use "haplotype" to indicate the collection of binary variable positions (dominant genotypes) for each individual at AFLP loci, and "epi-genotype" to indicate the collection of binary variable positions at MS-AFLP loci. We screened the 76 individuals for epigenetic variation with MS-AFLP using the same duplicate DNA extractions used for AFLP and the same selective bases on the two *EcoRI*+3 primers multiplexed with 25 pmol of the *HpaII/MspI*+TCAC primer (ATCATGAGTCCTGCTCGGTCAC). The MS-AFLP analysis used essentially the same protocol as the AFLP, but the *Mse*I enzyme was replaced with the same concentration of either *Msp*I or *Hpa*II, both of which are methylation sensitive, but vary in sensitivity. Both enzymes recognize and cleave CCGG sequences, but cleaving by *Hpa*II is blocked when the inner or outer C is methylated at both strands, while cleaving in *Msp*I is blocked when the outer cytosines are fully or hemi-methylated; cleaving in both enzymes is blocked when both cytosines are methylated. Four different types of variation can be scored (Salmon et al. [@b40]): Type I if both enzymes cut at the restriction site (no methylation), Type II if HpaII does not cut and MspI does cut (restriction site has a methylated internal C), Type III if HpaII does cut and MspI does not (restriction site has a methylated outer C), and Type IV if neither enzymes cuts (either both Cs are methylated or the restriction site has mutated). Recent work indicates that type II and III variation cannot be simply interpreted as CG versus CHG methylation, because what looks like CHG methylation is in fact often caused by differently methylated internal restriction sites nested within fragments (Fulnecek and Kovarik [@b14]). Therefore, we pooled data into two categories, methylated (Type II, Type III) or not methylated (Type I) restriction sites. We treated Type IV as missing data, because the methylation state cannot be specified (Salmon et al. [@b40]). We identified approximately 150 loci that could be reliably scored for *Msp*I and *Hpa*II for the HEX and 6FAM primer sets combined. Of these 150 loci, 39 were polymorphic. Statistical analysis -------------------- ### Phenotypic analyses We used REML models in Genstat (VSN International, 14th edn) with block, growth temperature (cool vs. warm chamber), and elevation (low, medium, or high) as fixed effects to analyze phenotypic data. We assessed maternal line effects by nesting families within elevation. Initial analyses included a continuous variable and a square term for elevation to account for nonlinear effects, but as this did not improve model fit, the term was not applied in the final model. Data were log transformed as needed to meet assumptions of normality. We calculated a modified Plasticity Index (PI) to compare cool and warm treatment values for each maternal line in each block (after Valladares et al. [@b51]). We calculated the plasticity index as (maximum − minimum)/maximum of the ln-transformed values for each pair of plants within a block\*maternal line combination (*n* = 5 pairs for each line). Pairs in which both individuals survived but neither set fruit (zero fitness), and pairs in which one plant died were excluded from the plasticity analysis (a total of 13 of the 75 pairs). Selection gradient analysis was performed to determine whether plasticity was adaptive, maladaptive, or neutral. The plasticity index was standardized to a mean of zero and standard deviation of one. Relative fitness was calculated as the mean fitness (capsule number) for a pair of plants divided by mean of all pairs and then log transformed (van Kleunen and Fischer [@b26]). The linear model included a term for the ln-transformed trait mean for each pair of plants (standardized as for PI), block, elevation, and maternal family nested within elevation. By pairing plants across blocks for our PI calculation, we were able to account for variation in plasticity within maternal line; conventional approaches that calculate plasticity by taking averages across plants for a given line within treatments cannot incorporate that variation. Preliminary analyses demonstrated strong interactions between measures of plasticity and elevation; therefore, plasticity was assessed as the interaction between elevation and PI for the trait. The partial regression coefficients for each elevation were used to assess slope and significance of the selection differentials. ### Genetic analysis: AFLP GENALEX version 6.41 (Peakall & Smouse [@b370]) was used to identify shared haplotypes among individuals and to determine the haplotype diversity (*h*-AFLP) for each sample. We compared *h*-AFLP between maternal plants and seedlings, and among four categories of seedlings: high-elevation mother grown cool, high-elevation mother in warm, low-elevation mother in cool, and low-elevation mother in warm. We also used GENALEX to calculate estimates of genetic differentiation between samples over all loci using an AMOVA framework. AMOVA was used first to compare variances among all individuals from either high or low elevation (Φ~ST~) and then to compare only seedlings between high and low elevation. Finally, we performed an AMOVA to compare variances among seedling treatments, both over all treatments and pairwise between treatments. For all AMOVA analyses, 9999 permutations were calculated to estimate statistical significance and the initial alpha = 0.05. ### Epigenetic analysis: MS-AFLP We analyzed epigenetic variation among individuals from the two elevations in the two temperature treatments. Statistical methods followed those in the genetic analysis. Given the small sample size for maternal plants and the differences in growing conditions, we appreciate that the epigenetic results for the mothers can only be compared loosely to the offspring but include these analyses as it is still interesting to assess the behavior of the markers. To further explore effects of experimental temperature treatment on MS-AFLP profiles, we compared the proportion of methylated loci (types II and III, see above) among all scorable loci (types I, II and III) between plants using a generalized linear model with a binomial distribution and logit link function (GENMOD procedure in SAS 9.2, The SAS Institute, Cary, NC). We tested effects of elevation, temperature treatment, maternal line (nested within elevation), temperature treatment x elevation, and temperature treatment x maternal line, on the proportion of methylated loci. To correct for overdispersion, standard errors were scaled using the Pearson chi-square (pscale option in GENMOD). Based on multivariate analysis of the MS-AFLP profiles, we tested for differences in multivariate dispersal between warm-grown and cool-grown plants using the PERMDISP program (Anderson et al. [@b1]; a multivariate analogue of the Levene\'s homogeneity of variances test). For this purpose, an epigenetic dissimilarity matrix was derived based on simple matching coefficients using the DISTANCE procedure in SAS 9.2. With this distance measure, shared methylations (type II or III fragments) and shared nonmethylations (type I fragments) at polymorphic MS-AFLP loci contribute equally to the similarity score between two individuals; both can capture relevant epigenetic information (Schulz et al. [@b42]). Mantel tests were used to test for correlations between epigenetic and phenotypic data. We made a trait-based Mahalanobis distance matrix for the experimental seedlings using all available measurements on plant height, rosette diameter, leaf number, flowering time, and capsule production (see section "seedling growth conditions"). Mahalanobis distances can handle variables that are correlated and/or measured at different scales by first extracting standardized principal components from the set of variables and calculating pairwise Euclidean distances based on these principal component scores (PRINCOMP and DISTANCE procedures in SAS 9.2). We ran mantel tests using Zt software (van de Peer [@b36]). Results ======= Effects of temperature on growth and reproductive traits -------------------------------------------------------- Seedlings in the cool temperature regime took longer to emerge than those in the warm regime; this pattern was not affected by the elevation at which the seed were developed (Appendix [A3](#a3){ref-type="app"}, *P* ≤ 0.0001). In addition, maternal lines differed in time to emergence and this variation was not dependent on growth temperature or elevation (Appendix [A4](#a4){ref-type="app"}, *P* ≤ 0.019). Not surprisingly, seedlings from the warm glasshouse, which had emerged earlier, were generally larger (taller with greater rosette diameter and more leaves) than those in the cool glasshouse when compared at constant times after sowing (for *P* values see Appendix [A4](#a4){ref-type="app"}, Fig.[A](#fig01){ref-type="fig"}--[C](#fig01){ref-type="fig"}). In addition, warm-grown seedlings showed higher relative growth increments in the juvenile phase, although the extent of this effect varied among blocks (Appendix [A4](#a4){ref-type="app"}). Even when compared at a standard time postgermination (90 days), warm-grown plants were on average taller with broader rosettes (Appendix [A4](#a4){ref-type="app"}, *P* ≤ 0.002, 0.003, respectively). With the exception of height growth increment at 90 days, there was no significant variation among maternal lines in growth parameters. {#fig01} Plants in the warm regime also flowered earlier and produced more capsules than those in the cool room (Fig.[D](#fig01){ref-type="fig"}, *P* ≤ 0.001, Appendix [A3](#a3){ref-type="app"}). In addition, capsules were larger under warm conditions (Fig.[E](#fig01){ref-type="fig"}, *P* ≤ 0.004, Appendix [A3](#a3){ref-type="app"}). When we estimated total mass of capsules from individual capsule mass and capsule number (for those plants on which mass was determined, see Appendix [A2](#a2){ref-type="app"}), we found that total capsule mass was also greater at warm temperatures (Fig.[F](#fig01){ref-type="fig"}, *P* ≤ 0.001, Appendix [A3](#a4){ref-type="app"}). More striking, however, was the elevation effects: seedlings from seed that had developed at lower elevations grew taller had greater rosette diameter and had more leaves than seedlings grown from seed that had developed at high elevations (Fig.[A](#fig01){ref-type="fig"}--[C](#fig01){ref-type="fig"}, Appendix [A4](#a4){ref-type="app"}). Seedlings from mid-elevations were generally not significantly different from high-elevation seedlings. The effect of elevation was significant for all growth measurements at each of the three measurement times except for rosette diameter at 14 weeks (Appendix [A4](#a4){ref-type="app"}). Note, however, that the elevation effects were not significant at 90 days postgermination, nor were they seen in the relative growth increments. Plants of low-elevation origin also produced more capsules than those from higher elevations (Appendix [A3](#a3){ref-type="app"} you, *P* ≤ 0.004, Fig.[1D](#fig01){ref-type="fig"}), but individual capsules were larger on plants from high-elevation families (Fig.[1E](#fig01){ref-type="fig"}), and so on average, seedlings from low-elevation families had lower total capsule mass (Fig[1F](#fig01){ref-type="fig"}). Selection gradient analysis of plasticity ----------------------------------------- Selection gradient analysis demonstrated a pattern of adaptive plasticity that was in accord with the previous results: In general, the more plastic low-elevation families also showed adaptive plasticity, whereas the less plastic high-elevation families were more likely to exhibit negative selection differentials for plasticity (Table[1](#tbl1){ref-type="table"}, Fig.[2A](#fig02){ref-type="fig"}). Mid-elevation families were intermediate in trait values and plasticity measures, and for these, plasticity was generally neutral. This pattern was observed to varying extents in several traits. Notably, one trait showed negative selection differentials for plasticity at all elevations: flowering time. Earlier flowering was associated with greater capsule production in all treatments (Fig.[2B](#fig02){ref-type="fig"}, Table[1](#tbl1){ref-type="table"}). ###### Selection gradient analysis for trait means and plasticity. (A) ANOVA table for analysis of effects of trait means and plasticity on relative fitness. Only traits with significant selection gradients for plasticity are included. Note that the covariate mean was also significant for height at 11 and 14 weeks and 90 days, leaf number at 90 days, and diameter at 8 and 14 weeks, but plasticity index was not. (B) Selection differentials (β) on plasticity at each elevation and across elevations. Underline text and negative numbers indicate negative selection differentials (costs); bold text indicates significant slopes (differentials, at *P* \< 0.05). Figures in italics are significant at *P* = 0.1 df Height 8 weeks Juv. height growth Rosette diam 90 days Leaf no. at 11 weeks Leaf no. at 14 weeks Days to flowering ---------------------------------- ---- -------------------------------------------------- ------------------------------------------------- -------------------------------------------------- -------------------------------------------------- ------------------------------------------------ ----------------------------------------------- \(A\) Analysis of variance table  Trait mean 1 **\<0.001** 0.156 \<0.001 **\<0.001** **\<0.001** **0.007**  Block (df = 4) 4 **0.027** 0.341 0.646 0.435 0.682 0.086  Elevation 2 0.565 0.455 0.67 0.651 0.818 0.267  Elevation. family 12 0.340 0.292 0.559 0.311 0.482 0.244  Trait *P* × elevation 3 **0.015** *0.073* 0.064 *0.057* 0.103 **0.010** \(B\) Selection differential, β  Mean **0.311**[**\*\***](#tf1-1){ref-type="table-fn"} −0.050 **0.515**[**\*\***](#tf1-1){ref-type="table-fn"} **0.332**[**\***](#tf1-1){ref-type="table-fn"} *0.268* −**0.281**[\*](#tf1-1){ref-type="table-fn"}  High −**0.498**[**\***](#tf1-1){ref-type="table-fn"} −**0.390**[**\***](#tf1-1){ref-type="table-fn"} −0.118 −0.081 0.006 −**0.559**[\*\*](#tf1-1){ref-type="table-fn"}  Medium 0.153 −0.157 0.203 −0.107 0.311 −0.196  Low 0.393 0.278 **0.502**[**\***](#tf1-1){ref-type="table-fn"} **0.452**[**\*\***](#tf1-1){ref-type="table-fn"} **0.536**[**\***](#tf1-1){ref-type="table-fn"} −0.194 \**P* \< 0.05, \*\**P* \< 0.01. {#fig02} Genetic diversity and structure ------------------------------- There were only nine variable positions among individuals from approximately 200 loci across the *Mse-I* AFLP selective PCR products, which formed 20 haplotypes. Haplotype diversity (*h*-AFLP) was similar between mothers and seedlings (*h*-AFLP = 0.019), while seedlings from low-elevation mothers had slightly, but significantly higher diversity than seedlings from high-elevation mothers (*P *≤* *0.05, Fig.[3](#fig03){ref-type="fig"}, Appendix [A4](#a4){ref-type="app"}). Likewise, there was little, but significant, genetic differentiation between individuals from low and high elevation among all individuals (Φ~ST~ = 0.027, *P *=* *0.047) and among only seedlings (Φ~ST~ = 0.038, *P *=* *0.027). There was no significant genetic differentiation among seedling treatments (Φ~ST~ = 0.022, *P *=* *0.133) or between any pairwise comparison of seedling treatments. {#fig03} Differences in epi-genotype between generations and as a function of elevation and growth temperature ----------------------------------------------------------------------------------------------------- Our analysis revealed more epigenetic than genetic diversity (more than four times as many variable positions), and the offspring had more epigenetic diversity than maternal plants. Thirty-nine variable positions were detected among 150 loci observed, which formed 62 epi-genotypes. There was no significant epigenetic differentiation between individuals from low and high elevation among all individuals (Φ~ST~ = 0.011, *P *=* *0.09) or among only seedlings (Φ~ST~ = 0.013, *P *=* *0.082). Nor was there significant epigenetic differentiation among seedling treatments (Φ~ST~ = 0.0001, *P *=* *0.475) or between any pairwise comparison of seedling treatments. In addition, there was no significant differentiation between maternal plants from low- and high-elevation sites, but note that sample size was very small for this comparison (*n* = 5 from each elevation, *F*~st~ 0.01, *P* = 0.05). Haplotype diversity for epi-genotypes (*h*-MS-AFLP), however, was higher in seedlings (*h*-MS-AFLP = 0.115) compared to maternal plants (*h*-MS-AFLP = 0.094). High-elevation maternal plants had slightly higher diversity (*h*-MS-AFLP = 0.099) compared to low-elevation maternal plants (*h*-MS-AFLP = 0.089). Notably, there was also significantly greater epigenetic diversity in seedlings from low-elevation origin when grown at warm temperatures (*h*-MS-AFLP = 0.158) compared to low-elevation maternal plants, the cool-grown low-elevation seedlings (*h*-MS-AFLP = 0.109), or the high-elevation origin maternal plants and seedlings (*h*-MS-AFLP = 0.098 & 0.093; Fig.[3](#fig03){ref-type="fig"}). Within the high-elevation seedlings, there was slightly more epigenetic diversity in the warm grown than the cool-grown plants, but the difference was not as strong as for low-elevation plants (Fig.[3](#fig03){ref-type="fig"}). Consistent with the difference in epigenetic haplotype diversity between warm-grown and cool-grown plants from low elevations, multivariate dispersal in MS-AFLP profiles based on pairwise distances was significantly larger within the low-elevation warm treatment than in the low-elevation cool treatment (average distance to centroid: 0.20 for cool and 0.27 for warm treatments, respectively; *P* = 0.037). We did not see this difference in multivariate dispersal between warm and cool treatments in the high-elevation seedlings (*P* = 0.87). In summary, although we did not detect differentiation (analogous to mean position in multivariate space) in response to elevation or treatment, we did find a difference in the amount of epigenetic variance within-groups in response to treatment. Mantel tests detected no correlations between phenotypic variation and epigenetic variation, either for high- (*n* = 31 plants from warm and cool treatments, *r* = −0.04, *P* = 0.40) or low-elevation plants (*n* = 33, *r* = 0.03, *P* = 0.42). These MS-AFLP--phenotype correlation tests were not appreciably affected by first accounting for possible correlations between MS-AFLP variation and temperature treatments (using partial mantel tests that test the MS-AFLP--phenotype correlation after controlling for a design matrix that codes plants from different treatments as distance = 1 and plants from the same treatment as distance = 0; data not shown). This suggests that global phenotypic variation and global MS-AFLP variation do not show consistent association in the experimental data, either within temperature environments or in the response to the temperature difference. Discussion ========== We set out to determine whether the distribution of adaptive plasticity in response to warming conditions differed depending on the elevation at which seed had developed in the alpine herb *W. ceracea*. Further, we assessed genetic and epigenetic differentiation between elevations and asked whether parents or offspring differed in patterns of DNA methylation depending on site of origin or experimental temperature regime. Warming did elicit a plastic response, accelerating seedling emergence and growth and leading to a shorter time to flowering and larger, more numerous capsules. More strikingly, although the maternal plants from which seed were collected were distributed over a very small geographic range, plants were differentiated at AFLP loci and the offspring showed evidence of significant trait differentiation along an elevation gradient. In addition to the genetic differentiation, plants grown from seed developed at lower elevation were larger, produced more seed capsules, and generally showed stronger plastic responses to temperature than those grown from high-elevation plants. The patterns of adaptive plasticity reflect our results for differentiation in plasticity: Plasticity was more likely to be selected for in the more plastic low-elevation seedlings. Finally, seedlings from low-elevation origin also showed the greatest propensity for changes in epigenetic marks in response to growth temperature. These results are consistent with the hypothesis that extreme but less variable conditions at high-elevation sites have led to canalization of growth traits, potentially via suppression of epigenetic change. Plants in lower elevation sites, in contrast, are perhaps likely to be exposed to more frequent extreme heat/cold events and have greater probability of encountering good conditions (e.g., periodic long, warm seasons, Briceño et al. [@b8]). Thus, the ability to respond to warming temperatures with vigorous growth has obvious value for low elevation. These results raise several interesting questions: How common is such small-scale differentiation in the distribution of adaptive plasticity, or the propensity for changes in methylation pattern? And, what does the association between epigenetic variation and adaptive plasticity tell us about the mechanisms underlying expression of phenotypic plasticity? Similar results of small-scale variation in selection for plasticity have been indicated for plasticity in leaf length and rosette circumference in alpine *Poa hiemata*, as indicated by patterns of co-gradient selection (Byars et al. [@b9]). Also, an altitudinal pattern of plant development persisted in a common garden environment in *Stylidium armeria* (Hoffmann et al. [@b19]), although this pattern was weaker than in the field, suggesting phenotypic plasticity in combination with local adaptation may contribute to survival of *Stylidium* in the field. Previous work has shown changes in the adaptive value of phenotypic plasticity depending on growth conditions or between sister species or disjunct populations (Jacobs and Latimer [@b22]). However, the present study is the first we know of to demonstrate contrasting selection gradients on plasticity over such a limited geographic range. In addition to being more plastic, the low-elevation plants showed a greater propensity to alter epigenetic signature in response to warming. Does this result indicate epigenetics-mediated adaptive phenotypic plasticity that might contribute to a bet-hedging strategy (Herman et al. [@b16])? Or might the epigenetic result be an artifact of the genetic differentiation (small though it was) between low- and high-elevation plants, as has been found in previous work (Joseph and Moritz [@b24]; Herrera and Bazaga [@b17]; Lira-Medeiros et al. [@b31])? Because the epigenetic difference was only apparent at high temperature, we think it is not simply a correlate of genetic differentiation. Although we did not find a majority response at the same loci (i.e., overall genome-wide pattern of differentiation), we did find a general tendency to change across several loci, which differed among maternal lines and elevations. Thus, we argue the evidence is more in favor of a link between epigenetic expression and the adaptive plastic response. To some extent, the ambiguity may reflect limitations of our method. The different maternal lines are also different genotypes and therefore different fragments across the individuals could be related to similar function. The MS-AFLP approach emphasized genome-wide patterns of variation, while epigenetically mediated functional responses may be restricted to a few specific loci. While this is a known limitation of the AFLP protocol, the approach does allow us to examine epigenetic response in this nonmodel plant for which the genome has not been sequenced (Schrey et al. [@b41]). Other studies examining epigenetic signatures have shown increases in variance in response to exposure to different environmental factors (Verhoeven et al. [@b52]; Dowen et al. [@b13]), with known effects on ecologically important phenotypes (Cubas et al. [@b11]; Johannes et al. [@b23]; Bossdorf et al. [@b4]; Zhang et al. [@b55]; Cortijo et al. [@b10]). Multigeneration experiments have shown that parental exposure to biotic or abiotic stresses resulted in modified DNA methylation in unexposed offspring (Boyko et al. [@b5]; Verhoeven et al. [@b52]). In dandelion, MS-AFLP showed that plants with identical genotypes exposed to different stresses had up to 30% change in polymorphic methylation sensitive markers compared to controls (Verhoeven et al. [@b52]). Bilichak et al. ([@b2]) showed that progeny of plants exposed to salt stress were globally hypomethylated, but hypermethylated at histone lysine methyltransferase genes. In response to water stress, Juenger et al. ([@b25]) found increased expression of several genes related to chromatin or epigenetic regulation. In our data, the larger variation in the low-elevation warm-grown seedlings may be due either to induced random modifications or may arise because different genotypes respond differently to the same environment. We went a step further than just documenting changes in epigenetic marks in that we examined correlations between these changes, changes in ecologically important traits, and adaptive plasticity therein. While our analysis is limited in scope, both in terms of sample size and the number of markers identified in the epigenetic analysis, we found that the pattern of adaptive plastic responses and propensity for epigenetic change occur within the same individuals (low-elevation progeny, especially when grown under warm conditions). We note, however, that our data did not reveal evidence for direct correlations between MS-AFLP variation and trait variation. Thus, based on these data the functional explanation of the MS-AFLP response remains unclear. The results here indicate adaptive plasticity in a range of complex quantitative traits, each of which is likely to be controlled by a range of genetic pathways, some potentially shared. The question of what the molecular mechanisms underlying plasticity might be is an old one, and only recently have we begun to unravel the answers. For some systems, the pathways of signal perception and response are well understood, for example, flowering time in *Arabidopsis thaliana* (Mouradov et al. [@b32]; Simpson and Dean [@b45]), but these are the minority. MS-AFLPs are a cost effective way to obtain data on methylation marks in many individuals and in nonmodel species and provide genome-wide patterns at anonymous loci. However, truly understanding the role of epigenetics requires a more powerful coverage of the genome and characterization of the behavior of specific genes or regulatory elements (e.g., *A. thaliana*, Slotkin and Martienssen [@b46]; Lippman et al. [@b30]). Genetic screening using AFLPs is rapidly being replaced by genotyping by sequencing (GBS) and restriction site-associated DNA sequencing (RAD-seq) approaches in nonmodel species (Narum et al. [@b33]). As has happened in model species, eventually these approaches will expand to DNA methylation profiling (e.g., reduced representation bisulphite sequencing or RRBS, Boyle et al. [@b6]), which will provide a powerful tool to explore how methylation patterns behave across different genomic regions in natural environments and will allow for more fine-scale resolution of sequence polymorphisms that we may have missed using AFLP markers. Conclusion ========== In conclusion, our study demonstrates significant differentiation in adaptive plasticity and in patterns of epigenetic expression within a species over a small geographic range. This variation may impact on the potential of the species to tolerate a warming climate. The AFLP analysis indicated only very little differentiation between the low- and high-elevation plants, suggesting that gene flow is quite high across the species range, however, the epigenetic and plasticity results indicate that some genotypes may have a higher potential to respond to stressful conditions which could be critical to surviving future climates. We thank Paulina Lobos Catalan and Chong Ren Ong for assistance with plant care, data-collection and with the hand pollination assays. This work was funded by an Australian Research Council grants to ABN, GLH and colleagues (LP0991593) and to ABN, CLR and colleagues (DP120100945). Data Accessibility ================== MS-AFLP and phenotype data: DRYAD entry doi:[10.5061/dryad.73r1b](10.5061/dryad.73r1b). Conflict of Interest ==================== None declared. {#fig04} Assessing Autogamy in *Wahlenbergia ceracea* To determine whether capsule production in the glasshouse was a good indicator of fitness, we assessed whether *Wahlenbergia ceracea* is autogamous (capable of self-pollination). Some *Wahlenbergia* species are autogamous (Mouradov et al. [@b32]; Simpson and Dean [@b45]); we assessed whether this is the case in *W. ceracea* by conducting a hand pollination experiment on a subset of plants. Hand pollination of flowers was conducted on no more than half the flowers on a given plant when the stigma was exposed and glossy in appearance. Mature pollen was moved from the subtending anthers to the stigma using a toothpick. Seed mass and number were assessed for capsules from 32 plants comparing hand pollinated and open capsules from a given plant (14 low, 10 mid, and eight high elevation, 14 of which came from cool and 18 from the warm glasshouse, respectively). On an additional 22 plants total seed mass (but not number) of hand pollinated versus open-pollinated capsules was assessed (eight low, two mid, and 12 high elevation, 14 of these came from cool and eight from the warm glasshouse, respectively). An Analysis of Variance for Unbalanced designs (growth temperature, elevation and pollination treatment and all interactions included as fixed terms) revealed no significant effect of hand pollination on either seed number per capsule or individual seed mass, regardless of inclusion or otherwise of the growth temperature or elevation terms (results not shown). Thus, we conclude that *W. ceracea* is autogamous and that capsule and seed production in the glasshouse is not likely to be limited by pollen. Hand- and open-pollinated capsules were therefore combined for all subsequent analyses and total number of capsules produced on each plant was used as a proxy for fitness. ###### Probabilities from REML analysis of emergence and reproductive traits. Values significant at *P* \< 0.05 are shown in bold Source n.d.f. Days to emergence (ln) Flowering date Capsule number (ln) Individual capsule mass Mg seed/capsule Total capsule mass ---------------------------------------- -------- ------------------------ ---------------- --------------------- ------------------------- ----------------- -------------------- Growth Temperature 1 **\<0.001** **\<0.001** **\<0.001** **0.004** 0.898 **\<0.001** Block 4 0.375 0.116 0.156 0.644 0.181 0.153 Growth Temperature × Block 4 0.137 0.174 **0.014** 0.318 0.122 **0.001** Elevation 2 0.535 0.509 **0.004** **\<0.001** 0.53 0.064 Elevation × Family line 12 **0.019** 0.128 **0.031** 0.573 0.483 **0.045** Growth Temperature × Elevation 2 0.249 0.229 0.551 0.017 0.336 0.612 Growth Temperature × Block × Elevation 12 0.08 0.629 0.132 0.261 0.059 0.536 ###### Probability values from REML analysis of growth parameters, (A) height, (B) diameter of rosette, and (C) leaf number at 8, 11, and 14 weeks after the first seedlings had emerged and at a standard development time: 90 days. Height, leaf number, and diameter data were log transformed. Juvenile growth increment is calculated from values at 8 and 11 weeks, mature growth increment from 11 and 14 weeks. The increments on mature growth for diameter and leaves required (log) transformation. Bolded values are significant at *P* = 0.05 8 weeks 11 weeks 14 weeks Juvenile growth increment Mature growth increment 90 days ------------------------------- ---- ------------- ------------- ------------- --------------------------- ------------------------- ------------- \(A\) Height  Growth Temperature 1 **\<0.001** **\<0.001** **\<0.001** **\<0.001** **\<0.001** **\<0.001**  Block 4 0.307 0.556 0.203 0.600 0.021 0.094  Growth Temperature × Block 4 **0.001** **0.001** **0.023** **0.054** 0.092 0.074  Elevation 2 **0.033** **0.009** **0.034** 0.225 0.303 0.074  Elevation × Family line 12 0.775 0.137 0.157 0.093 0.795 **0.028**  Growth Temp. × Elev. 2 0.179 0.572 0.475 0.465 0.730 0.414  Growth Temp. × Block × Elev. 12 0.787 0.280 0.587 0.158 0.317 0.268 \(B\) Rosette Diameter  Growth Temperature 1 0.063 **0.013** **0.005** 0.133 0.308 **0.003**  Block 4 0.460 0.344 0.222 0.405 0.659 0.151  Growth Temperature × Block 4 **0.001** 0.093 0.169 **0.009** 0.310 0.070  Elevation 2 **0.047** **0.036** 0.107 0.810 0.513 0.152  Elevation × Family line 12 0.211 0.162 0.114 0.417 0.065 0.121  Growth Temp. × Elev. 2 0.561 0.745 0.884 0.909 0.362 0.690  Growth Temp. × Block × Elev. 12 0.540 0.549 0.765 0.339 0.869 0.904 \(C\) Leaf number  Growth Temperature 1 **0.006** **0.011** 0.075 0.367 0.091 0.161  Block 4 0.736 0.427 0.12 0.069 **0.028** **0.049**  Growth  Temperature × Block 4 **\<0.001** **0.071** **0.06** **0.045** 0.543 0.118  Elevation 2 **0.007** **0.011** **0.007** 0.496 0.073 0.076  Elevation × Family line 12 0.657 0.211 0.203 0.077 0.706 0.253  Growth Temp. × Elev. 2 0.666 0.906 0.646 0.893 0.387 0.772  Growth Temp. × Block × Elev. 12 0.617 0.59 0.627 0.386 0.359 0.658 ###### Effects of elevation, maternal line, and treatment on the proportion of methylated MS-AFLP loci. *P* values in the Type III generalized linear model analysis are from Likelihood ratio tests. Bolded values are significant at *P* = 0.05 Source df Chi-square *P* value ----------------------------------------- ---- ------------ ----------- Elevation 1 0.30 0.583 Maternal line (Elevation) 8 5.58 0.694 Temperature 1 2.75 0.097 Temperature × Elevation 1 2.17 0.140 Temperature × Maternal line (Elevation) 8 16.23 **0.039** [^1]: **Funding Information** This work was funded by an Australian Research Council grants to ABN, GLH and colleagues (LP0991593) and to ABN, CLR and colleagues (DP120100945).

1. Introduction {#sec1-materials-11-00372} =============== Today, food technology is constantly evolving in response to different challenges. The changes in consumer demands and the necessity for the production of safe and high-quality foods are responsible for the innovation and improvement of already established food processes. In this sense, the introduction of new technologies could lead to a reduction of the processing time or an improvement in operating conditions, thereby decreasing both environmental and financial costs. Plasma treatments cause several chemical and physical changes on the plasma-polymer interface, which improve the surface properties \[[@B1-materials-11-00372],[@B2-materials-11-00372],[@B3-materials-11-00372],[@B4-materials-11-00372],[@B5-materials-11-00372],[@B6-materials-11-00372]\]. Plasma-induced effects on the polymer surface are nowadays exploited in surface functionalization of the packaging polymers for promoting adhesion or sometimes anti-adhesion \[[@B7-materials-11-00372]\], enhanced printability \[[@B8-materials-11-00372]\], sealability \[[@B9-materials-11-00372]\], assuring anti-mist properties, improving the polymer's resistance to mechanical failure \[[@B1-materials-11-00372]\], and adhesion of antibacterial coatings \[[@B10-materials-11-00372],[@B11-materials-11-00372],[@B12-materials-11-00372],[@B13-materials-11-00372],[@B14-materials-11-00372]\]. Polypropylene (PP) is an important commercial polymer which is often used for producing package films \[[@B15-materials-11-00372]\], because of its low cost and good thermal stability. Extruded PP film is amorphous, while the crystallization can be achieved by two-way stretching (monoaxially or biaxially orientated films) at elevated temperatures. Biaxial orientation (BO) slightly improves the silky structure of the film and significantly reduces turbidity, enhancing the barrier properties and flexural toughness at a low temperature. Biaxially oriented polypropylene (BOPP) film is often coated with an additional polymer layer to improve its mechanical properties or barrier properties against gases and moisture. If BOPP is used for food packaging applications, it is often coated with an acrylic layer or poly(vinylidene chloride) (PVDC). Acrylic coating (Ac) is durable, flexible, and resistant to degradation caused by ultraviolet rays \[[@B16-materials-11-00372]\]. If the PP foil is coated with a PVDC layer, this topcoat enhances PP barrier properties against water vapor and gases. Excellent protective properties of this layer make PP foil suitable for the packaging of confectionery products which require barrier protection from moisture \[[@B17-materials-11-00372]\]. As mentioned above, plasma can improve surface properties of polymers such as wettability and surface functionalization, and consequently also adhesion properties. This may be important for coating the food-packaging foils with antibacterial layers \[[@B18-materials-11-00372]\]. Many authors have investigated treatment of a pure PP foil rather than industrial-grade foils, which are often covered with an ultra-thin film of a protective coating. A reason for this might be unknown details regarding the composition and structure of the coating, let alone the method applied for deposition of the coating. Pandirayaj et al. \[[@B19-materials-11-00372]\] used a low pressure weakly ionized plasma created by a DC glow discharge to improve the wettability of the PP foil. The water contact angle dropped from the original value of 98° down to 58° upon treatment for 10 min. Similar results were reported by Choi et al. \[[@B20-materials-11-00372]\], who obtained 60° in low-pressure oxygen DC plasma. Additionally, Morent et al. \[[@B21-materials-11-00372]\] obtained a water contact angle of 60°, although he used an extremely weak plasma at the discharge power of solely 1.4 W and moderate pressure of 5 kPa. A dielectric barrier discharge (DBD) was applied. Leroux et al. \[[@B22-materials-11-00372]\] obtained the contact angle of 64° using plasma created in air at atmospheric pressure by classical DBD and a treatment speed of 2 m/min. Lower water contact angles were observed by some other authors. Aguiar et al. \[[@B23-materials-11-00372]\] achieved a water contact angle of 40° (initially 110°) on PP surface treated in oxygen plasma at 700 W and 6.7 Pa. Vishnuvarthanan et al. \[[@B24-materials-11-00372]\] observed that the contact angle depended on the discharge power (7.2--29.6 W) and treatment time (0--300 s). The lowest water contact angle was \~44°; however, the initial angle was just 74.5°, which indicates that the initial surface was probably already contaminated with surface impurities. Mirabedini et al. \[[@B25-materials-11-00372]\] obtained a minimal contact angle of 34.4° in RF oxygen plasma at 50 W and 0.35 × 10^5^ Pa. However, Wanke et al. \[[@B26-materials-11-00372]\] managed to achieve only 24° (initially 98°) at 15 min of treatment. Unlike other authors, who observed a decreasing water contact angle with increasing treatment time until reaching saturation, he observed that at long treatment times (after 15 min of treatment), the contact angle increased to 53°. The reason that some authors obtained such low contact angles can be associated with polymer overtreatment leading to the formation of low-molecular weight fragments (LMWOM) because of polymer degradation \[[@B27-materials-11-00372]\]. The main papers and key results are summarized in [Table 1](#materials-11-00372-t001){ref-type="table"}. Although oxygen plasma treatment causes beneficial effects such as improved wettability, it also causes other modifications of the surface and sub-surface layer which may not be tolerated. Oxygen plasma is rich in different reactive species and represents a source of ultraviolet (UV) radiation. The reactive gaseous species that interact with a polymer sample include positively charged molecular and atomic ions, neutral atoms in the ground and metastable states, and neutral molecules in both "a" and "b" metastable states and ozone. The major UV radiation occurs at the wavelength of 130 nm due to the transition from a highly excited 2s^2^2p^3^(^4^S°)3s^3^S° state to the ground state (2s^2^2p^4\ 3^P). The photon energy for this transition is 9.52 eV. The penetration depth of such UV radiation in a polymer material is around a micrometer \[[@B28-materials-11-00372]\]. The energetic photons cause bond scission and thus modification of the polymer properties well below its surface. Furthermore, there is always some water vapor in a low-pressure plasma reactor. The vapor is the major constitute of the residual atmosphere and is also formed due to chemical etching of the polymer upon oxygen plasma treatment. The water molecules dissociate under plasma conditions and the resulting OH and H radicals are excited upon inelastic collisions with energetic electrons. The excited states de-excite to the corresponding ground states by ration in the UV range: Lyman hydrogen series in the vacuum UV range and OH band of bandhead at 309 nm. All this radiation causes bond scission in the polymer film of a thickness of the order of several µm. The reactive species interact with dangling bonds on the polymer surface, causing the formation of LMWOM that are often volatile. Therefore, rather extensive etching is observed upon the treatment of a polymer material with oxygen plasma \[[@B29-materials-11-00372]\]. In fact, precise measurements of the oxidation rate for the same polymer exposed to oxygen plasma and only neutral O-atoms at the same O-atom flux on the sample surface showed a two orders of magnitude higher etching rate for the case where synergistic effects of radiation and reactive species were effective \[[@B30-materials-11-00372],[@B31-materials-11-00372]\]. Such synergies should therefore be avoided if functionalization of the polymer surface with oxygen functional groups is the goal. The aim of this research was to examine the effect of surface oxidation of commercial PP foils used for food packaging. Such foils are covered with a very thin acrylic coating. Unlike other authors, neutral reactive particles from late afterglow were used instead of gaseous plasma, because glowing plasma always causes the etching of polymers and the acrylic coating could have been removed by direct exposure to oxygen plasma \[[@B30-materials-11-00372]\]. Furthermore, in afterglow, a density of oxygen species interacting with the polymer can also be precisely determined. This allowed determination of the minimal oxygen atom fluence necessary for saturation of the surface with polar functional groups and thus optimal wettability at a minimal treatment time. 2. Materials and Methods {#sec2-materials-11-00372} ======================== 2.1. Materials {#sec2dot1-materials-11-00372} -------------- Biaxially oriented polypropylene (PP) films (Bicor 32MB777, ExxonMobil, Antwerp, Belgium) were used in the experiments. One side of the film had an acrylic acid coating and the other side was coated with a thin film of poly(vinylidene chloride) (PVDC), which means that plasma interacted with the coating and not with the PP substrate. Only the acrylic side was treated with plasma. The thickness of the foil was 32 µm. 2.2. Plasma Afterglow Treatment {#sec2dot2-materials-11-00372} ------------------------------- The polymer foil was cut into pieces of 2 × 2 cm^2^ and treated with reactive neutral oxygen species created in the center of the processing chamber. Oxygen species which were created in the surfatron plasma were passed through the narrow glass tube to the processing chamber. The experimental system is shown schematically in [Figure 1](#materials-11-00372-f001){ref-type="fig"}. The processing chamber was a pyrex cylinder with a diameter of 33 cm and a length of 40 cm. The chamber was pumped with a two-stage oil rotary pump of a nominal pumping speed of 40 m^3^·h^−1^ and ultimate pressure well below 1 Pa. A zeolite trap was used to prevent back-diffusion of the oil vapor. The pump was mounted on the flange at the bottom of the processing chamber via bellows of a large conductivity at the pressure of 20 Pa and above, and a manually adjustable shutter valve which allowed for suppressing the effective pumping speed in a gradual manner from the maximal speed (40 m^3^·h^--1^) down to zero. The upper flange of the Pyrex tube was equipped with a pressure gauge, a discharge tube, and a movable catalytic probe which was used for O-atom density measurements \[[@B36-materials-11-00372]\]. Oxygen of commercial purity 99.99% was leaked continuously in the discharge tube through a manually adjusted leak valve. A standard quartz tube with an inner diameter of 6 mm was used. The pressure was measured with an absolute gauge (baratron) calibrated for the pressure range 0.1--100 Pa. A microwave cavity of approximately 5 cm in length was mounted onto the discharge tube and connected to the microwave power supply. The configuration allowed for sustaining the gaseous plasma in the surfatron mode inside the discharge tube. The microwave power was set to 200 W. Continuous leakage of oxygen on one side and pumping of the processing chamber on the other side allowed for a drift of gas through the discharge into the processing chamber. Molecular oxygen from the flask partially ionized, dissociated, and excited in the plasma within the microwave cavity. Charged particles quickly neutralized and excited species relaxed on the way between the gaseous plasma and the processing chamber. Therefore, the only highly reactive oxygen species left for treatment of the polymer samples was neutral O atoms. The density of O atoms above the surface of the polymer samples was measured with a calibrated catalytic probe. The probe consists of a catalytic tip which is heated in the plasma because of the recombination of O atoms to O~2~ molecules on the surface of the catalyst \[[@B36-materials-11-00372]\]. The temperature of the catalyst is measured by a thermocouple. The heating rate of the probe is proportional to the flux of oxygen atoms. The O-atom density (*n*) was calculated from the probe temperature derivate using the following equation \[[@B37-materials-11-00372]\]: $$n = \frac{8 \cdot m \cdot c_{p}}{v \cdot W_{D} \cdot \gamma \cdot A} \cdot \left( \frac{dT}{dt} \right)$$ where *m* is the mass of the probe tip, *c*~p~ is its specific heat capacity, *W*~D~ is the dissociation energy of an oxygen molecule, *γ* is the recombination coefficient for O atoms on the catalyst surface, *A* is the area of the catalyst, and d*T*/d*t* is the time derivative of the probe temperature just after turning off the discharge. More details regarding the O-atom density calculation are explained in the works \[[@B36-materials-11-00372],[@B37-materials-11-00372]\]. We have used cobalt as the catalyst. This material is particularly suitable for the detection of atomic oxygen at a low density. The lower detection limit of the probe was approximately 2 × 10^18^ at a pressure above 10 Pa, whereas the upper at 10^22^ m^−3^. The experiments presented here were performed at the pressure of 20 Pa. At these conditions, the O-atom density in the system was 5.3 × 10^19^ m^--3^ when the shutter valve was fully open (the effective pumping speed was equal to the nominal pumping speed of the vacuum pump). By adjusting the shutter and leak valves simultaneously, it was possible to keep the pressure in the processing chamber constant but the O-atom density variable: less opened valves caused a lower atom density because the drift velocity of the gas through the discharge chamber was suppressed by closing valves. A detailed description of this effect was reported elsewhere \[[@B38-materials-11-00372]\]. Four adjustments of the O-atom density in the vicinity of the samples were chosen: 5.3 × 10^19^, 2.9 × 10^19^, 1.0 × 10^19^, 8.7 × 10^18^, and 3×10^18^ m^−3^. The corresponding fluxes of O-atoms onto the sample surface were calculated as: where *n* is the measured density of oxygen atoms and *v* is an average thermal velocity of O atoms at room temperature (*v* = 630 m·s^--1^). The fluence of O atoms to the surface of the sample was calculated as *j × t*, where *j* is the flux of oxygen atoms to the surface and *t* is the treatment time. Various treatment times were used for modification of the sample's surface. Such an experimental setup allowed for the treatment of samples in a broad range of fluences from 5 × 10^21^ to 3 × 10^24^ m^−2^---almost three orders of magnitude. 2.3. X-ray Photoelectron Spectroscopy (XPS) Characterization {#sec2dot3-materials-11-00372} ------------------------------------------------------------ Chemical composition of the samples was determined with an XPS instrument model TFA XPS (Physical Electronics, Ismaning, Germany) from Physical Electronics. Analyses were performed 15 min after the plasma treatment. Monochromatic Al Kα~1,2~ radiation at 1486.6 eV was used for sample excitation. Photoelectrons were detected at an angle of 45° with respect to the normal of the sample surface. XPS survey spectra were measured at a pass-energy of 187 eV using an energy step of 0.4 eV. High-resolution spectra of carbon C1s were measured at a pass-energy of 23.5 eV using an energy step of 0.1 eV. Because the samples are insulators, an electron gun was used for the additional charge compensation. The spectra were analyzed using MultiPak v8.1c software (Ulvac-Phi Inc., Kanagawa, Japan, 2006) from Physical Electronics. 2.4. Atomic Force Microscopy (AFM) Measurements {#sec2dot4-materials-11-00372} ----------------------------------------------- The surface morphology of the samples was analyzed with an AFM (Solver PRO, NT-MDT, Moscow, Russia). Images were recorded in a tapping mode using ATEC-NC-20 tips (Nano And More GmbH, Germany). A resonance frequency of the tip and the force constant were 210--490 kHz and 12--110 Nm^−1^, respectively. An average surface roughness of the samples (Ra) was determined by using the program Spip 5.1.3 (Image Metrology A/S). The average surface roughness was calculated from the images taken over an area of 5 × 5 µm^2^. 2.5. Contact Angle Measurements {#sec2dot5-materials-11-00372} ------------------------------- Changes of the surface wettability of the plasma-treated samples were determined immediately after the plasma treatment. An instrument by See System (Advex Instruments, Brno, Czech Republic) was used. A demineralized water droplet of a volume of 3 μL was applied to the surface. The measured contact angles were analyzed by the software supplied by the producer. For each sample, three measurements were taken to minimize the statistical error. 3. Results and Discussion {#sec3-materials-11-00372} ========================= [Figure 2](#materials-11-00372-f002){ref-type="fig"} illustrates the variation of the water contact angle of the acrylic coating versus the fluence of oxygen atoms. As mentioned earlier, the treatment was performed at several different densities of O atoms in the vicinity of the sample and at various treatment times. It seems that the water contact angle only depends on the fluence and not on the O-atom density because all measured points in [Figure 2](#materials-11-00372-f002){ref-type="fig"} follow the same curve. The contact angle at first decreases rapidly with the increasing fluence, but later the decrease becomes less and less rapid until the water contact angle becomes constant at approximately 40°. The particular measured points in [Figure 2](#materials-11-00372-f002){ref-type="fig"} are somehow scattered; however, the trend is obvious: no knee is observed in the curve which is only plotted for eye guidance. The contact angle becomes constant (approximately 40°) after the fluence of a few 10^22^ m^--2^ is used. Further exposure to O-atoms does not influence the wettability of this particular material. The exemptions are both measured points at very large fluences where the contact angles are approximately 30°. A feasible explanation for this effect will be presented and discussed later in this paper. [Figure 3](#materials-11-00372-f003){ref-type="fig"} represents the required treatment time for the fluences of 1 × 10^22^ and 1 × 10^23^ m^--2^. From this figure, one can conclude that the required treatment time for receiving the fluence of 1 × 10^22^ m^--2^ is only 6 ms at the atom density of 1 × 10^22^ m^--3^, which is typical for the extremely reactive oxygen plasma \[[@B39-materials-11-00372]\]. Such a short treatment time is achievable only when using pulsed discharges. Unfortunately, this experimental setup does not allow for verification of the calculated values presented in [Figure 3](#materials-11-00372-f003){ref-type="fig"}. Furthermore, in practice, such small treatment times are not very suitable, because the treated surface may be contaminated with impurities. This means that at such a short treatment time, plasma radicals interact with the contaminants rather than with a pure polymer surface. [Figure 4](#materials-11-00372-f004){ref-type="fig"} shows the variation of oxygen concentration and the O/C ratio on the polymer surface as determined by XPS. The oxygen concentration of the untreated sample was approximately 18 at %. The rather high concentration of oxygen in the surface film as detected by XPS (several nm thick) arises from the acrylic coating. After the treatment, the oxygen concentration on the surface increased. The increase is at first rapid but then less pronounced; however, the x-axis in [Figure 4](#materials-11-00372-f004){ref-type="fig"} is plotted in the logarithmic scale and therefore the measured points appear in a line. The oxygen concentration thus increases as a logarithm of the fluence. It is interesting that the oxygen concentration keeps increasing after the fluence that corresponds to the saturation of the wettability. Numerous explanations can be stated for this observation. A trivial one is that already approximately 30 at % of oxygen is enough for the optimal wettability. The second possibility is that the surface (which influences the wettability) is already saturated with the polar functional groups at a moderate fluence and oxidation of the sub-surface layers occurs at higher fluences. Yet another explanation could be the formation of oxides on the surface---this effect will be discussed later. The high-resolution spectra of the carbon C1s peak for selected samples are presented in [Figure 5](#materials-11-00372-f005){ref-type="fig"}. The spectra are normalized to the height of the main peak at 285 eV. The deconvolution of selected spectra is presented in [Figure 6](#materials-11-00372-f006){ref-type="fig"}. The untreated sample ([Figure 6](#materials-11-00372-f006){ref-type="fig"}a) contains three peaks: the main one at 285 eV corresponding to C--C, C--H bonds, and two small peaks at 286.5 and 289 eV corresponding to C--O and O=C--O groups, respectively. The spectrum in [Figure 6](#materials-11-00372-f006){ref-type="fig"}a supports the information that the original sample has the acrylic coating. [Figure 6](#materials-11-00372-f006){ref-type="fig"}b,c show an example of deconvolution of the sample treated at short (low oxygen fluence) and long (high oxygen fluence) treatment times. It can be observed that the intensity of C--O and O=C--O groups increased, especially for longer treatment times. It is difficult to judge about the formation of additional peaks corresponding to functional groups like C=O; however, if such groups develop upon treatment with the O atoms, their concentration on the polymer surface is much lower than the concentration of C--O and O=C--O groups. [Figure 5](#materials-11-00372-f005){ref-type="fig"} shows a gradual increase of the polar functional groups versus the fluence of the O-atoms, thus it is in good agreement with [Figure 4](#materials-11-00372-f004){ref-type="fig"}. The increase is not equal for C--O and O=C--O groups, though. This can be seen from [Figure 7](#materials-11-00372-f007){ref-type="fig"}, which shows the concentration of the functional groups versus the O-atom fluence. The highly polar O=C--O group increases somehow more intensively than the C--O group and actually prevails at the highest fluence. Interesting enough, this observation is not sound with the wettability presented in [Figure 2](#materials-11-00372-f002){ref-type="fig"}. Namely, on the basis of the results presented in [Figure 7](#materials-11-00372-f007){ref-type="fig"}, one would expect a monotonous decrease of the water contact angle with the increasing O-atom fluence. As mentioned above, this phenomenon could be related to surface saturation with the polar functional groups already at moderate fluences, and to oxidation of the sub-surface layers at higher fluences, or to the formation of Si oxides (discussed later). Another observation about the surface composition is worth stressing and discussing. [Figure 8](#materials-11-00372-f008){ref-type="fig"} represents survey XPS spectra for selected samples. Apart from carbon and oxygen, one can observe tiny peaks at binding energies of approximately 102 and 153 eV. The peaks correspond to silicon levels of Si 2p and Si 2s, respectively. The peaks are easily overlooked for the untreated sample (lowest curve in [Figure 8](#materials-11-00372-f008){ref-type="fig"}), but become more pronounced after the sample has received a large fluence (upper curve). Detailed spectrum in the range 88--188 eV is shown in the insert of [Figure 8](#materials-11-00372-f008){ref-type="fig"}. Doubtlessly, silicon is presented in the as-received sample and its concentration as detected by XPS increases with the increasing O-atom fluence. [Figure 9](#materials-11-00372-f009){ref-type="fig"} represents the concentration of Si in the surface of selected samples. Although the initial concentration is at the limit of this experimental technique, the trend is well justified. The origin of Si in the untreated sample is known to polymer scientists: i.e., silicon is often added to polymers as an anti-block or slipping agent in order to improve their performance. When the polymers are exposed to oxygen atoms, etching occurs. The effect has been elaborated elsewhere \[[@B40-materials-11-00372]\]. The oxygen atoms at first cause surface functionalization, but as the polymer surface becomes saturated with the O-rich functional groups, they form unstable molecular fragments which desorb from the surface. The polymer is thus slowly etched, leaving on the surface compounds that do not form volatile oxides. The effect is sometimes called plasma ashing \[[@B41-materials-11-00372]\]. Here, the acryl coating is slowly degraded and thus etched, leaving oxidized silicon nanoparticles on the surface. This effect explains the increase of Si concentration versus the O-atom fluence presented in [Figure 9](#materials-11-00372-f009){ref-type="fig"}. It may or may not be a coincidence that the Si concentration ([Figure 9](#materials-11-00372-f009){ref-type="fig"}) starts rising as the sample wettability becomes stable ([Figure 2](#materials-11-00372-f002){ref-type="fig"}). The virtual discrepancy between [Figure 2](#materials-11-00372-f002){ref-type="fig"} and [Figure 4](#materials-11-00372-f004){ref-type="fig"} can be attributed to the appearance of silicon on the polymer surface. As explained above, the wettability ([Figure 2](#materials-11-00372-f002){ref-type="fig"}) assumes a rather constant value after the fluence of about 3 × 10^22^ m^−2^, but the concentration of oxygen on the polymer surface still increases ([Figure 4](#materials-11-00372-f004){ref-type="fig"}). Taking into account the measured values of Si ([Figure 9](#materials-11-00372-f009){ref-type="fig"}) and assuming that silicon is in the form of oxide (SiO~2~), one can replot [Figure 4](#materials-11-00372-f004){ref-type="fig"} by considering that a part of oxygen is bonded to silicon, i.e., subtracting 2 × \[Si\] oxygen from the curves. The new plot of O concentration and the O/C ratio by considering this effect is plotted in [Figure 10](#materials-11-00372-f010){ref-type="fig"}. The behavior of the curve for oxygen in [Figure 10](#materials-11-00372-f010){ref-type="fig"} is now almost sound with the observations presented in [Figure 2](#materials-11-00372-f002){ref-type="fig"}. Namely, the oxygen concentration as determined from XPS results also approaches a constant value for large fluences. Unfortunately, the saturation in [Figure 10](#materials-11-00372-f010){ref-type="fig"} does not appear at the same fluence as in [Figure 2](#materials-11-00372-f002){ref-type="fig"}. The role of silicon dioxide on the sample wettability is worth discussing. [Figure 2](#materials-11-00372-f002){ref-type="fig"} represents numerous measured data that fit the curve well, but the two points at the highest fluences definitely do not fit the general behavior. The decrease of the WCA for the highest fluences could be explained by oxidized silica nanoparticles on the sample surface, because well activated silicon oxide (treated by oxygen plasma) is hydrophilic \[[@B42-materials-11-00372]\]. The hydrophilicity is, however, lost soon after the plasma treatment because of the adsorption of organic impurities. That is one of the reasons why wettability tests were performed just after the treatment of samples with the O-atoms; however, hydrophobic recovery cannot be excluded completely. In view of the upper discussion, let us also discuss the AFM images of selected samples. The images are shown in [Figure 11](#materials-11-00372-f011){ref-type="fig"}. The images were taken over the area of 5 × 5 µm^2^. The untreated sample ([Figure 11](#materials-11-00372-f011){ref-type="fig"}a) exhibits small un-evenly distributed particles of virtually the same lateral size protruding from the surface. The typical lateral dimension of the particles is almost 100 nm and the height as determined by AFM is several 10 nm. The origin of these particles is probably polymer additives containing silicon. According to the XPS results ([Figure 9](#materials-11-00372-f009){ref-type="fig"}), the density of the particles fits the concentration of silicon on the surface of the untreated sample. [Figure 11](#materials-11-00372-f011){ref-type="fig"}b is the image of the sample after receiving a small O-atom fluence. According to the upper results and discussion, the fluence received by this sample was too small to cause any detectable polymer etching. The image actually does not differ significantly from [Figure 11](#materials-11-00372-f011){ref-type="fig"}a. Also, the surface roughness of the sample shown in [Figure 11](#materials-11-00372-f011){ref-type="fig"}b did not change much (from the initial 5.8 nm it increased to 5.9 nm). One can qualitatively conclude that the concentration of the particles protruding from the sample surface is similar in [Figure 11](#materials-11-00372-f011){ref-type="fig"}a,b, which is sound with the observations presented in [Figure 9](#materials-11-00372-f009){ref-type="fig"}. The AFM images in [Figure 11](#materials-11-00372-f011){ref-type="fig"}c,d vary significantly from [Figure 11](#materials-11-00372-f011){ref-type="fig"}a,b. The particles protruding from the surface are now much denser, which could be a consequence of the polymer etching. Moreover, the surface roughness increased to 6.8 nm. From [Figure 11](#materials-11-00372-f011){ref-type="fig"}, one can therefore assume that the surface is enriched with silica nanoparticles, which has been proposed on the basis of the XPS results presented in [Figure 9](#materials-11-00372-f009){ref-type="fig"}. In [Figure 12](#materials-11-00372-f012){ref-type="fig"}, AFM topographic and phase images of the untreated sample and of one the selected treated sample recorded at a higher magnification of 2 × 2 µm^2^ are shown. The phase signal depends on the viscoelastic properties of the materials; therefore, the signal variation between the soft polymer surface and stiff silica particles can be observed. [Figure 12](#materials-11-00372-f012){ref-type="fig"} clearly shows a big difference in the variation of the phase signal for the treated sample in comparison to the untreated one. Many black spots with a big phase shift are observed on the treated sample, which confirms our conclusions about the presence of silica particles. 4. Conclusions {#sec4-materials-11-00372} ============== An early stage of activation of commercial acrylic coated polypropylene foils for food packaging has been elaborated. The results clearly show that the maximum achievable surface wettability is already obtained at a rather low fluence of O-atoms of the order of a few 10^22^ m^--3^. This information is particularly useful for users who want to activate the material without losing the acrylic surface film. Namely, larger fluences (in practice it means prolonged treatment time) has little or no effect on the surface wettability but causes etching of the thin acrylic film and thus loss of the functional properties of such foils. As stated in the introduction to this paper, the acrylic coating protects the polypropylene foil from external influences, and should therefore remain on the PP foil after accomplishing the activation procedure. This research was co-financed by the European Union from the European Regional Development Fund and by the Ministry of Education, Science and Sport (F4F "Food for future"). M.H. and T.V. conceived and designed the experiments under the supervision of M.M., M.S. and A.R.J.; M.H. and T.V. performed the experiments; M.H. performed AFM measurements; A.V. measured and analyzed the XPS results; M.S. contributed materials; M.M. wrote the paper; M.S. and A.R.J. participated in the discussion. The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results. {#materials-11-00372-f001} {#materials-11-00372-f002} {#materials-11-00372-f003} {#materials-11-00372-f004} {#materials-11-00372-f005} {#materials-11-00372-f006} {#materials-11-00372-f007} {#materials-11-00372-f008} {#materials-11-00372-f009} {#materials-11-00372-f010} {#materials-11-00372-f011} {#materials-11-00372-f012} materials-11-00372-t001_Table 1 ###### Treatment conditions and results obtained by other authors. ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Plasma Treatment Discharge Parameters Wettability Surface Analysis Reference -------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------- Low-pressure oxygen plasma −RF source 8--12 MHz−Power: 7.2, 10.2, 29.6 W−Pressure: 26.7--80 Pa−Flow: 5--10 scfh−Exposure time: up to 5 min WCA ^1^ was decreasing with increasing power and treatment time. From 74.5° it decreased to approx. 44° at the highest power of 29.6 W and at the longest treatment time 300 s.\ AFM ^3^: surface roughness RMS ^4^ increased from 1.5 to 7.3 nm.\ \[[@B24-materials-11-00372]\] SFE ^2^ increased from 56.5 to max. 94.1 mJ/m^2^. XRD ^5^: higher degree of crystallinity observed after oxygen plasma treatment.\ Mechanical properties: tensile strength decreased after plasma treatment.\ Barrier properties: water vapor transmission was increasing with increasing power and time. Low-pressure oxygen and argon plasma −commercial plasma reactor from Diener Co.−LF plasma 40 kHz−Power: 50 W−Pressure: 0.2 × 10^5^ Pa−Flow: 50 cm^3^/min−Exposure time: up to 5 min Increase of the surface energy of oxygen plasma-treated sample was higher than for the one treated in Ar.\ ATR-FTIR ^6^: carbonyl groups observed as well as C=C which could be a sign of crosslinking.\ \[[@B32-materials-11-00372]\] SFE was 70 and over 50 mJ/m^2^ for O~2~ and Ar plasma, respectively. AFM: O~2~ plasma caused higher roughness than Ar.\ Adhesion: High improvement of surface adhesion strength, especially for O~2~ plasma. Low-pressure oxygen plasma −RF source 13.56 MHz, capacitive−sample placed on the grounded electrode−Power: 70 W−Pressure: 6.7 Pa−Flow: 49 sccm−Exposure time: up to 5 min WCA decreased from 110° to 40°. AFM: surface roughness first decreased with treatment time. At longer treatment times, a significant increase is observed.\ \[[@B23-materials-11-00372]\] FTIR: C=O and --OH peaks observed for plasma-treated samples. Low-pressure air plasma −DBD plasma−sample placed on the grounded electrode−AC power source 10 kHz−Power: 1.4 W−Pressure: 5 × 10^3^ Pa−Flow: 20 sccm−Exposure time: 0.2--30 s−Energy load: up to 3.34 J/cm^2^ WCA decreased from 94.9° to 60°. Saturation reached after 10 s. XPS ^7^: Oxygen concentration increased from 4.3 to 13.7 at %. Nitrogen (0.8 at. %) was also found, the rest being carbon.\ \[[@B21-materials-11-00372]\] 8.2% C--O, 2.7% C=O and O--C=O and 86.4% C--C, C--H groups observed on plasma treated sample.\ ATR-FTIR: peaks attributed to OH and C=O in ketones, aldehydes and carboxylic acids. Atmospheric pressure air plasma −DBD plasma−"Coating Star" device from Ahlbrandt System−sample placed on the grounded electrode−30 kHz, 15 kV−Power: 300--1000 W−Pressure: atmospheric−Flow: 20 sccm−Treatment speed: 2--10 m/min−Energy load: up to 60 kJ/m^2^ WCA decreased from 104° to 64° even at 1.2 kJ/m^2^.\ XPS: O/C ratio increased over 0.16. Nitrogen (2 at %) was also found. After one month, O/C decreased to 0.12. Groups like C--O (22.5%), C=O or O--C--O (8.4%) and O=C--O (5.3%) were found. Maximum concentration was obtained at the lowest treatment speed.\ \[[@B22-materials-11-00372]\] SFE increased from 33.7 to almost 50 mN/m. AFM: Ra ^8^ increased from 5.8 to 12.9 nm. Bumps were observed on the surface. The height and width were increasing with treatment power and reached 60 and 500 nm, respectively. Low-pressure oxygen and argon plasma −RF source 13.56 MHz−Commercial K 1050X Plasma Asher Model from Emitech Co.−sample placed in the middle of the chamber on the glass substrate−Power: 10, 30, 50 W−Pressure: 0.35 × 10^5^ Pa−Flow: 15 mL/min−Exposure time: up to 5 min WCA was decreasing with increasing power and treatment time. The lowest WCA was 34.4° for O~2~ and 38.2° for Ar plasma (initially 98.3°).\ SEM ^9^ and AFM: topology and roughness changed significantly, especially for Ar plasma (nodules observed on the surface). RMS roughness increased from 3.6 to 6.9 and 6.1 nm for O~2~ and Ar plasma, respectively.\ \[[@B25-materials-11-00372]\] SFE increased to \~45 mN/m. ATR-FTIR: C=O stretching bond and C=C vibration observed. Some peaks attributed also to carboxylic/ester, aldehydes and ketone groups. Low-pressure oxygen plasma −RF power source−sample was on the tray in the middle of the chamber−Power: 500 W−Pressure: 13.3 Pa−Flow: 49 sccm−Exposure time: up to 40 min WCA decreased from 121.5° to 84° on PP nonwoven mats. Ageing for 90 days did not have significant effect on WCA.\ SEM: etching of PP fibers observed.\ \[[@B33-materials-11-00372]\] SFE increased from 13.7 to 29.2 mN/m. XRD: no significant effect on the crystallinity of the treated fibers. Low-pressure oxygen plasma −RF source 13.56 MHz, capacitive−Power: 0--150 W−Pressure: 0--120 Pa−Exposure time: 30 s--3 min−Ageing: 30 days The lowest WCA---bellow 10° was observed at 150 W, 3.33 Pa and 60 s.\ Ageing and crystallinity: Two polymers with different initial crystallinity were used. More crystalline PP was ageing slower---WCA after 30 days was for \~5° lower than for less crystalline one.\ \[[@B34-materials-11-00372]\] After 30 days of ageing WCA increased to \~50°. Degree of crosslinking was increased after the treatment for both polymers.\ XPS: \~25 at % of oxygen was found on less-crystalline polymer. O concentration on more crystalline polymer was few at % lower. However, after ageing the O concentration changed in favor of more crystalline one. Low-pressure oxygen plasma −RF power source−commercial reactor Inverse Sputter Etcher ISE 90 model 2001 (Von Ardenne Anlagentechnik−GmBh)−Power: 50 W−Pressure: 5.1 Pa−Exposure time: up to 40 min WCA decreased from 98° to 24°. At long treatment times, it increased to 53°. AFM: roughness RMS increased from \~ 12 nm to \~44 nm.\ \[[@B26-materials-11-00372]\] ATR-FTIR: OH, C=O and\ CO--C=O peaks observed for plasma treated samples. Low-pressure oxygen plasma −DC plasma (20 mA, 2 kV)−sample was placed on glass walls of discharge chamber positioned between the electrodes separated 42 cm−Pressure: 30 Pa−Exposure time: up to 200 s WCA decreased from 83° to 60°.\ ATR-FTIR: OH, C=O groups in ester, ketone and carboxyl groups, C=O groups in unsaturated ketones and aldehydes. \[[@B20-materials-11-00372]\] SFE increased from 25.7 to 43 mJ/m^2^. Low-pressure oxygen plasma −Capacitor plate plasma−commercial K1050 X Plasma Asher from Emitech Ltd.−sample positioned on the holder−Power: 0--100 W−Pressure: 60 Pa−Flow: 15 mL/min−Exposure time: up to 10 min−Ageing: 90 days in air or water WCA was decreasing with the increasing power and treatment time. Minimal achievable\ AFM: Roughness RMS increased after treatment from 2.1 nm to \~10 nm (in air) and \~5 nm (in water). Lower roughness of samples stored in water was explained by removing of water-soluble short-chain species. \[[@B35-materials-11-00372]\] WCA was 55.6° (initially 103°). Ageing in water was faster than in air. After 90 days, the WCA was 81.7° (in water) and 71.2° (in air). ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ^1^ Water contact angle (WCA); ^2^ Total surface free energy (SFE); ^3^ Atomic-force microscopy (AFM); ^4^ Root mean squared (RMS) roughness; ^5^ X-ray diffraction (XRD); ^6^ Attenuated total reflection Fourier transform-infrared spectroscopy (ATR-FTIR); ^7^ X-ray photoelectron spectroscopy (XPS); ^8^ Average roughness (Ra); ^9^ Scanning electron microscope (SEM).

Introduction ============ Middle ear surgery for chronic middle ear disease is aimed at eradicating inflammation and preventing disease recurrence. Preservation or improvement of hearing is an additional goal of middle ear surgery. However, surgery for middle ear disease of an ear that has a better hearing level than the opposite ear is challenging because of the importance of hearing preservation. Surgical management of the only hearing ear is stressful for the surgeon because there is a possibility of hearing deterioration. Conservative management, such as careful observation and hearing aid use, is one of the choices, but surgical management should be considered when there is refractory otorrhea or a destructive lesion. There have been a few reports into surgical interventions in only hearing ears. Many surgeons favor surgical intervention for only hearing ears with cholesteatoma.[@B1] Sanna, et al.[@B2] used a questionnaire to study how surgeons treat patients who have chronic ear disease in only hearing ears. The results indicated that there is no consensus about the type of surgical treatment and surgical indications. Accordingly, the aim of this study was to evaluate surgical interventions and hearing rehabilitation in patients with chronic middle ear disease of only hearing ears. Subjects and Methods ==================== We retrospectively reviewed the medical records of 31 patients who underwent middle ear surgery of only hearing ears by one senior surgeon in a single tertiary care center from 1992 to 2011. The age range of the patients was 25 to 65 years, with a mean of 45.3 years, with 21 female and 10 male patients. We examined the etiology of opposite deaf ears and the background of the decision making for surgery. Preoperative and postoperative pure tone and speech audiometry were checked to determine the hearing statuses of the patients. The average pure tone audiometry (PTA) was calculated from thresholds of 500 Hz, 1000 Hz, 2000 Hz, and 3000 Hz. A deaf ear was defined as that with an average air conduction threshold worse than 90 dB HL in PTA. Temporal bone computed tomography was preoperatively performed. We performed canal wall-up mastoidectomy with or without tympanoplasty for most patients with chronic otitis media and canal wall-down mastoidectomy for most patients with cholesteatoma. Conservative and minimal surgical procedures were preferred in only hearing ears except for those with destructive lesions such as large cholesteatoma. We classified the patients into three groups according to the hearing level. Group A contained patients with serviceable hearing levels-better than 40 dB HL-and only observation with-out other rehabilitation methods was recommended. Group B patients-with a hearing level between 40 dB HL and 70 dB HL-were recommended to use hearing aids for hearing rehabilitation. Group C contained patients with hearing levels worse than 70 dB HL. In addition, changes in hearing levels were also classified into three groups according to the difference between the preoperative and postoperative air conduction thresholds: improved, no change, and worsened. A significant hearing level change was defined as 15 dB. Results ======= Preoperative diagnoses of chronic middle ear disease in only hearing ears were as follows: 21 cases (67.7%) of chronic otitis media, 9 cases (29.0%) of chronic otitis media with cholesteatoma, and 1 case (3.2%) of cholesterol granuloma ([Table 1](#T1){ref-type="table"}). The main causes of surgery for chronic otitis media were refractory otorrhea recurring over three times during 1 year. The indication for surgery in patients with chronic otitis media with cholesteatoma and cholesterol granuloma was prevention of further destruction of surrounding structures. Before surgery, frequent dressing changes and antibiotic medication were administered according to bacteriologic cultures for preoperative management. The average hearing level of contralateral deaf ears was 97.3 dB \[25-75% interquartile range (IQR)=91.8-101.5\]. The most common etiology of contralateral deaf ears was surgery for chronic middle ear disease in 21 cases (67.7%). There were four cases (12.9%) of untreated chronic middle ear disease and six cases (19.4%) of sensorineural hearing loss. Surgical procedures are summarized in [Table 2](#T2){ref-type="table"}. Tympanoplasty was performed in five cases (16.1%) and all were of chronic otitis media. Canal wall-down mastoidectomy and tympanoplasty was performed in nine cases (29.0%), eight of which accompanied cholesteatoma; the other case was of recurred chronic otitis media after previous surgery. The most frequently performed procedure was canal wall-up mastoidectomy with tympanoplasty, which was performed in 17 patients (54.8%). Preoperative air and bone conduction PTAs were 55.3 dB HL (25-75% IQR=35.8-70.8) and 28.0 dB HL (25-75% IQR=9.2-46.7), respectively, and the air-bone gap was 27.4 dB (25-75% IQR=20.0-35.8). Postoperative air and bone conduction PTAs were 55.0 dB HL (25-75% IQR=27.5-74.2) and 31.3 dB HL (25-75% IQR=12.5-49.2), respectively, and the air-bone gap was 23.7 dB (25-75% IQR=10.0-36.7). Three cases (9.7%) showed hearing improvement after surgery: all of them were for chronic otitis media and a canal wall-up procedure was performed. Four cases (12.9%) had worsened hearing after surgery: three of them were of patients with cholesteatoma and a canal wall-down procedure was performed in three cases ([Table 3](#T3){ref-type="table"}). Preoperative and postoperative hearing levels and hearing rehabilitation recommendations are summarized in [Table 4](#T4){ref-type="table"}. Nine preoperative patients belonged to group A and eight of these were still group A after surgery. They did not need any hearing rehabilitation. One patient had hearing deterioration after surgery (to group B). Fourteen patients were preoperatively classified as group B, and three of them had hearing improvement after surgery (to group A). A total of 12 patients of postoperative group B were recommended to use hearing aids. All eight preoperative group C patients were still group C after surgery, and they were potential candidates for cochlear implantation because of bilateral severe or profound hearing loss. The speech discrimination scores of patients improved with hearing rehabilitation ([Table 5](#T5){ref-type="table"}). Postoperative group A did not require hearing rehabilitation. Postoperative group B patients were recommended to use hearing aids. We recommended hearing aids to 12 patients; although three of them refused to use these devices, the other nine patients wore hearing aids with successful rehabilitation. Speech discrimination scores improved with hearing aids. Postoperative group C was recommended to undergo cochlear implantation: three of them underwent cochlear implantation, three of refused these implants and wore hearing aids, and the remaining two patients refused hearing rehabilitation. The patients who underwent cochlear implantation showed extremely poor preoperative (precochlear implantation) discrimination scores, with a mean value of 10.7%. A clear benefit was seen after cochlear implantation, with discrimination scores improved to 50.7%. Patients who wore hearing aids showed an improvement in speech discrimination scores, from 76.0% to 88.0%. During the follow-up period over 1 year, recurrent tympanic membrane perforation was identified in two patients. Myringoplasty was done in these two cases. There was no recurrent otorrhea to disturb the use of hearing aids. As a result, all patients achieved dry ears after surgery. Discussion ========== Surgical management of only hearing ears remains challenging because of a possible aggravation of hearing after surgery. Control of disease and preservation of hearing are both important in this situation. Especially in cholesteatoma, in the aspect of cholesteatoma itself, it needs to be eliminated by appropriate surgery. However in the aspect of hearing, the removal of cholesteatoma has higher risk of hearing aggravation when compared to non-cholesteatoma chronic otitis media. In our study, 3 patients had hearing aggravation and 6 patients showed no change of hearing after surgery among 9 cholesteatoma patients. But considering the growth of cholesteatoma and high possibility of hearing aggravation due to cholesteatoma, just observing the patient without surgery could not be a proper management option, even in only hearing ear. However, Sakagami, et al.[@B3] reported via long-term follow-up results of chronic otitis media that the hearing of ears with chronic otitis media worsened significantly more than that of normal contralateral ears. Wait and see is not a good policy for only hearing ears with chronic ear disease because hearing impairments will be progressive, especially when the otorrhea is recurrent or persistent. Surgeons should carefully choose the surgical procedure, and either canal wall-up or canal wall-down mastoidectomy procedures or tympanoplasty can be used for chronic middle ear disease in only hearing ears. The type of mastoid surgery has been shown not to affect the hearing results of chronic otitis media patients.[@B4] Some previously published reports remarked on the type of surgery. Yamamoto, et al.[@B5] performed canal wall-up mastoidectomy with tympanoplasty in 16 of 30 patients in 1997. Perez de Tagle, et al.[@B6] performed canal wall-down mastoidectomy with tympanoplasty in 7 of 8 patients. Sakagsami, et al.[@B7] recommended a type 1 tympanoplasty without mastoidectomy for chronic otitis media and a canal wall-down mastoidectomy with ossicular reconstruction in patients with cholesteatoma. Battaglia, et al.[@B8] contended that one-stage canal wall-down procedures are preferable to minimize the number of times that the ear is placed at risk. Conclusion ========== All of our patients attained dry ears after surgery. Hearing aids were used in most patients with moderate to severe hearing loss and cochlear implants were used for auditory rehabilitation in patients with severe to profound hearing loss with successful results. Hence, proper preoperative evaluation and indications for surgery are important for successful eradication of inflammation and hearing preservation in only hearing ears. The association of cholesteatoma and recurrent otorrhea was a surgical indication in only hearing ears. Surgical interventions can achieve dry ears and enable further auditory rehabilitation using hearing aids and cochlear implants. ###### Preoperative diagnosis of only hearing ears and contralateral deaf ears  ###### Surgical procedures performed in only hearing ears  ###### Postoperative air conduction hearing level changes compared with preoperative levels  ^\*^a significant hearing level change was defined as a 15-dB air conduction threshold difference between preoperative and postoperative measurements ###### Hearing results and principles of postoperative hearing rehabilitation  ###### Results of hearing rehabilitation and changes in speech discrimination scores  Postop: postoperative

Background ========== Behaviour change among gay men in the early 1980s in response to the AIDS crisis contributed to a reduction in HIV transmission and that of other sexually transmitted infections (STIs) \[[@B1]\]. However, recent increases in STIs and outbreaks of less common STIs such as syphilis amongst gay men in several UK locations \[[@B2],[@B3]\] and elsewhere in Europe \[[@B4]\] raise concerns over complacency about safe sex among gay men. Moreover, STIs such as syphilis may interact with HIV and exacerbate HIV transmission \[[@B5]\]. The largest and most sustained syphilis outbreak in the UK, in Manchester, saw 306 individuals infected between January 1999 and June 2002 \[[@B6]\], at which point the outbreak showed no sign of slowing. The outbreak focuses on homosexual men (83% of cases), and a significant proportion have HIV (29%) \[[@B6]\]. We previously found that, compared to controls without syphilis, individuals infected with syphilis had more partners (particularly oral sex partners) and were more likely to seek partners in anonymous sex venues \[[@B7]\]. However, neither control nor syphilis groups were consistent in using condoms, and many controls had highly risky behaviour. Using a technique not previously used to investigate such an outbreak, we aim to quantify the behavioural overlap between gay men with syphilis (with or without HIV) and those without. Methods ======= Of 58 cases of syphilis presenting at genito-urinary medicine (GUM) departments in Greater Manchester between May 1999 and August 2000, 33 homosexual men (all those whom we were able to contact) were invited to take part and 23 (70%) gave information on their sexual and social behaviour for 12 months prior to their diagnosis. Three controls were recruited for each case. Controls were recruited through the voluntary sector (44/85 contacts) and directly from known gay social areas (18/42 contacts) and were matched on area of residence (first part postcode), age and ethnicity (all white). Interviews took place between December 2000 and March 2001. Information was gathered using a semi-structured questionnaire. Controls were offered a sexual health check comprising tests for gonorrhoea, syphilis, chlamydia and hepatitis B (and HIV if required) at a GUM department. Of 62 controls, 36 made appointments and 24 were screened: one chlamydia and one hepatitis B case was diagnosed. Behavioural variables (including sexual and drug taking behaviours and venues used to meet partners) for each individual (irrespective of infection status) were entered into a Detrended Correspondence Analysis (DCA) \[[@B8]\] run through PCOrd verson 4.10 (MjM Software, Oregon, USA) to identify and plot similarities in terms of behaviour between people. This technique was chosen over other ordination techniques because the data were nonlinear (cf principle components analysis), to avoid the arch effect produced when plotting the major axes of variation (cf correspondence analysis), and because of DCA\'s ability to utilise binary variables (cf multidimensional scaling) \[[@B9]\]. Binary variables were rejected if fewer than ten individuals fell into a category. The rejected variables were: ever injected drugs (8/85 individuals), been paid for sex (5/85 individuals), used a condom for oral sex (3/85 individuals), used crack cocaine (2/85 individuals). DCA produces an ordination plot of major axes of variation in a dataset, which enables a large number of variables to be combined. Commonly the first two major axes of variation are plotted to display all individuals in terms of the similarity (or otherwise) of the variables employed. A discriminant analysis (DA) using SPSS v11 was carried out on the first two ordination axes to aid interpretation of possible groupings between individuals with and without syphilis. Univariate statistics (chi-square and Mann-Whitney U tests) were used to compare the groups predicted by the DA. Individuals are labelled on the plot as to their infection category: those with syphilis alone (*n*= 16), those with syphilis and HIV (*n*= 7), HIV alone (*n*= 13) and those with neither infection (*n*= 49). Results and discussion ====================== Once DCA had been used to produce a two-dimensional plot of the major variation in the dataset, individuals were labelled with their infection status (Fig. [1](#F1){ref-type="fig"}). The closer the points are on the plot, the more similar the individuals\' behaviour. Inspection reveals a tight group of individuals towards the left who display similar behaviour. Discrimant analysis indicated that there was a high degree of discrimination between individuals infected with syphilis and those uninfected on the basis of their behaviours. Analysis showed that a higher degree of discrimination (about 77.6% -- see Table [1](#T1){ref-type="table"}) was achieved using both ordination axes compared to using the major axis of variation alone (71.8%). Two groups can therefore be identified on the ordination biplot (see diagonal line on Fig. [1](#F1){ref-type="fig"}). The group to the left comprises 88% of individuals with neither infection, 38% of those with HIV alone and relatively few individuals with syphilis (22%). Univariate analysis comparing the two groups reveals that those in the right-hand side of the figure are significantly more likely to have used a variety of drugs: cocaine, poppers, amphetamines, ecstasy viagra and Gamma Hydroxybutyrate (GHB) during sex (Table [2](#T2){ref-type="table"}). They were significantly more likely to have met partners in cruising areas (public areas used for sex), cottages (public lavatories) and darkrooms (private rooms in pubs); and were more likely to have met sexual partners abroad. In addition, they had more sexual partners (both oral and anal), and could name fewer of them (Table [3](#T3){ref-type="table"}). Groups on the left and right of Fig. [1](#F1){ref-type="fig"} did not differ in condom use for anal sex. {#F1} ###### Discriminant analysis on the basis of syphilis infected individuals using DCA axis 1 and axis 2 scores Predicted group membership Totals --------------------------- ------------------- ---------------------------- ------------ ---- Original group membership Not infected 48 (77.4%) 14 (22.6%) 62 Syphilis infected 5 (21.7%) 18 (78.3%) 23 ###### Chi square analysis of frequency of behaviour and predicted group membership on the basis of discriminant analysis. Df = 1 in all cases. Rows in bold are significant at *P*\< 0.05. ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Behaviour Chi square value *P* Percentage of predicted uninfected exhibiting behaviour\ Percentage of predicted syphilis infected group exhibiting behaviour\ (*n*= 53) (*n*= 32) -------------------------------- ------------------ ------------- ---------------------------------------------------------- ----------------------------------------------------------------------- Paying for sex 0.027 0.870 11.3 12.5 Using condom during anal sex 1.637 0.201 45.3 31.3 **Using cocaine** **5.647** **0.017** **30.2** **56.3** **Using poppers** **6.064** **0.014** **71.7** **93.8** Using alcohol 2.495 0.114 98.1 90.6 **Using amphetamines** **6.574** **0.010** **34.0** **62.5** **Using ecstasy** **5.930** **0.015** **41.5** **68.8** **Using Viagra** **11.715** **0.001** **7.5** **37.5** **Using GHB** **18.827** **\<0.001** **18.9** **65.6** Meeting partners in pubs 0.096 0.756 86.8 84.4 Meeting partners in clubs 2.315 0.128 73.6 87.5 Meeting partners in saunas 0.960 0.327 45.3 56.3 **Meeting through cruising** **13.443** **\<0.001** **34.0** **75.0** **Meeting through cottaging** **17.805** **\<0.001** **13.2** **56.3** **Meeting in dark rooms** **11.384** **0.001** **28.3** **65.6** Meeting through chat rooms 3.355 0.067 17.0 34.4 Meeting on the internet 0.113 0.737 18.9 21.9 **Meeting partners abroad** **18.093** **\<0.001** **5.7** **43.8** Meeting partners in Manchester 3.732 0.053 9.4 25.0 ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ###### Mann-Whitney tests between groups predicted to be syphilis infected and uninfected using discriminant analysis on the basis of sexual behaviour. ----------------------------------------------------------------------------------------------------------------------------------------------------------------------- Behaviour *z*value *P* Median value for predicted uninfected group\ Median value of predicted syphilis infected group\ (*n*= 53) (*n*= 32) ---------------------------------------------- ---------- --------- ---------------------------------------------- ---------------------------------------------------- Number of sexual partners in 12 months 6.22 \<0.001 5 30 Proportion of names of sexual partners known 6.90 \<0.001 60 5.5 Number of anal sex partners in 12 months 3.56 \<0.001 3 10 Number of oral sex in 12 months 5.39 \<0.001 5 30 ----------------------------------------------------------------------------------------------------------------------------------------------------------------------- The power of the technique is demonstrated by the fact that one of the individuals with both syphilis and HIV who appears on the left of the graph had very few partners and claimed to have been infected with syphilis when forced to perform oral sex. One in eight of those with neither infection and nearly three fifths of HIV positive individuals showed behaviour patterns that were more similar to that of the majority of those with syphilis (i.e. on the right hand side of Fig. [1](#F1){ref-type="fig"}). These individuals are putting themselves at risk of contracting HIV or other STIs. Indeed, the two controls diagnosed with an STI (chlamydia and hepatitis B) during screening at a GUM clinic following the interview both fall to the extreme right of Fig. [1](#F1){ref-type="fig"}. This technique provides an objective method of using behaviour to discriminate between infected and uninfected controls and to identify which individuals are putting themselves at risk. It is particularly useful in analysis of outbreaks of infectious diseases since a relatively large number of binary variables can be combined across a relatively small number of individuals. Conclusions =========== The recent UK sexual health strategy set new targets for reducing the levels of STIs. However, the Manchester syphilis outbreak continues unabated, raising the possibility that the infection will become endemic. The rate of partner change in the population needs to be high for bacterial sexually transmitted infections to become endemic (since individuals are infectious for only a relatively short period of time) \[[@B10]\]. Here, we have identified high levels of partner change, even among one in eight uninfected HIV negative controls. Interventions are urgently required to address the apparent increase in complacency about safe sex among homosexual men. The at-risk group is variable in terms of potentially risky behaviours (right-hand side of Fig. [1](#F1){ref-type="fig"}), thus targeting must be flexible. The presence of syphilis in the population may foreshadow an increase in HIV, since those with syphilis are 2 to 5 times more likely to acquire or pass on HIV \[[@B5]\]. List of abbreviations ===================== DA Discriminant analysis DCA Detrended Correspondence Analysis GUM genitourinary medicine HIV human immunodeficiency virus STI sexually transmitted infection UK United Kingdom Competing interests =================== None declared. Authors\' contributions ======================= All the authors participated in the design of the study. CPW carried out the Detrended Correspondence Analysis and the discriminant analysis, interpreted the results and drafted the manuscript. PAC assisted in statistical analysis, interpretation of the results and in drafting the manuscript. PC coordinated the study, recruited participants and designed and carried out the questionnaires. MAB and QS conceived the study, and QS provided epidemiological support. MAB assisted in questionnaire design and aided in the interpretation of the results. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2458/3/34/prepub> Acknowledgements ================ We thank the staff from the GUM departments of North Manchester General Hospital, Manchester Royal Infirmary and Withington Hospital. The GUM consultants, Steve Higgins, Ashish Sukhankar, Penny Chandiok and Debashis Mandel provided considerable support for this project, as did Diana Leighton (Liverpool John Moores University) and Ann Hoskins, Alan Jones, David McKelvey and Bridget Hughes from Manchester Health Authority. The cooperation of the voluntary organisations Body Positive North West, George House Trust and the Lesbian and Gay Foundation was paramount to recruiting controls. We would also like to thank two referees (Mark Hill and Michael Rekart) for comments on an earlier version of the paper. Finally, we thank all the individuals who sacrificed their time to be interviewed as part of this study. This study was funded by Manchester Health Authority.

Introduction {#s1} ============ > ***'And now let doctors quit the centre stage*** > > ***To usher in the prophylactic age'*** > > From: 'Superfluous Doctors' in: Poems from a Prisoner of War Camp > > AL Cochrane 1942 Lifestyle and health-related behaviours are powerful determinants of morbidity and mortality worldwide [@pone.0081877-Ford1], [@pone.0081877-Reeves1], and unhealthy behaviours lie at the root of many chronic and disabling diseases [@pone.0081877-Danaei1]. Major concerns focus on smoking, body mass, physical activity, diet and alcohol consumption, and the concept of a healthy lifestyle is usually defined with reference to combinations of these factors. A number of studies have shown that the following of lifestyles based on these factors is strongly associated with reductions in the incidence of certain chronic diseases [@pone.0081877-Khaw1] [@pone.0081877-Stampfer1] [@pone.0081877-Chiuve1] [@pone.0081877-Kvaavik1]. Indeed, as one writer put it: 'Healthy living is the best revenge' [@pone.0081877-Ford2]. Lifestyles, and in particular physical activity [@pone.0081877-Buchman1] [@pone.0081877-Liu1] have been shown to be associated with cognitive health [@pone.0081877-Sabia1] [@pone.0081877-Small1]. Several reviews however have commented that the quality of the evidence is low and have called for more long-term cohort studies before conclusions can be drawn with confidence about the role of healthy behaviours in cognitive health [@pone.0081877-Williams1], [@pone.0081877-Plassman1] [@pone.0081877-Snowden1]. A major issue is reverse causation. While this is a concern with all the outcomes, it is an especial concern in relation to dementia, because the pathophysiological processes associated with Alzheimer\'s Disease are known to begin many years prior to detectable cognitive impairment [@pone.0081877-Sperling1]. Evidence from short-term studies is therefore limited, seriously in the case of cognitive impairment, and long-term studies are greatly to be preferred. The personal and public health benefits from healthy behaviours have been shown to have enormous potential so it is important that the uptake is carefully monitored. Evidence from a number of studies in the UK [@pone.0081877-Khaw1] [@pone.0081877-Kvaavik1] and elsewhere [@pone.0081877-Chiuve1] [@pone.0081877-Hu1] [@pone.0081877-Stampfer1] show that the uptake of truly healthy lifestyles is very poor. The Caerphilly Prospective Study (CaPS) is based on a cohort of men in a typical small town in South Wales UK. The men have been repeatedly questioned and examined for over 30 years. In this report we summarise evidence on relationships between healthy lifestyles at baseline and the incidence of diabetes, vascular disease, cancer, all-cause mortality, cognitive impairment and dementia during follow-up. We also examine changes in the following of healthy behaviours over the thirty years. Methods {#s2} ======= Ethics approval was obtained from the South Wales Research Ethics Committee D, and each subject signed their agreement to be involved. Attempts were made to include in the cohort all men aged 45-59 years living in Caerphilly. Extensive data were collected at baseline (Phase I), including smoking history, self-reported physical activity and alcohol consumption, and, with help from his partner, each man completed a food frequency questionnaire. Extensive clinical data were also recorded at baseline. The present or most recent occupation of each man was used to derive social class as in the Office of National Statistics classification, and is summarised as 'non-manual' and 'manual'. The definitions of health related behaviours are based on answers to questions asked at baseline in 1979, but we have tried to match definitions used in recent studies, including the annual Welsh Health Survey [@pone.0081877-Anonymous1], and other cohort studies [@pone.0081877-Khaw1] [@pone.0081877-Chiuve1]. Thus, Smoking: men not smoking, including ex-smokers; body mass index (BMI): 18 to under 25 Kg/m2; Diet: consumption of fruit and vegetables was low in the community, therefore three or more portions of fruit and/or vegetables a day was accepted as 'healthy', together with less than 30% of calories from fat; Physical activity: walking two or more miles to work each day, or cycling ten or more miles to work each day, or 'vigorous' exercise described as a regular habit; Alcohol: three or fewer units per day, with abstinence not treated as a healthy behaviour. The uptake of each of these behaviours, and combinations of them were later compared with similar data obtained in 2009 in the Welsh Health Survey [@pone.0081877-Anonymous1], based upon self-reports from 15,000 adult subjects across Wales. Approximately every five years after baseline the men were re-questioned and re-examined and primary care and hospital records inspected to identify new cases of diabetes type 2, and vascular events. Diabetes was self-reported. Clinical details of all possible ischaemic heart disease events, including electrocardiogram (ECG) and appropriate enzyme levels were evaluated against standard diagnostic criteria, and details of possible stroke symptoms together with computerized tomography (CT) scans were evaluated by two expert observers against standard diagnostic criteria. Notifications of deaths and cancer registrations were obtained from the Office of National Statistics. In the second re-examination of the men, when they were aged 55--69 years, cognitive function testing was introduced into the study [@pone.0081877-Elwood1]. In 2004, when the survivors of the original cohort were aged 70--85 years, they were assessed in detail for cognitive impairment and for dementia, and the medical records of all the other men in the cohort were examined for evidence of cognitive impairment. Those who had a score on the Cambridge Cognitive Examination (CAMCOG) of less than 83, or a decline in CAMCOG score of 10 or more since the earlier cognitive examination, were selected for a clinical assessment. Full details of the clinical assessment of the selected men are reported elsewhere [@pone.0081877-Fish1]. In brief: this included a modified CAMDEX interview [@pone.0081877-Roth1]; the Rosen-revised Hachinski Ischaemic Score [@pone.0081877-Rosen1]; a neurological examination with Frontal Assessment Battery [@pone.0081877-Dubois1]; the Clinical Dementia Rating [@pone.0081877-Berg1], and the Informant Questionnaire on Cognitive Decline in the Elderly [@pone.0081877-Jorm1]. Statistical methods {#s2a} ------------------- Lifestyle was defined as the number of healthy behaviours practiced. Odds ratios (OR) and 95% confidence intervals (CI) were calculated for different numbers of healthy behaviours; number of behaviours was modeled as an ordinal variable. Men who practiced no healthy behaviours were used as the reference group. Trend was tested using an extension to the Wilcoxon Rank-Sum Test [@pone.0081877-Cuzick1]. Adjustments were made for age and social class due to their strong relationships to all the outcomes and to the healthy behaviours. For cognitive impairment and dementia, additional adjustments were made for pre-morbid cognitive ability using the National Adult Reading Test (NART) [@pone.0081877-Nelson1] assessed at initial cognitive assessment (phase 3). The main analyses compare incident cases with the remainder of the cohort. Sensitivity analyses were conducted by excluding men with evidence of disease at baseline (diabetes, a history of angina, chest pain, clinical or ECG evidence of infarction, stroke, high blood pressure). Reverse causation for cognitive outcomes was investigated by omitting men with evidence of early cognitive impairment: i.e. CAMCOG score monotonically declining from the initial cognitive assessment. Rate advancement [@pone.0081877-Brenner1] gives an estimate of the average time by which any risk could, in theory, be postponed through the practice of a number of healthy behaviours. These times are calculated by comparing the effects of age on the outcomes with the effects of the healthy behaviours. A limitation in rate advancement is that it is only useful in the study of outcomes the rate of which increases with age. If the age relationship is weak, the rate advancement estimation is unreliable and not a useful measure. It is also worth noting that the estimates are conditional on the absence of competing risks. For example, if one disease is postponed, another disease may occur. Rate advancement could not be calculated for diabetes or dementia as no dates of onset were available. Results {#s3} ======= At baseline, 2,235 men, representing 89% of the defined population, were examined. 46% were non-smokers and 35% had a BMI of 18 to under 25. Only fifteen men consumed five or more portions of fruit and/or vegetables daily, so the definition of this behaviour was reduced to three or more portions per day, and 18% of men satisfied this criterion. 39% took regular exercise, and 59% stated that they drank within the guidelines ([Table 1](#pone-0081877-t001){ref-type="table"}). 10.1371/journal.pone.0081877.t001 ###### Odds ratios for individual healthy behaviours at base-line and the various outcomes during 30 years in the Caerphilly Prospective Study. {#pone-0081877-t001-1} Healthy Disease Cognitive function Death ------------------------------------------------ ---------------- ---------------- ---------------- ---------------- ---------------- ---------------- **Non-smoking** 1.36 0.70 0.65 0.74 0.95 0.42 ** **(1,017 men) (1.02 to 1.82) (0.58 to 0.84) (0.54 to 0.79) (0.50 to 1.08) (0.54 to 1.69) (0.35 to 0.51) **BMI**: 18 to under 25 0.31 0.69 1.37 1.00 1.06 1.27 ** **(791 men) (0.21 to 0.45) (0.57 to 0.83) (1.13 to 1.66) (0.66 to 1.51) (0.58 to 1.94) (1.05 to 1.53) **Fruit/vegetable consumption 3+portions/day** 0.91 0.95 0.97 0.79 0.80 0.82 ** **(394 men) (0.62 to 1.33) (0.75 to 1.21) (0.76 to 1.23) (0.52 to 1.20) (0.40 to 1.61) (0.65 to 1.03) **Regular exercise** 0.63 0.89 1.06 0.62 0.41 0.82 ** **(874 men) (0.46 to 0.85) (0.74 to 1.07) (0.88 to 1.28) (0.41 to 0.92) (0.22 to 0.77) (0.68 to 0.99) **Alcohol intake** within limits 1.02 0.97 0.94 0.68 0.65 0.87 ** **(1,320 men) (0.77 to 1.37) (0.81 to 1.16) (0.78 to 1.14) (0.46 to 1.00) (0.37 to 1.16) (0.72 to 1.04) All odds ratios have been adjusted for age and social class. [@pone.0081877-Nelson1]. Cognitive function has also been adjusted for the National Adult Reading Test Vascular disease includes ischaemic heart disease and ischaemic stroke. **Smoking**  =  men not smoking, including ex-smokers. **BMI** (Body Mass Index)  =  18 to 25 Kg/m2. **Fruit and vegetable consumption**  =  3 or more portions per day, plus less than 30% of calories from fat. The criterion for fruit and vegetable intake had to be reduced because only 15 men consumed five portions per day). **Regular exercise**  =  walking two or more miles to work each day, or cycling ten or more miles to work each day, or 'vigorous' exercise described as a regular habit. **Alcohol intake**  =  three or fewer units per day, with abstinence not treated as a healthy behaviour. The numbers of men judged to be following a healthy lifestyle were as follows: 179 (8%) followed none of the five behaviours, 702 (31%) followed one behaviour, 814 (36%) followed two, 429 (19%) followed three, 111 (5%) followed four or five behaviours and only two (0.1%) followed all five behaviours ([Table 2](#pone-0081877-t002){ref-type="table"}). 10.1371/journal.pone.0081877.t002 ###### Odds ratios (95% CI) for healthy lifestyles and cognitive impairment, dementia chronic diseases and all-cause death adjusted for age and social class for each outcome (number of incident events). {#pone-0081877-t002-2} Healthy lifestyles (Number of behaviours) Numbers (%) of men (Total 2,235) Diabetes (214 men) Vascular Disease (752 men) Cancer (648 men) Cognitive Impairment (219 men) Dementia (79 men) All-cause Deaths (1,208 men) ----------------------------------------------- ---------------------------------- -------------------- ---------------------------- ------------------ -------------------------------- ------------------- ------------------------------ None 179 (8%) 1.00 1.00 1.00 1.00 1.00 1.00 1.35 0.85 1.00 0.69 0.80 0.78 Any one 702 (31%) (0.79--2.31) (0.60--1.19) (0.71--1.47) (0.32--1.49) (0.23--2.77) (0.55--1.12) 0.84 0.58 0.88 0.47 0.61 0.64 Any two 814 (36%) (0.48--1.45) (0.41--0.82) (0.60--1.23) (0.22--1.00) (0.18--2.10) (0.45--0.92) 0.62 0.53 1.00 0.36 0.32 0.48 Any three 429 (19%) (0.33--1.16) (0.37--0.77) (0.68--1.48) (0.16--0.81) (0.08--1.24) (0.33--0.71) 0.50 0.50 0.87 0.36 0.36 0.40 Four or five[\*](#nt109){ref-type="table-fn"} 111 (5%) (0.19--1.31) (0.30--0.84) (0.51--1.48) (0.12--1.09) (0.07--1.99) (0.24--0.67) Significance of trend \<0.0005 \<0.0005 0.60 0.001 0.02 \<0.0005 Data for the two men following all five behaviours have been included with those following four. All the relationships have been adjusted for age and social class, and those for cognitive impairment and dementia have additionally been adjusted for NART (27) score at baseline. Vascular disease includes ischaemic heart disease and ischaemic stroke. The risk of diabetes declined with increasing numbers of healthy behaviours followed, up to an OR of 0.50 (95% CI: 0.19, 1.31, [Table 2](#pone-0081877-t002){ref-type="table"}) with four or five behaviours, and the trend with increasing numbers of behaviours was significant (P\<0.0005). For vascular disease the risk was decreased up to an OR of 0.50 (95% CI: 0.30, 0.84) with four or five behaviours, and again the trend was significant (P\<0.0005). Cancer incidence was not related to lifestyle although there was a reduction in cancer associated with non-smoking alone (OR: 0.65; 95% CI 0.54, 0.79, [Table 1](#pone-0081877-t001){ref-type="table"}). There was a significant association between lifestyle and all-cause mortality, with an OR of up to 0.40 (95% CI: 0.24, 0.67; P for trend \<0.0005) with four or five behaviours. Cognitive function was assessed in 1,225 men (75% of the survivors), and examination of the medical records of the other men identified a further 219 with evidence of impairment. Seventy-nine of these men were judged by two experienced clinicians to have dementia. After adjustment for age, social class and NART, the OR for cognitive impairment in men following four or five behaviours was 0.36 (95% CI: 0.12, 1.09; P for trend \<0.001), and for dementia was 0.36 (95% CI: 0.07, 1.99; P for trend \<0.02, [Table 2](#pone-0081877-t002){ref-type="table"}). A detailed examination of relationships with individual behaviours suggested that exercise is an important predictor of both cognitive impairment (OR 0.64 95% CI 0.41, 0.92; P\<0.04, [Table 1](#pone-0081877-t001){ref-type="table"}) and dementia (OR 0.41 95% CI 0.22, 0.77; P\<0.005). A sensitivity analysis was conducted by omitting 896 men with evidence of diabetes or vascular disease at baseline. The association with all-cause mortality remained unchanged. In the men following four or five of the behaviours, the OR for diabetes decreased from 0.47 to 0.27 (95% CI: 0.06, 1.26; P for trend \<0.002), and for vascular disease from 0.50 to 0.46 (95% CI: 0.22, 0.93; P for trend \<0.0005). A further sensitivity analysis was conducted with cognitive outcomes by omitting 144 men with evidence of early cognitive decline. This included 26 men with dementia. Due to the reduced numbers of men with dementia the upper two healthy behaviour groups were combined to form a three, four or five healthy behaviour group. For these men the OR for cognitive impairment was 0.48 (95% CI: 0.30, 0.79; P for trend \<0.001) before omitting men with evidence of early impairment and 0.52 (95% CI: 0.30, 0.93; P for trend \<0.001) after omitting men with evidence of early decline. For dementia, in men with between three and five healthy behaviours the OR was 0.40 (95% CI: 0.18, 0.86; P for trend \<0.007) before omitting men with evidence of early impairment and 0.47 (95% CI: 0.18, 1.23; P for trend \<0.05) after omitting men with evidence of early decline. Over time there were changes in the uptake of the behaviours leading to a 'dilution' of the associations with disease. When questioned in the repeated follow-up examinations 1,023 men did not report substantial changes in the behaviours they followed. Amongst those who consistently reported either four or five healthy behaviours the trends with increasing numbers of behaviours remained significant. For diabetes the OR was 0.27 (95% CI: 0.04, 1.66; P for trend \<0.002), for vascular disease the OR was 0.22 (95% CI: 0.05, 0.89; P for trend \<0.0005), and for all-cause death the OR was 0.46 (95% CI: 0.14, 1.54; P for trend \<0.001). For cognitive outcomes, due to small numbers 3--5 behaviours groups were collapsed, in which the OR for impairment was 0.82 (95% CI: 0.50, 1.36) and the OR for dementia was 0.75 (95% CI: 0.35, 1.61). Rate advancement periods were estimated. For vascular disease, men following two, three and four healthy behaviours had delays of approximately 9.3, 10.8 and 11.9 years, respectively. For mortality the delays were 2.5, 4.6 and 6.3 years respectively. Finally, the uptake of healthy behaviours and healthy lifestyles within the cohort in 1979 were compared with estimates of uptake in 2009 based on data from the Welsh Health Survey [@pone.0081877-Anonymous1]. Individual healthy behaviour uptake changed over time: smoking declined, more fruits and vegetables were consumed, overweight increased and exercise declined. The adoption of a healthy lifestyle by men was however low and appears not to have changed during the subsequent 30 years, with under 1% of men following all five of the behaviours and 5% reporting four or more in 1979 and in 2009 ([Table 3](#pone-0081877-t003){ref-type="table"}). 10.1371/journal.pone.0081877.t003 ###### Uptake of behaviours in the Caerphilly Prospective study and in the Welsh Health Survey [@pone.0081877-Anonymous1]. {#pone-0081877-t003-3} Healthy behaviours in 1979 1979 2009 2009 ------------------------------ -------------------------------------------- --------------- ---------------- **Number of subjects** 2,235 men 1,927 men 15,810 persons **Non smoking** 1,017 *(46%)* 1,405 *(73%)* 12,201 *(77%)* **BMI** (18 to under 25) 791 *(35%)* 461 *(24%)* 6,027 *(38%)* **Fruit & veg. consumption** 394 (18%)[\*](#nt112){ref-type="table-fn"} 646 *(34%)* 5,431 *(34%)* **Regular activity** 874 *(39%)* 680 *(35%)* 4,519 *(29%)* **Light/Mod drinking** 1,320 *(59%)* 825 *(43%)* 8,638 *(55%)* **Healthy Lifestyle** \- **No healthy behaviours** 179 *(8%)* 135 *(7%)* 845 *(5%)* \- **Any two behaviours** 814 *(36%)* 650 *(34%)* 5,251 *(33%)* \- **Any three behaviours** 429 *(19%)* 350 *(18%)* 3,008 *(19%)* \- **Any four behaviours** 109 *(5%)* 98 *(5%)* 1,042 *(7%)* \- **All five behaviours** 2 (0.1%) 9 *(0*.*5%)* 132 *(0.8%)* The criterion for fruit and vegetable intake had to be reduced to 3 portions a day for the Caerphilly cohort because only 15 men in the 1979 cohort consumed five portions per day. Discussion {#s4} ========== Within a representative sample of middle-aged men, the following of increasing numbers of healthy behaviours was associated with increasing reductions in several important chronic diseases and mortality: an estimated 50% reduction in diabetes, 50% in vascular disease and 60% for all-cause mortality. These results therefore confirm previous studies and provide further data on the association of lifestyle with cognitive impairment and dementia, with a reduction of about 60% in cognitive impairment and about the same in dementia. These reductions, and especially those in cognitive function, are of enormous importance in an ageing population. When the 'dilution' effect of changes in lifestyle during follow-up is allowed for, the relationships we describe are closely similar to those reported from other cohorts. Thus the subjects amongst the 43,000 US Health Professionals who had adopted the five healthy behaviours experienced an 87% reduction in heart disease (Relative risk (RR): 0.13 CI: 0.09--0.19) [@pone.0081877-Chiuve1] and among the 84,000 women in the US Nurses\' Health Study the risk of coronary events was reduced 85% (RR: 0.17; 95% CI: 0.07--0.41) by following the five behaviours [@pone.0081877-Stampfer1]. In another US study, a 61% reduction in type 2 diabetes was attributable to a low BMI alone [@pone.0081877-Hu1] and this compares with 69% less diabetes in men in the CaPS with a BMI at baseline within the guidelines. Similar effects have been reported in men and women in the European EPIC study: subjects following four healthy behaviours having approximately one-quarter the mortality of those who followed none (OR: 0.25; 95% CI: 0.18,0.34), equivalent to about 14 years difference in chronological age [@pone.0081877-Khaw1]. In a study of 4,886 adult British subjects, the following of four healthy behaviours led to hazard ratios of 0.29 (95% CI: 0.19, 0.43) for cardiovascular disease and 0.29 (95% CI: 0.19, 0.43) for death [@pone.0081877-Kvaavik1]. The absence of any reduction in cancer in CaPS, other than by non-smoking, is surprising. Other papers have shown reductions attributable to the other healthy behaviours, indeed in some the reduction is large, up to a hazard ratio of 0.30 (95% CI: 0.15, 0.60) [@pone.0081877-Ford2]. On the whole however the reduction in cancer appears to be highly variable and usually small. For example, in 112,000 non-smoking subjects, the 4% of subjects who achieved a high score based on body weight, activity, diet and alcohol intake, showed a reduction of only 14% in incident cancer (RR: 0.86; 95% CI: 0.78, 0.94) during a follow-up of 14 years [@pone.0081877-McCullough1]. The postponement of vascular disease by healthy living for up to 12 years and up to six years for death is of interest, and again, these estimates are derived from subjects, all but two of whom followed four, rather than all five behaviours. A report based on two thousand 60 year old college alumni, estimated that the onset of 'disability' in eight common daily activities was delayed 8.3 years by following three healthy behaviours [@pone.0081877-Chakravarty1]. A study based on subjects within the US NHANES cohort who had adopted four healthy behaviours, estimated a rate advancement of 11.1 years for all-cause mortality [@pone.0081877-Ford1], and a UK study of five thousand adults following four behaviours showed an all-cause mortality risk equivalent to being 12 years older than those following no healthy behaviour [@pone.0081877-Kvaavik1]. Of particular interest is the association of lifestyle with cognitive outcomes. Studies on elderly populations followed for short periods have generally shown an association between lifestyle and cognitive impairment [@pone.0081877-Buchman1] [@pone.0081877-Gow1] [@pone.0081877-Iwasa1] [@pone.0081877-Ruscheweyh1], but the likelihood of reverse causation makes the interpretation of such short --term studies difficult. In fact, few studies have followed middle-aged populations over extended periods. Within the Whitehall study obesity, alcohol and smoking have been show to affect cognitive function over 10 years [@pone.0081877-HaggerJohnson1], [@pone.0081877-SinghManoux1]. In the Honolulu-Asia Aging Study a healthy lifestyle was related to reduced risk of dementia over 25 years (OR 0.36, 95% CI 0.15, 0.84) [@pone.0081877-Gelber1]. Our data are consistent with these results. The fact that the significant trends in our study for cognitive impairment and dementia were retained after omitting men with evidence of early cognitive impairment suggests that reverse causation is unlikely to contribute much to the relationships we describe. Although it can still be argued that even 30 years follow-up may not fully remove the full impact of cognitive change on behaviour, early pathological changes that have been described decades prior to the onset of dementia do not necessarily lead to cognitive change [@pone.0081877-Alexopoulos1]. Reverse causation is unlikely to contribute much to the relationships with diabetes and vascular disease, or with mortality. These relationships remain significant following the omission of men with evidence of disease at baseline, and furthermore, omission of events during the first five years of follow-up was found to make no material difference to the associations (data not shown). An important limitation in the CaPS study is the fact that the full impact of healthy lifestyles was underestimated because of small numbers of men adhering to all five healthy behaviours. In addition, the full effect of dilution due to changes in behaviour that were not fully monitored constantly throughout follow-up. A strength of CaPS however is that surveillance of the subjects in CaPS was close throughout the study, and intense efforts were made to identify incident disease involving both questioning of the subjects and inspection of all the available medical records approximately every five years. These findings must be balanced against an earlier, more detailed analysis which, following extensive adjustment for health, social and other lifestyle factors, found no evidence for an independent association of physical activity with dementia [@pone.0081877-Morgan1]. This more detailed analysis suggests any benefit of physical activity on cognition is likely to be mediated via vascular metabolic and emotional pathways rather than being independent of these factors. While the independence of the relationship we describe can therefore be questioned, the public health message is still that regular physical activity is highly beneficial. Residual confounding cannot be discounted in these data; however, the analyses we present investigate overall effects of lifestyle, rather than attempt to identify causal pathways between lifestyle and health. Non-causal explanations include confounding by social factors such as education and marital status. Nevertheless, our findings confirm that there is a substantial health benefit associated with a healthy lifestyle. Writing about wellbeing in the community, Huppert [@pone.0081877-Huppert1] puts the case for a population approach, aiming for a small shift in an unhealthy behaviour, rather than focussing interventions on individuals who need help. Thus: assuming that the associations we report are causal and reversible, quite small increases in the uptake of healthy behaviours could considerably reduce the population burden of vascular disease, dementia and death. Had the two and a half thousand men in CaPS each been urged at baseline to adopt one additional healthy behaviour, and if only half them had complied, then during the following 30 years there would have been a 13% reduction in dementia, a 12% drop in diabetes, 6% less vascular disease and a 5% reduction in total mortality. The area chosen for the CaPS cohort studies was selected because the social class distribution of the residents had been shown to be similar to that of the UK [@pone.0081877-Elwood2], and it can reasonably be assumed to be similar to that of the whole of Wales, a substantial part of the UK. Comparisons with the Welsh Health Survey show that there has been little change in the prevalence of healthy lifestyles during the 30 years covered by CaPS ([Table 3](#pone-0081877-t003){ref-type="table"}). During this period it has been estimated that unhealthy living accounts for 10% of the costs of the National Health Service in Wales, while the annual expenditure on prevention and public health services in Wales has been estimated to have been around £ 280 millions (USD 436 millions) [@pone.0081877-Hale1]. Despite this, and despite increasing knowledge of the relevance of lifestyle to health and to survival, the proportion of the adult Welsh population following all five healthy behaviours was, and remains under 1%. At the same time, the prevalence of all five healthy behaviours within health workers in the USA, Nurses [@pone.0081877-Stampfer1] and other Health Professionals [@pone.0081877-Chiuve1], is estimated to be only 3%. Clearly there is an urgent need for new strategies in health promotion to be developed and evaluated. The costs of health services are increasing globally, and are likely to become unsustainable unless members of the public become more fully engaged and take a greater responsibility for their own health [@pone.0081877-Elwood3]. Personal prevention measures, such as we describe, could have a large impact on the costs of healthcare services [@pone.0081877-Huppert1]. Ultimately however, decisions about behaviours lie with the individuals and there is therefore an urgent need to establish a more effective partnership between health services and citizens. The Caerphilly Prospective Study was conducted by the former MRC Epidemiology Unit (South Wales). The Caerphilly archive is now maintained by the School of Social and Community Medicine in Bristol University. We thank the Medical Research Information service of the National Health Service Information Centre for helping us maintain long term follow-up with the cohort. We thank all the men who have given their time to be participants in CaPS. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: PE J. Galante AB YB-S. Performed the experiments: PE J. Galante AB. Analyzed the data: PE JP J. Gallacher. Contributed reagents/materials/analysis tools: J. Galante SP YB-S ML J. Gallacher. Wrote the paper: PE J. Galante JP ML J. Gallacher.

INTRODUCTION {#s1} ============ Humans orient their bodies vertically by integrating visual, somatosensory, and vestibular system information in their gravity environment[@r1]^)^. The important cognitive aspects related to this vertical orientation are the subjective visual vertical (SVV), the subjective postural vertical-eyes open (SPV-EO), and the subjective postural vertical (SPV)[@r2],[@r3],[@r4]^)^. The mean bias demonstrating the direction of tilt (hereafter "tilt direction") and the standard deviation demonstrating instability in the perception of verticality (hereafter "variability") are considered important indicators of the characteristics of the above-described perceptions of verticality. These parameters have been indicated to change as a result of cerebrovascular accidents and other neurological disorders[@r2],[@r3],[@r4],[@r5]^)^. Declines in postural balance and activities of daily living are also indicated to be associated with these parameters[@r6],[@r7],[@r8]^)^. Recently, SPV in the sagittal plane has been shown to deviate backward with age[@r9]^)^. Furthermore, in a study that investigated SPV in the sagittal plane on the influence of starting direction, the characteristics of this tilt were indicated to differ in young adults and elderly adults[@r9]^)^. Although this result indicates that already-acquired verticality perception declines with age, age-related changes in the subjective vertical and starting direction characteristics in the frontal plane have not been determined. The purpose of the present study was to determine age-related differences in the subjective vertical perception in the frontal plane in healthy adults. SUBJECTS AND METHODS {#s2} ==================== The subjects comprised 13 young adults (age: 25.1 ± 2.3 years \[mean ± SD\], 22--30 years \[range\]; 6 men, 7 women; all right-handed) and 13 elderly adults (age: 67.0 ± 5.1 years, 60--74 years; 7 men, 6 women; all right-handed). The inclusion criteria consisted of no past history of bone and joint disease, neurological disorders, psychiatric disorders, or dementia. All subjects in the elderly group walked without a cane and had no history of falls. The present study was conducted with the approval of the Saitama Medical University International Medical Center Institutional Review Board (approval number: 14-117). Recruitment for study participation was conducted openly. All subjects received an explanation of the study and provided informed consent in writing. SVV was measured using computer software ([Fig. 1](#fig_001){ref-type="fig"}Fig. 1.Assessment of subjective visual vertical. These two computers were linked with an USB cable so that the image displayed on the computer controlling the SVV was also displayed on the computer screen that the subjects watched. a cylindrical tube was placed in such a manner that the vertical portion of the screen frame could not provide clues to verticality.). Measurements were conducted with the subjects seated and their feet flat. Two computers were used for measurements. These two computers were linked with an USB cable so that the image displayed on the computer controlling the SVV was also displayed on the computer screen that the subjects watched. In accordance with the method described by Pavan et al.[@r10]^)^, a cylindrical tube was placed in such a manner that the vertical portion of the screen frame could not provide clues to verticality. The subjects' trunks were fixed, and their feet were flat. The visual indicator was at the subject's eye level, 50 cm in front of them. Subjects watched the visual indicator through the cylindrical tube placed in front of the computer screen. The SVV was controlled as follows. The experimenter rotated the visual indicator from a horizontal position leftward or rightward toward a vertical position at a rate of 5°/s. The rotation was stopped when the subject verbally reported that he had perceived true verticality, and the deviation from true verticality was recorded. Eight trials were performed in an ABBABAAB sequence. SPV-EO and SPV were measured using a vertical board (VB) that had a semicircular rail attached to the bottom ([Fig. 2](#fig_002){ref-type="fig"}Fig. 2.Measurement of subjective postural vertical and subjective postural vertical-eyes open. SPV-EO and SPV were measured using a vertical board (VB) that had a semicircular rail attached to the bottom.). The sides and backs of the subjects' trunks were covered with non-stretchable cloth, and the subjects sat on the VB with their feet not in contact with the ground. The subjects' trunks were fixed, their arms were crossed in front of their chest, and the positions of their heads and legs were not fixed. The VB was controlled by two experimenters. The experimenters rotated the seat of the VB from a position of 15° or 20° tilt toward a vertical position at a rate of approximately 1.5°/s. The tilt of the seat when the subjects verbally reported that he had reached a true verticality was recorded with a digital inclinometer. Eight trials were performed in an ABBABAAB sequence so that the starting position and angle would be pseudo-random. Trials for SPV-EO and SPV were conducted with the subjects' eyes open and closed, respectively. A true vertical position was considered 0°, while rightward tilt and leftward tilt were treated as positive and negative, respectively. The mean (Tilt direction) and the SD (Variability) of the eight trials were calculated. In order to determine the starting point effect, means were calculated for the four trials that started from the right (Right side position) and the four trials that started from the left (Left side position), respectively. Verticality perception parameters were compared between the young group and the elderly group using the unpaired t-test. Statistics were processed using PASW Statistics ver. 18.0 (SPSS Inc., Tokyo, Japan), with the level of significance set at 5%. RESULTS {#s3} ======= Results for tilt direction and variability are shown in [Table 1](#tbl_001){ref-type="table"}Table 1.Results for tilt direction and variabilityVariableYoung (n=13)Elderly (n=13)p valueTilt directionSVV--0.3 ± 1.3--0.5 ± 1.8SPV-EO0.5 ± 0.50.5 ± 0.9SPV0.1 ± 0.6--0.1 ± 1.1VariabilitySVV0.9 ± 0.41.1 ± 0.5SPV-EO1.7 ± 0.82.9 ± 0.8\*SPV1.9 ± 0.83.2 ± 1.1\*SVV: subjective visual vertical; SPV-EO: subjective postural vertical-eyes open; SPV: subjective postural vertical, \*p\<0.05. Significant differences in tilt direction were not observed in any parameter. In terms of variability, no difference was observed between young and elderly subjects in SVV. However, the elderly group demonstrated significantly higher variability in SPV-EO and SPV (p\<0.05). [Table 2](#tbl_002){ref-type="table"}Table 2.Results for starting point effectVariableYoung (n=13)Elderly (n=13)p valueRight side positionSVV--0.7 ± 1.3--0.7 ± 1.6SPV-EO1.7 ± 1.22.8 ± 1.3\*SPV1.6 ± 1.12.7 ± 1.5\*Left side positionSVV0.1 ± 1.20.1 ± 1.8SPV-EO--0.8 ± 1.3--2.0 ± 1.3\*SPV--1.3 ± 1.2--2.9 ± 1.5\*SVV: subjective visual vertical; SPV-EO: subjective postural vertica-eyes open; SPV: subjective postural vertical, \*p\<0.05 indicated results for starting point effects. In SVV, there was no difference between the groups in right and left side position, while the elderly group demonstrated significant bias against starting position in SPV-EO and SPV (p\<0.05). In addition, the tilt of SPV-EO and SPV tended to occur in the starting direction. DISCUSSION {#s4} ========== In the present study, examinations of differences in the subjective vertical in the frontal plane between elderly adults and young adults revealed that elderly adults demonstrated significantly greater variability in SPV-EO and SPV, as well as greater bias toward the starting direction. Mean tilt direction was not significantly different between the young and elderly groups. Pérennou et al.[@r4]^)^ have reported that the normal range of SVV and SPV in healthy adults is −2.5° to 2.5°. While nothing was demonstrated with regard to the normal range of SPV-EO, all of the verticality perception data obtained in the present study was near perfect. Although visual, somatosensory, and vestibular system functions have long been known to decline with age, it was indicated that these declines do not bring about any specific directional abnormalities in verticality perception. As for variability, although there was no difference between young and elderly subjects in SVV, the elderly group demonstrated significantly greater variability in SPV-EO and SPV. Saeys et al.[@r11]^)^ investigated the association between SPV and somatosensory system function in stroke patients. They reported that deviation in SPV is related to somatosensory loss and is associated not with deep sensation, but with superficial sensation. In addition, Bisdorff et al.[@r5]^)^ demonstrated that while SPV tilt direction is perfectly vertical in patients with vestibular disorders, SPV variability in patients with vestibular disorder is significantly greater among elderly patients. Bergmann et al.[@r12]^)^ examined SPV in the sagittal and frontal planes during standing. They found that elderly subjects demonstrated significantly greater variability than young subjects in the sagittal plane; however, they found no difference between groups in SPV variability in the frontal plane. In the present study, the elderly group demonstrated significantly higher SPV variability while seated; this result differs from previous studies, which found no difference in variability while standing. These divergent results indicate that information processing for verticality perception differs between standing and sitting. In examinations of verticality perception using starting direction, the young and elderly groups demonstrated no difference in SVV; however, for SPV-EO and SPV, both young and elderly subjects tilted in the starting direction. These findings on SPV-EO and SPV are consistent with the results reported by Mazibrada et al.[@r13]^)^ reported that SPV was slightly deviated to stating position in normal subjects. This phenomenon of verticality perception tilting in the direction of postural tilt may demonstrate the Aubert effect. Barbieri et al.[@r9]^)^ examined verticality perception in the sagittal plane by measurement of the starting direction. In young subjects, SPV tilted backward when measurement started in a backward position relative to the vertical, while SPV tilted forward when measurement started in a forward position. The phenomenon of these subjects' perceptions of the postural vertical conforming to the starting direction was surmised as a possible manifestation of the Aubert effect. In elderly adults, however, SPV-EO and SPV were tilted backward regardless of the starting direction. In the present study, SPV-EO and SPV were tilted in the measurement starting direction regardless of age; therefore, the frontal plane was surmised to possess tilt characteristics that differ from those in the sagittal plane. Additionally, the greater bias toward the starting direction in the elderly group than in the young group may have been due to age-related degenerative changes in vestibular and somatosensory function. The vestibular and somatosensory systems have long been indicated to have great influence on the Aubert effect[@r14],[@r15],[@r16]^)^. Degenerative changes in vestibular and somatosensory function have been reported in elderly individual[@r17], [@r18]^)^; thus, elderly subjects may have been more affected by the Aubert effect than young individuals. In contrast, SVV is measured in an upright position, which may have nullified the Aubert effect. Based on the above results, tilt direction in SPV-EO and SPV were offset by left and right bias, thus indicating that it is necessary to consider different analysis methods for different starting directions. Moreover, variability in verticality perception is believed to reflect the characteristics of the starting direction. It is conceivable that this is why the elderly group demonstrated greater bias toward the starting direction than did the young group, therefore demonstrating greater variability. The present study showed that age increases variability in the postural vertical and results in greater bias toward starting direction. This finding is considered important, as it suggests that starting direction characteristics and age-related changes should be taken into account when assessing the characteristics of verticality perception in patients with cerebrovascular accidents and other neurological disorders. One limitation of the present study is that the sample size was small. Additionally, the subjects comprised adults aged 20--39 years and 60--79 years; therefore, the characteristics of verticality perception in adults aged 40--59 years remains unknown. Going forward, it may be necessary to include larger numbers of subjects and investigate verticality perception characteristics with respect to age. In conclusion, the present study examined age-related differences in the subjective vertical in the frontal plane in healthy individuals. The results demonstrated that elderly adults show greater bias in the postural vertical than do young adults, suggesting that the postural vertical declines with age. We hope to use the present study as a foundation for determining the characteristics of the subjective vertical in patients with cerebrovascular accidents. Conflict of interest -------------------- The authors declare no conflicts of interest. Funding ------- No financial support was received for this study.

1. Introduction {#sec1-materials-13-02377} =============== Physical properties of Ce-based intermetallic compounds are mainly determined by two competing interactions: Kondo effect, characterized by a temperature $T_{K} \propto exp\left( - \frac{1}{\mid J_{fc}N\left( E_{F} \right) \mid} \right)$, and Ruderman--Kittel--Kasuya--Yosida (RKKY) interaction, related to $T_{RKKY} \propto J_{fc}^{2}N\left( E_{F} \right)$. In both expressions, $N\left( E_{F} \right)$ stands for density of states (DOS) at Fermi level $E_{F}$, and $J_{fc} \sim V_{fc}$ is the coupling constant between $4f$ and conduction (c) electron states, where $V_{fc}$ represents on-site hybridization energy given by *f*--*c* hybridization matrix element. According to the Schrieffer--Wolff transformation \[[@B1-materials-13-02377]\], $J_{fc}$ is defined as $J_{fc} = \frac{2V_{fc}^{2}}{|E_{f} - E_{F}|}$, where $E_{4f}$ stands for energy of $4f$ level. The energy $V_{fc}$ determines filling of the $4f$ shell, and thus governs the character of magnetic ground state. In the Ce-based intermetallics, the hybridization $V_{fc}$ results in a variety of intriguing properties such as heavy-fermion behavior, unconventional superconductivity, various magnetic ordering, non-Fermi liquid, and quantum critical phenomena \[[@B2-materials-13-02377]\]. For a number of Ce-based compounds reported in the literature, Ce ions occupy a single position in their crystallographic unit cells. If the $4f$-electron states are strongly localized, i.e., the Kondo interaction is weak, generally, a kind of magnetic ground state is expected. Most often, the compounds order antiferromagnetically, yet, a few ferromagnets are also known, e.g., Ce${}_{2}$RuGe${}_{2}$ \[[@B3-materials-13-02377]\], CeRuPO \[[@B4-materials-13-02377]\], Ce${}_{3}$RhSi${}_{3}$ \[[@B5-materials-13-02377]\], CePd${}_{2}$Al${}_{8}$ \[[@B6-materials-13-02377]\], CeCrGe${}_{3}$ \[[@B7-materials-13-02377]\], or Ce${}_{11}$Pd${}_{4}$In${}_{9}$ \[[@B8-materials-13-02377]\]. However, the situation becomes less obvious when there is more than a single inequivalent Ce site in the crystal structure. Different local environments of the Ce ions can lead to dissimilar hybridization strengths, which spark the possibility of having distinctly different ground states for each individual inequivalent Ce ion. Recently, investigation of Ce-based compounds bearing multiple inequivalent Ce sites has received considerable attention, and for a few of them, diverse unusual low-temperature properties were established. Prominent examples are Ce${}_{5}$Ni${}_{6}$In${}_{11}$, with separate antiferromagnetic orderings in two different Ce atom sublattices \[[@B9-materials-13-02377]\]; Ce${}_{3}$Pd${}_{20}$Si${}_{6}$, with dipolar and quadrupolar antiferromagnetic orders associated with inequivalent Kondo sites \[[@B10-materials-13-02377]\]; or Ce${}_{3}$PtIn${}_{11}$ and Ce${}_{3}$PdIn${}_{11}$, where two different Ce atom sublattices host antiferromagnetism and heavy-fermion superconductivity \[[@B11-materials-13-02377],[@B12-materials-13-02377],[@B13-materials-13-02377],[@B14-materials-13-02377]\]. Another exciting case is the coexistence in a single material of long-range magnetic ordering and valence fluctuations, each phenomenon emerging in a separate Ce atom sublattice. Recently, this very rare situation was reported to occur, e.g., in Ce${}_{2}$RuGe \[[@B15-materials-13-02377]\], with two independent Ce atom sites in its crystallographic unit cell, and Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$, which possesses as many as three inequivalent Ce atom sublattices \[[@B16-materials-13-02377],[@B17-materials-13-02377]\]. Remarkably, in both compounds one of the different Ce atoms is coordinated by its Ru neighbors at a distance of $\sim 2.2$ Å and $\sim 2.4$ Å, respectively, which is much shorter than the sum of the covalent radii of the Ce and Ru atoms. As discussed in detail in a series of our previous papers on the Ce-Ru-*X* intermetallics (*X* = Ge, Ga, Al) \[[@B15-materials-13-02377],[@B16-materials-13-02377],[@B17-materials-13-02377],[@B18-materials-13-02377],[@B19-materials-13-02377],[@B20-materials-13-02377],[@B21-materials-13-02377],[@B22-materials-13-02377]\], strong Ce--Ru bonding brings about a significant instability of the electronic $4f$ shell, and thus intermediate valence behavior may arise. At the same time, the Ce ions with Ce-Ru distances of regular length remain their trivalent character that promotes localized magnetism with possible magnetic ordering at low temperatures. The present research was aimed at verification of the electronic character of the particular Ce ions in Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$ by means of X-ray photoelectron spectroscopy and ab initio band structure calculations. Our results fully support the scenario of the dual nature of the $4f$ electrons in this material. 2. Experimental and Computational Details {#sec2-materials-13-02377} ========================================= X-ray photoelectron spectroscopy (XPS) experiments were carried out on a polycrystalline sample Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$ used before for magnetization, magnetic susceptibility, specific heat, and resistivity measurements \[[@B17-materials-13-02377]\]. The XPS spectra were obtained at room temperature in vacuum of \~$10^{- 10}$ Torr using a Physical Electronic PHI 5700/600 ESCA spectrometer (Physical Electronics, Inc., Chanhassen, MN, USA) with monochromatized Al K$\alpha$ radiation. The sample was broken in high vacuum of $6 \times 10^{- 10}$ Torr, immediately before the spectra were recorded. Calibration of the spectral data was performed in a manner described in \[[@B23-materials-13-02377]\]. Binding energies were referenced to the Fermi level ($E_{F} = 0$). The electronic band structure of Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$ was calculated using the full-potential linearized augmented plane waves (FP-LAPW) method \[[@B24-materials-13-02377]\] implemented in the WIEN2k computer code (WIEN2k_18.1, released on 30 June 2018, Institute of Materials Chemistry, TU Viena, Austria) \[[@B25-materials-13-02377]\] (for details on similarly made computations see, e.g., in \[[@B26-materials-13-02377]\]). In the performed calculations, we assumed the following electronic configurations of strongly-bound core level states (SC), weakly-bound states (WC), and valence band states in the particular atoms; Ce: \[Kr\]${}_{SC}${$4d^{10}5s^{2}5p^{6}$}${}_{WC}$($4f^{1}5d^{1}6s^{2}$)${}_{VB}$; Ru: \[Ar + $3d^{10}$\]${}_{SC}${$4s^{2}4p^{6}$}${}_{WC}$($4d^{7}5s^{1}$)${}_{VB}$; and Ga: \[Ne+$3s^{2}$\]${}_{SC}${$3p^{6}3d^{10}$}${}_{WC}$($4s^{2}4p^{1}$)${}_{VB}$. The fully relativistic formalism was implemented for the SC states, while local orbital (LO) states and VB electrons were treated within the scalar-relativistic Kohn--Sham approach. The spin-orbit (SO) interaction was applied within the second variational approach \[[@B24-materials-13-02377]\] for calculation of the VB and LO states. The revised Perdew--Burke--Ernzerhof (PBEsol) generalized gradient approximation (GGA) \[[@B27-materials-13-02377]\] was applied for the exchange correlation (XC) potential. To determine theoretically the magnetic properties of the individual Ce ions in Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$, the following procedure was applied. First, the PBEsol XC potential was corrected by the Hubbard-like correlation interaction using the approach developed by Anisimov et al. \[[@B28-materials-13-02377],[@B29-materials-13-02377]\] with correlation energy parameter *U* = 1.5, 2.25, and 3 eV \[[@B30-materials-13-02377]\]. Then, the ab initio calculations were made within FP-LAPW approach, assuming the muffin-tin (MT) model for crystal potential. The radii of MT spheres, $R_{MT}$, were taken equal 0.121 nm, 0.101 nm, and 0.111 nm for Ce, Ru, and Ga ions, respectively. The accuracy of the performed calculations was determined by the following parameters; $l_{\max} = 10$, $G_{\max} = 14$, and $K_{\max} = 9/R_{MT} \simeq 8.17\mspace{600mu}{nm}^{- 1}$. A number of 324$\overset{\rightarrow}{k}$ vectors in the irreducible Brillouin zone used in the calculations was found to ensure a total energy convergence of the order of 0.01 eV. The structural data assumed in the initial calculations were taken from work in \[[@B16-materials-13-02377]\]; however, an atomic relaxation was performed to reach the equilibrium structure. [Figure 1](#materials-13-02377-f001){ref-type="fig"} shows the crystal structure of Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$, which was the basis for our calculations. In the crystallographic unit cell, there are three inequivalent Wyckoff positions for cerium atoms: $2a$ site with Ce1 atoms, $8d$ site with Ce2 atoms, and another $8d$ site with Ce3 atoms \[[@B16-materials-13-02377]\]. Throughout the present paper we adopted the Ce atoms labels introduced in Table 2 in Ref. \[[@B16-materials-13-02377]\]. One should note, however, that in the text of the latter publication and in its figures the Ce1 atom was mistakenly switched with the Ce2 atom (we thank Dr. Elena Murashova, a coauthor of Ref. \[[@B16-materials-13-02377]\], for giving us comprehensive information about that error). 3. XPS Results {#sec3-materials-13-02377} ============== The X-ray absorption near-edge structure (XANES) spectroscopy, performed for Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$ near its Ce $L_{3}$ edge, revealed a mixed valence state of the Ce ions, giving an average valence of Ce ions to be about 3.1 at room temperature \[[@B16-materials-13-02377]\]. In order to corroborate that finding, we measured Ce $3d$ and Ce $4d$ core-level XPS spectra and analyzed the results in terms of the Anderson theory \[[@B31-materials-13-02377]\]. For a system with partial filling of the Ce $4f$ shell, the theory predicts the appearance of the $f^{0}$ and $f^{2}$ final states as a result of intra-atomic hybridization between $4f$ and conduction band states. The $3d$ XPS spectrum recorded at room temperature is presented in [Figure 2](#materials-13-02377-f002){ref-type="fig"}a. The main lines correspond to the $3d_{5/2}^{9}4f^{1}$ and $3d_{3/2}^{9}4f^{1}$ final states, separated by spin-orbit (SO) interaction $\Delta_{SO} = 18.6$ eV. Most remarkably, the spectrum also shows satellites $3d_{5/2}^{9}4f^{n}$ and $3d_{3/2}^{9}4f^{n}$ with *n* = 0 and 2, separated by the same energy $\Delta_{SO}$. According to the Gunnarsson--Schönhammer (GS) model \[[@B32-materials-13-02377],[@B33-materials-13-02377]\], the $3d4f^{0}$ line arises due to the intermediate valence effect, whereas $3d4f^{2}$ reflects the on-site hybridization strength, which is expressed by the energy $\Delta_{fc} = \pi V_{fc}^{2}N\left( E_{F} \right)$ \[[@B31-materials-13-02377]\]. It is possible to separate of the overlapping peaks on the basis of the Doniach--Šunjić theory \[[@B34-materials-13-02377]\], and $\Delta_{fc}$ can be estimated from the intensity ratio $I\left( f^{2} \right)/\left\lbrack I\left( f^{1} \right) + I\left( f^{2} \right) \right\rbrack$ of the respective Ce $3d$ XPS lines \[[@B33-materials-13-02377]\]. In turn, the intensity ratio $r = I\left( f^{0} \right)/\left\lbrack I\left( f^{0} \right) + I\left( f^{1} \right) + I\left( f^{2} \right) \right\rbrack$ gives an estimate for the $4f$ shell mean occupation number $n_{f}$ \[[@B33-materials-13-02377]\]. The accuracy of determining $\Delta_{fc}$ and $n_{f}$ is usually less than 20% \[[@B35-materials-13-02377],[@B36-materials-13-02377]\] (the limitations were discussed in details, e.g., in \[[@B33-materials-13-02377]\]). Moreover, one should note that these two quantities are interrelated. In the case of Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$, we found from the GS approach $\Delta_{fc} \approx 210$ meV. In order to determine the ground-state $4f$ occupation, we used the theoretical method proposed by Fuggle et al. in Ref. \[[@B33-materials-13-02377]\], where the *r* ratio is calculated as a function of the initial *f* occupation $c_{(f^{0})}$ for different values of $\Delta_{fc}$. Assuming $n_{f} \approx 1 - c_{(f^{0})}$ and $c_{(f^{0})}$ equal to wave function amplitude of the initial $f^{0}$ configuration state \[[@B33-materials-13-02377]\], we derived the fractional $4f$ electron count $n_{f} \approx 0.88$. The fractional valence of Ce ions in Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$ was further corroborated by inspection of the Ce $4d$ XPS spectrum (see [Figure 2](#materials-13-02377-f002){ref-type="fig"}b), which exhibits two lines near 120 and 124 eV, characteristic of the Ce${}^{4 +}$ states \[[@B33-materials-13-02377]\]. 4. Calculated Electronic Structure {#sec4-materials-13-02377} ================================== The atomic positions in the crystallographic unit cell of Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$, obtained as a result of minimizing interatomic forces, are presented in [Table 1](#materials-13-02377-t001){ref-type="table"}, and the so-derived local environments of the Ce1, Ce2, and Ce3 atoms are given in [Table 2](#materials-13-02377-t002){ref-type="table"}. All the respective interatomic distances are very similar to those reported in the literature \[[@B16-materials-13-02377]\] (see also Ref. citeremark ). The electronic bands in Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$, calculated assuming the correlation energy $U = 1.5$ eV and 2.25 eV, are shown in [Figure 3](#materials-13-02377-f003){ref-type="fig"} in a form of the total density of states (TDOS). In addition, the calculations were performed for a model in which different *U* values were attributed to distinct Ce atoms, and [Figure 3](#materials-13-02377-f003){ref-type="fig"} displays the result obtained setting $U = 3$ eV for the Ce1 atom and $U = 2.25$ eV for the Ce2 and Ce3 atoms. As can be inferred from the figure, the DFT data hardly depend on *U*, except a narrow range of binding energies $- 1$ eV $< E < E_{F}$. [Figure 4](#materials-13-02377-f004){ref-type="fig"}a shows the spin-resolved TDOS calculated for the latter values of *U* compared with the valence band of of Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$ determined experimentally. [Figure 4](#materials-13-02377-f004){ref-type="fig"}b, with an expanded energy scale, presents the same theoretical and XPS data together with the partial TDOS due to the particular atoms in the unit cell. Clearly, all the features present in the XPS spectrum are properly reproduced in the computed data. The main contribution due to the Ru $4d$ states is distributed between $E_{F}$ and the binding energy of 4 eV. In turn, the Ga 4*p* states form bands located near the binding energy of about 6 eV. The Ce $4f$ states are responsible for a broad and fairly weak feature near $E_{F}$. At the binding energy of about 17 eV and 19 eV, the calculated Ce $5p$ electronic states show SO-separated features, which are displaced in respect to the measured data by \~1 eV. The discrepancy can be attributed to Ce $5d$-electron correlations, which usually shift the calculated Ce $5p$ states to lower binding energies \[[@B26-materials-13-02377],[@B37-materials-13-02377],[@B38-materials-13-02377]\]. The main numerical results of the performed PBEsol+U calculations are listed in [Table 3](#materials-13-02377-t003){ref-type="table"}. For various *U*, the mean occupancy of the 4f shell of the Ce1 atoms is close to 1, while $n_{f}$ computed for both the Ce2 and Ce3 atoms is notably smaller than 1. The effect of *U* on the obtained 4f electron count is almost negligible. Taking into account the multiplicity of the particular Ce sites, one obtains an average filling of the 4f shell in Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$ equal to 0.86--0.89, in perfect agreement with the experimental result $n_{f} \approx 0.88$ (see above). The calculated total magnetic moment $m_{Ce1}$ amounts to about 1 $\mathsf{\mu}_{B}$, regardless of the value of *U*, whereas the magnetic moment found at the Ce2 and Ce3 sites is significantly smaller, namely, $m_{Ce2} \approx 0.3$ $\mathsf{\mu}_{B}$ and $m_{Ce3} \approx 0.5$ $\mathsf{\mu}_{B}$. The fractional valence of the Ce2 and Ce3 ions likely results from the on-site hybridization effect as well as some intersite hybridization between the Ce $4f$ and Ru *d*-electron states. As can be inferred from [Figure 5](#materials-13-02377-f005){ref-type="fig"}a--c, the on-site f--c hybridization causes a significant increase in the number of Ce2 and Ce3 5d-electron states, whereas the Ce1 4f electrons remain well localized at the binding energy of about 1 eV. At the same time, the DFT calculations clearly revealed strong inter-band hybridization of the Ce $5d$ and Ru $4d$ electron states for the Ce2 and Ce3 atoms, whereas the latter effect is negligibly small for the Ce1 atom (see [Figure 5](#materials-13-02377-f005){ref-type="fig"}d--f). In order to visualize the intersite hybridization and the atomic bonds in the unit cell of Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$, we calculated the charge densities (setting $U_{Ce1} = 3$ eV and $U_{Ce2,Ce3} = 2.25$ eV). [Figure 6](#materials-13-02377-f006){ref-type="fig"} displays the electron density map within the crystallographic (010) plane. The map clearly shows almost isotropic distribution of valence electrons around the Ce1 and Ga atoms. In contrast, the charge distribution near the Ce2, Ce3, and Ru atoms is strongly anisotropic with strong accumulation of the electronic density along the bonds Ce2-Ru and Ce3-Ru. The strongest covalent bonding occurs between the Ce2 and Ru atoms, in concert with the crystal structure refinement, which revealed abnormally short Ce2-Ru interatomic distance \[[@B16-materials-13-02377]\]. Thus, the DFT calculations fully corroborated the scenario developed before \[[@B16-materials-13-02377],[@B17-materials-13-02377]\], in which the IV behavior in Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$, evidenced in the spectroscopic and thermodynamic properties of the compound, can be associated primarily with the Ce2 atoms. 5. Conclusions {#sec5-materials-13-02377} ============== The XPS experiment performed for Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$ confirmed the fractional valence of the Ce ions, noticed before in the $L_{3}$ XANES spectroscopy \[[@B16-materials-13-02377]\] and bulk thermodynamic measurements \[[@B17-materials-13-02377]\]. The compound forms with a crystallographic unit cell that hosts three inequivalent Wyckoff positions for Ce atoms, thus the experimentally derived filling of the $4f$ shell ($n_{f} \approx 0.88$) was an average over those three sites. The DFT calculations allowed for inspecting the $4f$ electron counts at each Ce atom. The results indicated that the Ce1 ion located at the $2a$ site is trivalent ($n_{f}$ is close to 1). In contrast, the Ce2 and Ce3 ions, placed at the $8d$ sites, were found intermediate valent with $n_{f}$ notably smaller than 1. The ab initio calculated mean occupation of the $4f$ shell in Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$ is 0.86--0.89, which is in very good agreement with the experimental finding. The electronic instability of the $4f$ shell in the Ce2 ion gives rise to the IV character of the compound, established before in the study on its low-temperature bulk physical properties \[[@B17-materials-13-02377]\]. Most remarkably, the IV features were found to coexist with a long-range antiferromagnetic (AFM) ordering that sets in below $T_{N}$ = 3.7 K \[[@B17-materials-13-02377]\]. As suggested by our group in an earlier study \[[@B17-materials-13-02377]\] these two phenomena are spatially separated, i.e., they develop in different Ce ions sublattices. The present DFT results have corroborated such a scenario. Due to dissimilar strength of the intra-site band hybridization, the calculated magnetic moments are distinctly different for Ce1 (∼1 $\mathsf{\mu}_{B}$/atom), Ce2 (∼0.3 $\mathsf{\mu}_{B}$/atom), and Ce3 (∼0.5 $\mathsf{\mu}_{B}$/atom). Therefore, it is reasonable to attribute the AFM state principally to the Ce1 ions, with a possible contribution due to the Ce3 ions, while the Ce2 ions remain nonmagnetic. The energy $\Delta_{fc} \sim 200$ meV is for Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$ quite large, however, $V_{fc} = \left( \frac{\Delta_{fc}}{\pi N\left( E_{F} \right)} \right)^{1/2} \sim 44$ meV. One can also estimate the coupling constant $J_{fc} = \frac{2V_{fc}^{2}}{|E_{f} - E_{F}|} \approx 5$ meV \[[@B1-materials-13-02377]\] between the nearest Ce magnetic moments, which well correlates with low $T_{N}$ temperature-scale. In order to verify this tempting conjecture, neutron diffraction experiment is compulsory. Undoubtedly, the coexistence of intermediate valent and trivalent cerium ions makes Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$ an interesting material for further comprehensive experimental and theoretical investigations. We thank Elena Murashova for providing us with a polycrystalline sample of Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$, utilized in the XPS experiment. conceptualization, A.Ś.; J.D. and D.K.; methodology, A.Ś. and J.D; validation, A.Ś.; J.D. and D.K.; formal analysis, A.Ś.; investigation, A.Ś.; resources, A.Ś.; data curation, A.Ś.; writing---original draft preparation, A.Ś.; writing---review and editing, A.Ś.; J.D. and D.K.; visualization, A.Ś.; supervision, A.Ś., D.K.; project administration, A.Ś, D.K.; funding acquisition, A.Ś., J.D. All authors have read and agreed to the published version of the manuscript. This research received no external funding. The authors declare no conflict of interest. {ref-type="table"}.](materials-13-02377-g001){#materials-13-02377-f001} {#materials-13-02377-f002} {#materials-13-02377-f003} {#materials-13-02377-f004} {#materials-13-02377-f005} {#materials-13-02377-f006} materials-13-02377-t001_Table 1 ###### Relaxed atomic positions in Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$. Calculations were performed assuming the experimental lattice parameters $a = b = 10.1132$ Å and $c = 8.1212$ Å. Wyckoff Atom Coordinates --------- ------ ---------- ------------- ---------- $2a$ Ce1 0.000000 0.000000 0.141812 $8d$ Ce2 0.290732 0.000000 0.498706 $8d$ Ce3 0.287464 0.000000 0.871542 $8d$ Ru 0.347568 0.000000 0.203546 $2a$ Ga1 0.000000 0.000000 0.699153 $8c$ Ga2 0.217667 0.217667 0.178952 materials-13-02377-t002_Table 2 ###### Interatomic distances (Å) in the crystal structure of Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$. Ce1 Atom Distance Ce2 Atom Distance Ce3 Atom Distance ----- ------ ---------- ----- ------ ---------- ----- ------ ---------- 4Ga2 3.1355 Ru 2.4237 3Ru 2.8233 Ga1 3.3769 Ce3 3.0114 Ru 2.9531 4Ce3 3.4661 2Ru 3.0792 Ce2 3.0114 4Ru 3.6321 2Ce3 3.1721 2Ce2 3.1721 4Ce2 4.3802 2Ga2 3.2187 Ga1 3.2437 Ga1 4.7443 Ga1 3.3633 2Ga2 3.2791 2Ga2 3.5047 2Ga2 3.3838 Ce2 4.1355 Ce1 3.4661 2Ce2 4.2269 2Ce3 4.0909 Ru 4.2978 Ce3 4.3277 Ce1 4.3802 Ru 4.5993 materials-13-02377-t003_Table 3 ###### Results of the LSDA+U calculations performed for Ce${}_{9}$Ru${}_{4}$Ga${}_{5}$ with different values of the correlation energy *U*, where $n_{f}$ stands for the number of $4f$ electrons and *m* is the total magnetic moment per atom. *U* (eV) $\mathbf{\mathbf{U}_{\mathbf{Ce}1,\mathbf{Ce}2,\mathbf{Ce}3}}$ = 1.5 $\mathbf{\mathbf{U}_{\mathbf{Ce}1,\mathbf{Ce}2,\mathbf{Ce}3}}$ = 2.25 $\mathbf{\mathbf{U}_{\mathbf{Ce}1}}$ = 3 $\mathbf{\mathbf{U}_{\mathbf{Ce}2,\mathbf{Ce}3}}$ = 2.25 ---------- -------- ---------------------------------------------------------------------- -------- ----------------------------------------------------------------------- -------- --------------------------------------------------------------------------------------------------- Ce1 0.9833 0.9644 0.9880 0.9951 0.9873 1.0149 Ce2 0.8860 0.2589 0.8446 0.2808 0.8451 0.3099 Ce3 0.8732 0.4031 0.8309 0.4602 0.8423 0.5295 Ru −0.0396 −0.0446 −0.0477 Ga1 0.0014 0.0023 0.0027 Ga2 0.0013 0.0024 0.0035

Introduction {#s1} ============ Apple (*Malus x domestica* Borkh.) is one of the most cultivated fruit crops in temperate climates. The major constraint of apple cultivation is the apple scab, a fungal disease caused by *Venturia inaequalis*, which can lead to important crop losses if not properly controlled [@pone.0078457-Holb1]. In apple, at least 17 major resistance genes (*Rvi1* to *Rvi17*) against *V. inaequalis* have been found [@pone.0078457-Bus1]. However, only *Rvi6* (previously *Vf* from *Malus floribunda* 821) has been extensively used for resistance breeding to date [@pone.0078457-Gessler1]. Since the breakdown of the *Riv6* gene in the early nineties [@pone.0078457-Parisi1] new breeding programmes have started to investigate other resistance genes for future resistance breeding [@pone.0078457-Gygax1]--[@pone.0078457-Erdin1]. An increase in resistance with increasing apple leaf age (ontogenic resistance) has been observed in all apple genotypes and is known to act against all known *Venturia inaequalis* strains. To date, no report of the breakdown of this type of resistance has been found in the literature; thus, ontogenic resistance is considered durable [@pone.0078457-MacHardy1]. Goethe [@pone.0078457-Goethe1] and Aderhold [@pone.0078457-Aderhold1] are believed to have been the first researchers noticing age-related resistance in the *Malus*-*Venturia* pathosystem. The authors observed a decrease of leaf susceptibility with increasing tissue age. Nearly three decades later, Keitt and Jones [@pone.0078457-Keitt1] showed an increase in incubation period and a decrease of disease severity by increasing leaf age. Following these observations, many researches have been carried out on *Malus*-*Venturia* interaction during leaf infection. Gessler and Stumm [@pone.0078457-Gessler2], Li and Xu [@pone.0078457-Li1], and Gusberti et al. [@pone.0078457-Gusberti1] showed that the fungus grew faster in young leaves compared to old ones. The first unfurled and expanding leaf is considered susceptible to the apple scab disease, while the fifth leaf (starting from the top of the shoot) is considered fully resistant [@pone.0078457-Gessler2]--[@pone.0078457-MacHardy2]. Disease resistance mechanisms during tissue ontogenesis have been studied in different plant pathogen systems and some factors have been suggested to be correlated to the observed age-related resistance. Among them, the most important appears to be chemical compounds such as salicylic acid [@pone.0078457-Kus1], [@pone.0078457-Hugot1] and pathogenesis-related proteins [@pone.0078457-Hugot1], [@pone.0078457-Wyatt1], physiological barriers like the cuticle [@pone.0078457-Peries1]--[@pone.0078457-Ficke2], lenticels [@pone.0078457-Kennelly1], restricted phloem movement [@pone.0078457-GarciaRuiz1] or a limiting nutritional substrate for fungal infection [@pone.0078457-Juen1]. However, since a different mechanism for age-related resistance is described in each crop plant, much work remains to unveil the mechanism underlying this type of resistance in other plants. In apple, several aspects have been investigated in order to unveil the nature of ontogenic resistance. Physiological barriers like the cuticle and papillae [@pone.0078457-Stadler1], were not correlated to the age-related resistance. Research focusing on chemical barriers like melanoproteins, phenols, flavonols, polygalacturonases-inhibition proteins and the activity of different enzymes (e.g. phenylalanine ammonia lyase, polyphenoloxidase, β-glucosidase, chitinase, and fungal polygalacturonases) has been performed [@pone.0078457-MacHardy2], [@pone.0078457-Treutter1]. However, despite nearly half a century of research, no clear patterns for ontogenic resistance of apple were found. Thus, since physical and biochemical barriers have been exhaustively studied in this pathosystem without any clear pattern for the observed ontogenic resistance, other aspects to be considered are genes that are differentially expressed between young and old leaves. Furthermore, analyses to unveil the constitutive or pathogen-induced nature of the ontogenic resistance are needed. Today, the rapidly evolving sequencing techniques based on total RNA sequencing (RNA-seq), have decreased costs of analysis and increased the precision of results, allowing the researcher to maximise data outputs minimising their laboratory work and manipulation bias [@pone.0078457-Wang1], [@pone.0078457-Morozova1]. RNA-seq uses next generation sequencing (NGS) technology (Illumina\'s Genome Analyser, SOLID from Applied Biosystems, or the 454 Genome Sequencer) to sequence and quantify transcripts [@pone.0078457-Morozova1]. With the technical progress of this rapidly evolving technology, some studies have focused the research at the transcriptome level to find fungal effectors [@pone.0078457-Petre1],[@pone.0078457-deJonge1] and mechanisms involved in plant defences against microorganisms like chemical defences [@pone.0078457-WanderleyNogueira1] and structural defences [@pone.0078457-Xu1]. Thus, today, NGS appears to be the most promising methodology to study plant pathogen interactions in non-model species [@pone.0078457-Morozova1] like *Malus x domestica*. Moreover, the genome sequence of *Malus x domestica* 'Golden Delicious' has been recently published [@pone.0078457-Velasco1]. The aim of this work was to identify and characterise genes that are significantly differentially expressed during the shift from the susceptibility of young leaves to the resistance of old fully expanded leaves (ontogenic resistance) of the apple plant. Moreover, the constitutive or induced mechanism behind ontogenic resistance were studied by comparing inoculated and uninoculated leaves in the early phase of fungal colonisation at 72 and 96 hours post-inoculation. The data presented in this work will add more knowledge to the *Venturia-Malus* pathosystem and provide new insight for future researches on ontogenic resistance in apple. Materials and Methods {#s2} ===================== Plant material {#s2a} -------------- Plant material comprised young *Malus x domestica* (Borkh.) 'Golden Delicious' saplings grafted on M9T337 rootstock. Plants were kept in active growth with a 16h/8h (day/night) photoperiod and a minimum of 65.5 µmol m^-2^ s^-1^ light intensity provided with white fluorescent lamps (Philips Master TL-D 36W/830); they were fertilised once with Obstdünger 12∶8∶16∶2∶0.2 (N∶P∶K∶Mg∶B; OBA-Lanze, Hauert AG, Switzerland) one week prior to the start of the experiment. Plants were grown under greenhouse conditions at 20±2°C and 70% RH until 10 to 15 leaves stadium. Rape-seed oil (Maag AG, Switzerland) mixed with Alaxon 50 (Maag AG, Switzerland) and two weeks later Vertimec (Maag AG, Switzerland) were used as chemical treatments before starting the experiment to keep the plants free of pest-insects. Before inoculation, leaves were numbered and marked from the top of the shoot toward the base, with leaf 1 being the youngest unfurled and expanding leaf. Previous work has shown that leaf 1 is completely susceptible [@pone.0078457-Gessler2]--[@pone.0078457-MacHardy2] and that leaf 5 (starting from the top of the shoot) is already expressing ontogenic resistance [@pone.0078457-Gusberti1]. In this experiment, we selected leaf 7 (*i.e*. 4 to 5 days older than leaf 5 [@pone.0078457-Valsangiacomo1], [@pone.0078457-Lindhagen1]) in order to ensure clear differences between the leaf age classes. Inoculum, inoculation procedure and sampling {#s2b} -------------------------------------------- For inoculation, conidia of *Venturia inaequalis* (Cooke) single spore isolate no. 1639 [@pone.0078457-Bus2], [@pone.0078457-Bus3] were first multiplied on potato dextrose agar medium (Difco, USA) over 90-mm filter-papers (Whatmann International Ltd., USA); conidia were then suspended in sterile distilled water and stored in the refrigerator at -20°C until use. The second step of conidia multiplication was performed *in planta* on susceptible cultivar 'Golden Delicious' until enough conidia were produced. Sporulating leaves were then dried at room temperature in paper boxes and stored in plastic bags in the refrigerator at -20°C until inoculation. Inoculation, with a spore suspension of 5×10^5^ spores ml^-1^, was performed at 17±2°C and \>98% RH, as previously described [@pone.0078457-Gusberti1]. Plants were incubated under these conditions for 48 h following inoculation allowing the fungus to penetrate the cuticle and establish the primary sub-cuticular stroma. Half of the plants were challenged with the apple scab pathogen and half were mock inoculated. After the treatment (scab or mock inoculation), plants were grown at 17±2°C and 70% RH until sampling. The first sets of samples were collected 72 hours post-inoculation (hpi). This enabled the plants to acclimatise at the lower relative humidity (70%) for one day after the incubation period. The second sets of samples have been collected at 96 hpi, which is the moment when the pathogen is first recognised by the plant [@pone.0078457-MacHardy3]. Samples have been collected in biological triplicate from leaf 1 (L1) and leaf 7 (L7) by removing the leaf tip (\<100 mg) of each leaf from inoculated and uninoculated shoots, with 12 samples for each time point (72 and 96 hpi), making 24 in total. Moreover, to enable the validation of RNA-seq data using the real-time quantitative reverse-transcription PCR (qRT-PCR), independent samples of mock inoculated *M. x domestica* 'Golden Delicious' leaf 1 and leaf 7 were collected at 72 and 96 hpi. Total RNA isolation {#s2c} ------------------- Fresh leaf tissue (\<100 mg) was sampled from inoculated and uninoculated leaf samples at 72 and 96 hpi and collected in 2-ml Eppendorf tubes (Eppendorf, Germany), previously prepared with 5 to 10 2-mm sterile glass beads, immersed in liquid nitrogen immediately after sampling and stored at −80°C until processing. Tissues were ground twice with the FP 120 Fast-Prep machine (Bio 101 Savant Instruments Inc., Qbiogene, France) for 30 s at a speed of 5.5 m s^-1^ with an intermediate immersion in liquid nitrogen between the two grinding steps. RNA was extracted with the SV Total RNA Isolation System (Promega Corporation, USA) and column purified following the manufacturer\'s instructions. After the addition of RNA Lysis Buffer, samples were homogenised with the FP 120 Fast-Prep machine again for 30 s at a speed of 5.5 m s^-1^. RNA integrity and quality was tested with the Agilent 2100 Bioanalyzer RNA 6000 NANO assay (Agilent Technologies, Palo Alto, CA, USA). After RNA isolation and quality assessment, samples were stored at -80°C until cDNA library construction and transcriptomic assay. Libraries preparation for Illumina HiSeq 2000 {#s2d} --------------------------------------------- Complementary DNA (cDNA) libraries were constructed, starting from 1 µg of total RNA, at the Functional Genomic Center Zurich (FGCZ) following the TruSeq RNA Sample preparation protocol v.2 instructions (Low Throughput protocol, Illumina, Inc.). The quality of the isolated RNA was further determined with a Qubit® (1.0) Fluorometer (Life Technologies, CA, USA) and a Bioanalyzer 2100 (Agilent Technologies). Only those samples with a 260 nm/280 nm ratio between 1.8--2.1 and a 28S/18S ratio within 1.5--2 were further processed. The TruSeq RNA Sample Prep Kit v2 (Illumina, Inc., CA, USA) was used in the succeeding steps. Briefly, total RNA samples (1 µg) were polyA-enriched and then reverse-transcribed into double-stranded cDNA. TruSeq adapters were ligated to double-stranded cDNA. Fragments containing TruSeq adapters on both ends were selectively enriched with PCR. The quality and quantity of the enriched libraries were validated using Qubit® (1.0) Fluorometer and the Caliper GX LabChip® GX (Caliper Life Sciences, Inc., USA). The products resulted in a smear with an average fragment size of approximately 260 bp. The libraries were normalised to 10 nM in Tris-HCl 10 mM, pH 8.5 with 0.1% Tween 20. Sequencing and processing of RNA-Seq data {#s2e} ----------------------------------------- Bar-coded libraries were spread over 4 Illumina HiSeq 2000 lanes, avoiding biological replicates in the same lane to assure the same instrument variation for the entire experiment. The TruSeq PE Cluster Kit v3-cBot-HS (Illumina, Inc., California, USA) was used for cluster generation using 2 pM of pooled normalized libraries on the cBOT. Sequencing was performed on the Illumina HiSeq 2000 paired-end at 2×101 bp using the TruSeq SBS Kit v3-HS (Illumina, Inc.). RNA-seq reads were quality-checked with fastqc which computes various quality metrics for the raw reads. Reads were aligned to the genome and transcriptome with Tophat v 1.3.3. Before mapping, the low quality ends of the reads were clipped (3 bases from the read start and 10 bases from the read-end). Tophat was run with default options. The fragment length parameter was set to 100 bases with a standard deviation of 100 bases. Based on these alignments the distribution of the reads across genomic features was assessed. Isoform expression was quantified with the RSEM algorithm (<http://www.biomedcentral.com/1471-2105/12/323>) with the option for estimation of the read-start position distribution turned on. All raw data were deposited in the European Nucleotide Archive (ENA: <http://www.ebi.ac.uk/ena/data/view/ERP003589>) and experimental meta-data are available in the ArrayExpress database ([www.ebi.ac.uk/arrayexpress](http://www.ebi.ac.uk/arrayexpress)) under accession number E-MTAB-1726. The analysis of Tophat files was performed on the CLC Genomics Workbench v. 5.5.1 (CLC bio, Aarhus N, Denmark), following the manufacturer\'s instructions. Sequences were then analysed with the RNA-seq analysis program of the CLC platform and mapped against unannotated *M. x domestica* 63541 genes set reference v.1.0 (<http://genomics.research.iasma.it/>). The insert size for paired-end reads was set between 150 and 250 bp and normalisation of expression values was performed using RPKM values [@pone.0078457-Mortazavi1]. All other parameters were kept at default levels. The CLC Genomic Workbench was further used to perform a principal component analysis with all differentially expressed genes (DEGs) found in each cDNA library. Identification of DEGs was based on normalised gene expression calculated as RPKM, analysed using the Baggerley\'s test [@pone.0078457-Baggerley1] and filtered with the False Discovery Rate (FDR) *P*-value correction of 0.0001 (one false discovery in 10000 discoveries). The resulting DEGs were then loaded on Blast2Go v. 2.5.1 (<http://www.blast2go.com/b2glaunch>; [@pone.0078457-Conesa1]) for Blastx and gene ontology analysis, separated using the Gene Ontology (GO) vocabulary (<http://www.geneontology.org/>). Ontology annotations were then refined using InterPro Scan, ANNEX, GoSlim and KEGG (Kyoto Encyclopaedia of Genes and Genomes; <http://www.genome.jp/kegg>) functions of the Blast2Go platform. The 20 most abundant transcripts for each cDNA library were filtered using the RPKM normalisation procedure on the CLC Genomic Workbench. Sequences were then loaded on Blast2Go v.2.5.1 for gene ontology functional annotation using level 2 GO vocabulary for biological process terms, molecular function terms, and cellular component terms. Moreover, a fasta file with all DEGs was generated and sent through the Mercator webtool (<http://mapman.gabipd.org/web/guest/app/mercator>) for Bincode mapping, and through webtool MapMan v.3.5.1 (<http://mapman.gabipd.org/web/guest/mapman>, [@pone.0078457-Thimm1]) for pathway analysis. Default parameters were used and JGI Chlamy Augustus models, TIGR5 rice proteins, InterProScan, and a Blast cut-off of 50 were selected. Pathway analysis was performed using the KEGG function of the Blast2Go webtool. KEGG pathway maps were then enriched by inserting the most significant DEGs found with the MapMan webtool. The analysis focused on signalling and hormone pathways, genes encoding chemical defences like pathogenesis-related proteins, genes encoding physical barriers like cuticle, waxes and callose, genes acting on the biosynthesis or transport of substances connected to fungal nutrition and genes regulating the acidity of leaf tissues like proton transporters and cation/anion co-transporters. Real-Time quantitative PCR validation of RNA-seq data {#s2f} ----------------------------------------------------- Specific primers for the five candidate genes were designed ([File S1](#pone.0078457.s001){ref-type="supplementary-material"}) using the Primer-BLAST webtool (<http://www.ncbi.nlm.nih.gov/tools/primer-blast>) of the National Center for Biotechnology Information (NCBI: <http://www.ncbi.nlm.nih.gov>) and verified against *Malus x domestica* genome v. 1.0 using the BLAST function of the Genome Database for Rosaceae (GDR: [www.rosaceae.org](http://www.rosaceae.org)). Independent RNA samples (*M. x domestica* 'Golden Delicious' uninoculated L1 and L7 leaves collected at 72 and 96 hpi) were reverse-transcribed in triplicate using the RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific, Fermentas, Hilden, Germany), following the manufacturer\'s instructions. The concentration of cDNA samples was quantified using a NanoDrop ND-8000 spectrophotometer (Thermo Scientific, Wilmington, USA). Preliminary specificity tests were performed using the end-point PCR performed with each of the five primer pairs (MT3, LOX, LTP, PX and EDS, [File S1](#pone.0078457.s001){ref-type="supplementary-material"}) in a 20 µl reaction. The Mastermix comprised 1× PCR Buffer (Fermentas, Hilden, Germany), 0.1 mM dNTPs, 0.05 µM primer pairs, 0.07 U µl^-1^ DreamTaq DNA Polymerase (Fermentas) and 5 µl (20.1±3.9 ng µl^-1^) cDNA. PCR thermo-cycler conditions were: 3 min at 94°C, followed by 35 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 30 s. The PCR products were then loaded on a 1% (w/v) agarose gel in 0.5× Tris-Borate-EDTA (TBE) buffer at 125 Volts for 1.5 h. Real-time quantitative PCR analyses were then performed on the ABI 7500 Fast Real-Time Sequences Detection System (Applied Biosystems, Foster City, CA, USA). Amplification conditions were 15 min at 95°C, followed by 45 cycles of 30 s at 95°C and 1 min at 60°C. Reaction mix (10 µL reaction) comprised 1× Hot FirePol EvaGreen qRT-PCR Mix Plus (ROX) buffer (Solis BioDyne, Tartu, Estonia), 10 µM forward and reverse primer pairs and 3 µl (201.3±39.3 ng µl^-1^) cDNA. Melting curve analysis was performed to confirm the specificity of the amplification product. Threshold line was set manually at 0.2 in every analysis, performed using the Sequence Detection Software v. 2.0.6 (Applied Biosystems). Each 96-well plate was loaded with No Template Controls, No RT control and positive controls in triplicate. Ubiquitin conjugating enzyme (UBC: MDP0000223660) was chosen as a housekeeping internal standard gene during qRT-PCR, as published in a previous work [@pone.0078457-Pagliarani1], and the relative expression of the five candidate genes has been calculated using the 2^−ΔΔCT^ method as previously described [@pone.0078457-Livak1]. Before comparing the fold- change between qRT-PCR and RNA-seq data, all L1 values were normalised to L1~average~ = 1. After the normalisation procedure, the differential gene expression between qRT-PCR and RNA-seq data was assessed using the one-tailed T-test (*p*\<0.05) using JMP v. 10.0.2 (SAS Institute Inc., Cary, US) on Windows 7. Results and Discussion {#s3} ====================== Approximately 30--55 million paired-end reads for each cDNA library were obtained ([Table 1](#pone-0078457-t001){ref-type="table"}). The principal component analysis (PCA) of DEGs showed two main clusters between old leaves (L7) and young ones (L1) for both inoculated and uninoculated leaves ([Figure 1](#pone-0078457-g001){ref-type="fig"}). {#pone-0078457-g001} 10.1371/journal.pone.0078457.t001 ###### Reads and genes mapped in each generated cDNA library. {#pone-0078457-t001-1} Hours post inoculation Condition [a](#nt101){ref-type="table-fn"} Biological replicate Reads mapped in pairs [b](#nt102){ref-type="table-fn"}(Mio.) Reads mapped in broken pairs (Mio.) Reads not mapped (Mio.) Total reads (Mio.) Mean reads per library (Mio.) SD Genes mapped by unique reads (%) [c](#nt103){ref-type="table-fn"} ------------------------ -------------------------------------------- ---------------------- -------------------------------------------------------------- ------------------------------------- ------------------------- -------------------- ------------------------------- ------ ------------------------------------------------------------------- 72 L1 I 1 4.9 11.8 19.5 36.2 20917 (33) 2 4.5 11.6 29 45.1 39.4 4.95 21626 (34) 3 5.5 10.5 20.9 36.9 22501 (35) L1 N 1 3.7 10.1 21.9 35.7 21401 (34) 2 2.3 6.7 21.7 30.7 34.0 2.89 19534 (31) 3 6.5 11.2 18 35.7 24080 (38) L7 I 1 4.5 12.2 33.6 50.3 20359 (32) 2 2.7 7.9 26.8 37.4 42.1 7.10 17211 (27) 3 6.2 12 20.5 38.7 21652 (34) L7 N 1 4.6 13.1 26.3 44 20023 (32) 2 5 11.7 25.3 42 46.3 5.86 20816 (33) 3 7.9 17.5 27.6 53 22210 (35) 96 L1 I 1 10.7 9 17.2 36.9 25046 (39) 2 10.8 9.6 22 42.4 41.8 4.63 24708 (39) 3 9.9 10.5 25.7 46.1 24849 (39) L1 N 1 9 7.8 29.1 45.9 24714 (39) 2 20.6 15.7 19.2 55.5 52.0 5.30 27606 (43) 3 18.1 14.7 21.8 54.6 27172 (43) L7 I 1 13.8 12.4 17.4 43.6 22604 (36) 2 16.8 14.3 23.8 54.9 45.1 9.10 24473 (39) 3 7.1 7 22.8 36.9 22088 (35) L7 N 1 9.9 9.9 22.3 42.1 22196 (35) 2 12.3 11.3 22 45.6 44.7 2.32 22765 (36) 3 14.1 14.1 18.3 46.5 23635 (37) Conditions: L1 and L7 correspond to Leaf 1 and Leaf 7 (enumeration starting from the top of the shoot toward the base) for *V. inaequalis* inoculated (I) and uninoculated (N) leaves at 72 and 96 hours post inoculation; Reads mapped, not mapped and total reads obtained with the CLC Genomics Workbench v. 5.5.1; Genes mapped by at least one unique read and proportion of genes mapped by at least one unique read out 63541 apple genes. Values obtained with the CLC Genomics Workbench v. 5.5.1. Differential expression analysis of *M. x domestica* with an FDR *P*-value correction of 0.0001 resulted in 6 and 16 DEGs between uninoculated (L1N) and inoculated (L1I) young leaves at 72 and 96 hpi, respectively, while between uninoculated (L7N) and inoculated old leaves (L7I), 56 and 6 DEGs were found at 72 and 96 hpi, respectively. The analysis performed on leaves of different ages (L1 vs. L7) resulted in 3119 and 1784 DEGs at 72 hpi, for uninoculated and inoculated leaves, respectively. At 72 hpi, 1027 DEGs were present in both uninoculated (33%) and inoculated (57%) old leaves. At 96 hpi, DEGs between leaves of different ages for uninoculated and inoculated plants were 3750 and 2490, respectively. At this time point, 1877 DEGs were found in both uninoculated (50%) and inoculated (75%) old leaves. Young leaves analysed at 72 and 96 hpi showed 52 DEGs, while old leaves showed 99 DEGs. Nine common DEGs were found at both time points, corresponding to 17% and 9% of the DEGs found for young and old leaves, respectively. In total, we obtained 5823 DEGs among the ten conditions tested ([File S2](#pone.0078457.s002){ref-type="supplementary-material"}). Bin annotation and mapping of the 5823 DEGs resulted in 23.93% unannotated genes with a blast cut-off value of 50 ([Files S3](#pone.0078457.s003){ref-type="supplementary-material"}, [S4](#pone.0078457.s004){ref-type="supplementary-material"}). Results of 72 and 96 hpi were discordant for inoculated old leaves. In the last sampling point (96 hpi), relatively fewer DEGs were found compared to the same situation at 72 hpi. Some of the DEGs were found for inoculated and uninoculated old leaves at 72 hpi but not at 96 hpi with an FDP *P*-value correction of 0.0001. These differences were connected to the highly stringent FDP *P*-value correction used during the analyses, which may have hidden the effect of some DEGs at 96 hpi. Differentially expressed genes putatively involved in ontogenic resistance {#s3a} -------------------------------------------------------------------------- In this part of the work, we examined DEGs during leaf ontogenesis using an RNA-seq approach to find a possible explanation to the observed ontogenic resistance in apple against *Venturia inaequalis*. The analysis was performed with three biological replicates for young and old leaves, either challenged or not with the pathogen at two time points (72 and 96 hpi). The biological variability in preceding transcriptomic experiments has been found to be low [@pone.0078457-Lu1], [@pone.0078457-Savory1]. However, our experiment displays a biological variability of single gene\' RPKM values between 1% and 98% of the average ([File S5](#pone.0078457.s005){ref-type="supplementary-material"}), showing the importance of taking at least three biological replicates in this type of experiments. Moreover, the accumulation of metabolites is possible when the corresponding biosynthetic pathway genes are highly expressed or not modulated in young tissues compared to old ones (e.g. callose, lignin, wax, flavonoids, phenols, salicylic acid, and tocopherol). This is highly dependent on the rapport between the rapidity of production and degradation of the compounds; with this type of experiment, the production, accumulation or degradation steps could not be investigated. Thus, quantification of compounds in young and old leaves, inoculated and uninoculated, by means of proteomic or metabolomics approaches, may be more informative than RNA-seq experiments. Differential gene expression analysis was performed between young (L1) and old (L7) leaves for inoculated (I) and uninoculated (N) shoots at both time points (72 and 96 hpi) on the 5823 DEGs found with a FDR *P*-value correction of 0.0001 ([File S2](#pone.0078457.s002){ref-type="supplementary-material"}). Results of DEGs in the different conditions tested have been summarized in the [figures 2](#pone-0078457-g002){ref-type="fig"} and [3](#pone-0078457-g003){ref-type="fig"}. {#pone-0078457-g002} {#pone-0078457-g003} KEGG analysis resulted in 127 pathways ([File S6](#pone.0078457.s006){ref-type="supplementary-material"}). The most abundant sequences were found in the starch and sucrose metabolism (147 sequences; 6%), followed by purine metabolism (101 sequences; 4%), glycolysis and glucogenesis (75 sequences; 3%), carbon fixation in photosynthetic organisms (66 sequences; 3%) and pentose and glucuronate interconversions (65 sequences; 3%). ### Signalling {#s3a1} Hormones and signalling mechanisms were investigated in *Arabidopsis thaliana* [@pone.0078457-Kus1]; the authors showed that the accumulation of salicylic acid (SA) was the only factor which correlated with the observed age-related resistance. In our experiment, we found that genes involved in the biosynthesis of hormone precursors were, in general, up-regulated in old leaves if uninoculated and down-regulated upon inoculation (e.g. [files S7](#pone.0078457.s007){ref-type="supplementary-material"}, [S8](#pone.0078457.s008){ref-type="supplementary-material"}). For both the inoculated and uninoculated old leaves, a down-regulation of transmembrane amino acid transporters encoding genes was observed. This may be correlated to the cessation of cell enlargement and cell division hormone response (auxin, cytokinine, brassinosteroid) in the old leaves. Moreover, we could identify an enhanced disease susceptibility 1 (EDS1: MDP0000253215) protein, first observed in a mutant of *A. thaliana* mutant that was susceptible to *P. parasitica* [@pone.0078457-Parker1]. EDS-silencing increased disease resistance in *Arabidopsi*s [@pone.0078457-Parker1], [@pone.0078457-Aarts1]. The EDS gene has been found to be necessary for the functionality and signal transduction of other resistance genes in *Arabidopsis* plants [@pone.0078457-Parker1]. In our experiment, the EDS1 gene encoding protein was down-regulated in both inoculated and uninoculated old leaves only at 96 hpi ([Table 2](#pone-0078457-t002){ref-type="table"}), which was confirmed by qRT-PCR analysis. However, the high variability between the three biological replicates does not enable us to detect significant differences between young and old uninoculated leaves at 72 hpi ([Figure 4](#pone-0078457-g004){ref-type="fig"}). The EDS1 gene encoding protein has been classified under biotic stress signalling molecule during the MapMan analysis; however, the mechanism behind EDS in apple plants remains to be unveiled. {#pone-0078457-g004} 10.1371/journal.pone.0078457.t002 ###### Summary of the candidate genes for the ontogenic resistance in Apple found at 72 and 96(hpi). {#pone-0078457-t002-2} Fold-change at 72 hpi [b](#nt105){ref-type="table-fn"} Fold-change at 96 hpi --------------- ----------------------------------- ------ ---------- ------ -------------------------------------------------------- ----------------------- -------- ------- MDP0000149327 peroxidase 3 978 9.4E-167 0.78 28.5 23.36 15.87 NDE MDP0000253215 enhanced disease susceptibility 1 1947 0 0.64 NDE [c](#nt106){ref-type="table-fn"} NDE −6.82 −2.72 MDP0000312397 lipoxygenase 3363 0 0.86 9.07 23.67 23.81 NDE MDP0000466190 metallothionein-like protein 201 8.4E-33 0.79 21.53 16.20 19.22 13.55 MDP0000940078 lipid transfer protein 342 5.5E-34 0.66 466.32 206.56 743.78 NDE Mean protein similarity; Fold-change of the genes differentially expressed between Leaf 1 (L1) and Leaf 7 (L7) for inoculated (I) and uninoculated (N) leaves at 72 and 96 hours post inoculation. Gene expressions of L1 have been normalised to L1~average~ = 1 before calculating the fold-change for L7; NDE: Not Differentially Expressed genes with the FDR *P*-value correction of 0.0001 procedure. ### Structural defences {#s3a2} Genes involved in the synthesis of cell wall precursors, wax precursors (Eceriferums) and lignin precursors (Phenylalanine ammonia lyase: PAL and cinnamyl-alcohol-NADPH dehydrogenase: CAD) were down-regulated in old leaves at both time points. Thus, a reduced production of phenylpropanoid alcohols (coumaryl-, caffeyl-, coniferyl-, hydroxyconiferyl- and sinepsyl-alcohols) could be expected (e.g. [files S9](#pone.0078457.s009){ref-type="supplementary-material"}, [S10](#pone.0078457.s010){ref-type="supplementary-material"}). The studies of Stadler [@pone.0078457-Stadler1] and Valsangiacomo and Gessler [@pone.0078457-Valsangiacomo1] showed that physical barriers, such as cuticle and papillae, were not linked to ontogenic resistance in apple. Our analysis confirmed this result since callose synthase genes, wax biosynthesis genes and lignin biosynthesis genes were found to be down-regulated in old leaves of both inoculated and uninoculated leaves compared to young ones. ### Chemical defences {#s3a3} Chemical barriers, such pathogenesis related (PR-) proteins, were linked to age-related resistance in some pathosystems [@pone.0078457-Hugot1], [@pone.0078457-Wyatt1]. In apple leaves, PR-proteins (β-glucanase, chitinase, endochitinase, thaumatine-like, defensin, oxalate oxidase, and protease inhibitor) encoding genes were found to be down-regulated in old leaves (L1 vs. L7) while other genes involved in metabolite production (peroxidase, lipid transfer proteins, and lipoxygenase) were up-regulated in general. At 72 hpi ([Figure 2](#pone-0078457-g002){ref-type="fig"}, [Table 2](#pone-0078457-t002){ref-type="table"}), a PR-protein gene (LTP: lipid transfer protein; MDP0000940078) showed increased gene expression in old leaves of both uninoculated and inoculated old leaves. At 96 hpi ([Figure 3](#pone-0078457-g003){ref-type="fig"}, [Table 2](#pone-0078457-t002){ref-type="table"}), LTP was up-regulated only in uninoculated old leaves (L1N vs. L7N), and not in the inoculated (L1I vs. L7I) old ones with an FDR *P*-value correction of 0.0001. However, without considering the FDR *P*-value correction, we also found a statistically significant (*p* = 0.022) up-regulation of LTP also for the inoculated old leaves. The results of the qRT-PCR confirmed the differential LTP expression between young and old uninoculated apple leaves at both time points ([Figure 4](#pone-0078457-g004){ref-type="fig"}). LTPs have been suggested to be linked to antifungal activity through different possible paths upon pathogen attack [@pone.0078457-Blein1]. A recent study [@pone.0078457-IsaacKirubakaran1] on LTPs showed the potential inhibition of germination and fungal growth *in vitro*. Moreover, the authors produced a transgenic tobacco plant overexpressing the LTP gene, resulting in inhibition of pathogen growth in this plant [@pone.0078457-IsaacKirubakaran1]. In *Malus-Venturia* pathosystem, the effect of LTPs may be connected to fungal growth inhibition rather than to inhibition of fungal germination, as no difference in conidia germination on leaves of different age has been observed [@pone.0078457-Gessler2]. Peroxidases (PXs) were linked to avirulent-microbe defence [@pone.0078457-Choi1]. The biochemical functions of PXs were connected to the lignin and suberin biosynthesis [@pone.0078457-Quiroga1] and to the regulation of reactive oxygen species (ROS) [@pone.0078457-Kawano1]. Silencing plant PX resulted in an increase in plant susceptibility [@pone.0078457-Choi1] and its overexpression enhanced plant resistance [@pone.0078457-Choi1], [@pone.0078457-Choi2]. In *Capsicum annuum*, PX expression increased upon pathogen attack [@pone.0078457-Choi1]. However in apple, PX3 (MDP0000149327) increased gene expression in old leaves has been observed in both inoculated and uninoculated leaves at 72 hpi. At 96 hpi, we could observe a significant up-regulation of PX3 for the uninoculated old leaves ([Table 2](#pone-0078457-t002){ref-type="table"}) with the stringent FDR *P*-value correction procedure. However, without the FDR *P*-value correction procedure, also the marginally significant (*p* = 0.047) up-regulation for the inoculated old leaves could be observed. The analysis performed with the qRT-PCR confirmed the differential PX3 expression between young and old uninoculated apple leaves at both time points ([Figure 4](#pone-0078457-g004){ref-type="fig"}). The effect of PXs on *V. inaequalis* growth may be linked to strengthening of the cell wall and consequently reducing the nutrient availability necessary for the fungal growth [@pone.0078457-Spann1]. Lipoxygenases (LOX) have been reported in numerous plant species [@pone.0078457-Porta1] and are involved in the first step of jasmonate biosynthesis pathway [@pone.0078457-Wasternack1]. Transformation of plants by LOX-silencing resulted in enhanced susceptibility to some microbial pathogens [@pone.0078457-Renc1], [@pone.0078457-Hwang1]. In our experiment, LOX (MDP0000312397) encoding protein showed an increased expression in both uninoculated and inoculated old leaves at 72 hpi. At 96 hpi, we could observe a statistically significant up-regulation of LOX for the uninoculated old leaves but not for the old inoculated ones ([Table 2](#pone-0078457-t002){ref-type="table"}) and as observed with other genes encoding proteins, the up-regulation of LOX in the inoculated old leaves was still significant (*p* = 0.024), but was discharged due to the stringent FDR *P*-value correction procedure. However, qRT-PCR results were in discordance with RNA-seq data ([Figure 4](#pone-0078457-g004){ref-type="fig"}). The increased expression of the LOX encoding protein may inhibit fungal growth by the production of fungal inhibitor oxylipin substances (e.g. hexanal and colnelenic acid) or by its own antimicrobial activity, as previously described [@pone.0078457-Porta1], [@pone.0078457-Vaughn1]. Metallothioneins (MT) are proteins connected to heavy metal detoxification in stressed plants [@pone.0078457-Cobbett1]. In apple plants, MT protein encoding genes have been found [@pone.0078457-Degenhardt1], [@pone.0078457-Degenhardt2]. In our work, we observed an up-regulation of MT3-like protein encoding gene (MDP0000466190) in old leaves at both time points ([Table 2](#pone-0078457-t002){ref-type="table"}). The results of the qRT-PCR confirmed the differential MT3 expression between young and old uninoculated leaves at both time points ([Figure 4](#pone-0078457-g004){ref-type="fig"}). The role of MTs in response to biotic stress is not fully understood; however, some suggestions were made: the up-regulation of MT3 may inhibit fungal growth through metal ion sequestration, leading to an unsuitable habitat for fungal growth, or by decreasing the fungal enzymatic activity [@pone.0078457-Poschenrieder1]; thus, in both situations, an inhibition of fungal growth may be expected. However, since ontogenic resistance in old senescing leaves is no longer functional [@pone.0078457-Kollar1], further studies on MTs at the senescence stadium must be performed. Tocopherol, part of the vitamin E group, has been postulated to have antioxidant qualities to maintain the chemical and physical properties of the epicuticular waxes [@pone.0078457-Collakova1]. This substance was found to improve fruit quality by decreasing disease incidence [@pone.0078457-Noga1], [@pone.0078457-Schmitz1]. In the work of Bringe et al. [@pone.0078457-Bringe1], an increase of tocopherol between leaf one and leaf seven has been observed. In our work, we found a constant down-regulation of genes involved in the biosynthesis of tocopherol in old leaves at both time points. This result does not contrast with the findings of Bringe et al. [@pone.0078457-Bringe1] if tocopherol, as suggested previously [@pone.0078457-Rise1], accumulates in old leaves. However, at the onset of autumn, leaves lose their ontogenic resistance [@pone.0078457-Kollar1]; therefore, it is unlikely that tocopherol plays an important function in this resistance mechanism. Phenols and flavonoids have been extensively studied in the past five decades in apple tissues [@pone.0078457-Treutter1], [@pone.0078457-Barnes1]--[@pone.0078457-Treutter2] apparently without any conclusive answer to the observed age-related resistance. In the present work, flavonoids and phenols precursor genes were down-regulated or not differentially expressed in old leaves of the conditions tested. ### Old leaves as suitable substrate for fungal growth {#s3a4} After analysing structural and chemical defences in apple leaves, we focused on the suitability of old leaves for fungal growth in the early phase of tissue colonisation. Fungal growth in artificial media has been investigated in the past. Leben and Keitt [@pone.0078457-Leben1] showed that the best carbon sources for *V. inaequalis* were sugars and alcohols and the most suitable nitrogen sources were amino acids (arginine, glutamic acid, histidine and proline), urea and ammonia compounds (-sulphate and --phosphate). Thiamine was the only vitamin essential for fungal growth. In the present study, we observed a general up-regulation of sugar biosynthesis genes and sugar transporter genes in old leaves at 72 hpi ([Figure 2](#pone-0078457-g002){ref-type="fig"}), while at 96 hpi ([Figure 3](#pone-0078457-g003){ref-type="fig"}), sugars transporter genes were down-regulated in uninoculated old leaves and up-regulated in the inoculated ones. Thus, it does not seem probable that sugar amount in old leaves plays an important role in apple ontogenic resistance. Thiamine synthesis precursor genes did not show any differential regulation between young and old leaves at both time points, indicating that thiamine is not the limiting factor leading to ontogenic resistance. Nitrate, ammonium, sulphate and phosphate transporter genes were up-regulated in both uninoculated and inoculated old leaves in most of the tested conditions. Genes involved in amino acid biosynthesis and cell wall precursor synthesis (cellulose and proline-rich proteins) appeared in general to be down-regulated in old leaves compared to young ones at 72 hpi, while at 96 hpi they displayed up-regulation in uninoculated old leaves (L1N vs. L7N) and down-regulation in the inoculated (L1I vs. L7I) old leaves. The transmembrane amino acid transporter encoding genes showed a down-regulation in both inoculated and uninoculated old leaves. However, the action of peroxidases and the consequent cell wall lignification may limit the diffusion of these nutrient compounds between the cell and the sub-cuticular space, limiting therefore the fungal growth [@pone.0078457-Spann1]. Fothergill and Ashcroft [@pone.0078457-Fothergill1] showed that *V. inaequalis* growth was stimulated at pH values above 5.8. Later works [@pone.0078457-Raa1], [@pone.0078457-Raa2] suggested a different pH between young (pH = 6) and old leaves (pH = 5). With these works it may be suggested that the fungal growth, as a result of sub-optimal growth conditions in old leaves, may be inhibited. In our work, we observed an up-regulation of proton transporter precursor genes in both uninoculated and inoculated old leaves at 72 hpi, while at 96 hpi they were down-regulated in both uninoculated and inoculated old leaves. Magnesium ion transmembrane transporter genes did not show any differential expression at 72 hpi and were down-regulated in both uninoculated and inoculated old leaves at 96 hpi. Potassium, sodium, chlorine, and calcium ion transporter genes showed a down-regulation in old leaves compared to young ones at both time points. Other unspecified anion transporter genes were down-regulated. However, in the work of Raa [@pone.0078457-Raa1] and Raa and Overeem [@pone.0078457-Raa2], the difference in pH between leaves of different ages was determined with leaf homogenates, which make the assumption of a different pH between young and old leaves difficult to prove with RNA-seq experiments and to connect to *V. inaequalis* growth. In fact, this pathogen invaded only the sub-cuticular space of the leaf, thus the acidity of the sub-cuticular space would be a better factor to analyse in future researches. Functional annotation of the most abundant transcripts {#s3b} ------------------------------------------------------ Analysis with the level 2 GO vocabulary of young (L1) and old (L7), inoculated (I) and uninoculated (N) leaves at both time points (72 and 96 hpi) was performed for the 20 most abundant transcripts found in each tested condition ([File S11](#pone.0078457.s011){ref-type="supplementary-material"}), using the three GO classes, *i.e*. biological process, molecular functions, and cellular components ([Figure 5](#pone-0078457-g005){ref-type="fig"}). At 72 hpi, 50% of the 20 most abundant transcripts were present in both uninoculated and inoculated young leaves (L1), while 80% of the transcripts were present in both inoculated and uninoculated old leaves (L7). At this time point, only one gene encoding protein (5%) was present in all tested conditions ([File S11](#pone.0078457.s011){ref-type="supplementary-material"}). At 96 hpi, 95% of the 20 most abundant transcripts were common in both inoculated and uninoculated young (L1) leaves. The same proportion could also be observed in the old leaves (L7). At this time point, eleven (55%) gene encoding proteins could be found in all tested conditions ([File S11](#pone.0078457.s011){ref-type="supplementary-material"}). {#pone-0078457-g005} ### Biological process {#s3b1} At 72 hpi, inoculation of young leaves (L1N vs. L1I) resulted in a decrease of biological regulation terms (GO:0065007), cellular process (GO:0009987), metabolic process (GO:0008152), and response to stimulus (GO:0050896), while localisation terms (GO:0051179) increased. Inoculation of old leaves (L7N vs. L7I) resulted in an increased response to stimulus (GO:0050896), and signalling (GO:0023052), while metabolic process (GO:0008152), cellular process (GO:0009987), developmental process (GO:0032502), localisation components (GO:0051179), and multicellular organismal process (GO:0032501) decreased ([Figure 5A](#pone-0078457-g005){ref-type="fig"}). Inoculation of old leaves (L7N vs. L7I) led to an increased response to stimulus and signalling of biological components, while metabolic processes showed the most significant decrease. Signalling components indicate the transmission of information within a biological system and ends up with a cellular response. In our work, signalling may be connected to the pathogen perception and induction of resistance response. The response to stimulus component was connected to high expression of two PR-10 protein genes (both coding for Mal d 1.0105) found in old inoculated leaves at 72 hpi. The biological function of PR-10 proteins is not entirely known, but some work suggests that the PR-10 protein possesses a ribonuclease activity [@pone.0078457-Liu1], [@pone.0078457-Park1], which may prevent fungal growth [@pone.0078457-Galiana1] in host plants. The underrepresentation of metabolic process terms in inoculated old leaves was connected to reduced metabolism, which was probably also influenced by the decreased photosynthesis detected upon inoculation, as observed in other crop plants [@pone.0078457-Hao1]. At 96 hpi, we did not observe any major changes between inoculated and uninoculated young leaves: a slight decrease of cellular process (GO:0009987) and localisation terms (GO:0051179) was observed. Old inoculated leaves showed a decrease in response to stimulus terms (GO:0050896) and biological regulation terms (GO:0065007) ([Figure 5A](#pone-0078457-g005){ref-type="fig"}). ### Cellular components {#s3b2} Inoculation of young leaves (L1N vs. L1I) resulted in a decrease of cellular components (GO:0005623), extracellular matrix components (GO:0031012), organelle components (GO:0043226), membrane components (GO:0016020), and macromolecular complex components (GO:0032991) at 72 and 96 hpi ([Figure 5B](#pone-0078457-g005){ref-type="fig"}). The "macromolecular complex" components showed a decrease of terms from 9 to 1 between uninoculated and inoculated young leaves. The macromolecular complex term implies a stable assembly of two or more macromolecules in which both constituents function together. In our experiment, this function is highly underrepresented in inoculated young leaves. In the context of leaf inoculation, this was shown by decreased photosynthesis due to a decrease of photosystem I reaction centre subunit XI, chlorophyll a-b binding protein 4 precursor, photosystem II reaction centre w chloroplastic, and light-harvesting complex II protein lhcb3. The inhibition of photosynthesis upon pathogen attack has already been reported [@pone.0078457-Sabri1]. Inoculation of old leaves (L7N vs. L7I) did not show any major difference at any time point (72 and 96 hpi). Between young and old leaves, differences were found principally between the cell components, which decreased from young to old leaves at both time points ([Figure 5B](#pone-0078457-g005){ref-type="fig"}). ### Molecular functions {#s3b3} At 72 hpi, inoculation of young leaves (L1N vs. L1I) showed only minor changes in structural molecule activity terms (GO:0005198) and nutrient reservoir activity terms (GO:0045735), while at 96 hpi, differences were found only for transporter activity terms (GO:0005215). Inoculation of old leaves (L7N vs. L7I) showed an increase of structural molecule activity terms (GO:0005198), of nutrient reservoir activity terms (GO:0045735), of receptor activity terms (GO:0004872), and catalytic activity terms (GO:0003824) at 72 hpi, while no differences between inoculated and uninoculated leaves were found at 96 hpi ([Figure 5C](#pone-0078457-g005){ref-type="fig"}). Conclusions {#s4} =========== The RNA-seq technology is becoming an important research tool to investigate plant-pathogen interaction at the transcriptomic level. In this work, we described for the first time genes that are differentially expressed during the shift from susceptibility of young leaves to the resistance of old leaves of *Malus x domestica*, challenged or not with *Venturia inaequalis*, using a transciptomic approach. In this RNA-seq experiment the importance of using at least three biological replicates was shown: the standard deviation of the averaged RPKM values ranged from 1% to 98%. The high biological variability was also observed during the validation of the five candidate gene encoding proteins by means of the real-time quantitative reverse-transcription PCR. Nevertheless, we could propose five candidate gene encoding proteins linked to ontogenic resistance of apple: a metallothionein 3-like, a lipoxygenase, a lipid transfer protein, an enhanced disease susceptibility 1 protein and a peroxidase 3. The results presented in this study suggest that the ontogenic resistance in apple is the consequence of the fungal growth inhibition, either due to metal ion sequestration and inhibition of pathogens\' enzymatic activity (MT3), low mineral diffusion between the cell and the sub-cuticular space (PX3), secondary substances produced (LOX), or by the direct action of specific enzymes (LTP and LOX). Moreover, the transduction signal effect due to EDS1 needs further studies to determine the influence of this gene on the ontogenic resistance in apple. Further works, to test the relative expression of the candidate genes in several unrelated apple genotypes is needed. Additionally, the correlation between gene expression, protein and metabolite levels by means of proteomic and metabolomics approaches is desired to further determine the contribution of these genes to ontogenic resistance in apple. Supporting Information {#s5} ====================== ###### **List of primer pairs used for qRT-PCR validation of RNA-seq data.** (DOC) ###### Click here for additional data file. ###### **Complete list of the genes differentially expressed in all tested conditions.** Data were generated with the CLC Genomics Workbench v. 5.5.1 with a FDR *P*-value correction of 0.0001. (XLSX) ###### Click here for additional data file. ###### **Mercator\'s Bins mapping used for the functional annotation and pathway analysis of the 5823 differentially expressed genes obtained with the CLC Genomics Workbench v. 5.5.1 with a FDR** ***P*** **-value correction of 0.0001.** (TIF) ###### Click here for additional data file. ###### **Mercator\'s Bins mapping file of the 5823 differentially expressed genes of apple.** (DOCX) ###### Click here for additional data file. ###### **Biological variation of the transcription level observed for the five candidate genes involved in the ontogenic resistance of apple.** (XLS) ###### Click here for additional data file. ###### **Overview of the 127 pathways assigned with the KEGG pathway analysis of the 5823 differentially expressed genes obtained with the CLC Genomics Workbench v. 5.5.1 with a FDR** ***P*** **-value correction of 0.0001.** (XLSX) ###### Click here for additional data file. ###### **KEGG map for plant hormone signal transduction enriched with the genes differentially expressed found with MapMan v. 3.5.1 at 72 hpi between uninoculated leaf 1 and leaf 7.** (JPG) ###### Click here for additional data file. ###### **KEGG map for plant hormone signal transduction enriched with the genes differentially expressed found with MapMan v. 3.5.1 at 72 hpi between inoculated leaf 1 and leaf 7.** (JPG) ###### Click here for additional data file. ###### **KEGG map for cell wall and lignin precursors enriched with the genes differentially expressed found with MapMan v. 3.5.1 at 72 hpi between uninoculated leaf 1 and leaf 7.** (JPG) ###### Click here for additional data file. ###### **KEGG map for cell wall and lignin precursors enriched with the genes differentially expressed found with MapMan v. 3.5.1 at 72 hpi between inoculated leaf 1 and leaf 7.** (JPG) ###### Click here for additional data file. ###### **List of the 20 most abundant transcripts found in each cDNA library.** (XLSX) ###### Click here for additional data file. The quality control of RNA samples and qRT-PCR analyses were performed at the Genetic Diversity Center of ETH Zurich, which we gratefully acknowledge. We express our gratitude to the staff of the Functional Genomic Center Zurich: C. Aquino and Dr S. Alnuri for cDNA library preparation and Dr H. Rehrauer and Dr G. Russo for read quality control, alignment and preliminary bioinformatic analysis. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: MG CG GALB. Performed the experiments: MG. Analyzed the data: MG GALB. Contributed reagents/materials/analysis tools: MG. Wrote the paper: MG. Revision of the manuscript: CG GALB.

Introduction {#s1} ============ Chagas disease, a neglected tropical disease caused by the protozoan parasite *Trypanosoma cruzi*, is considered to be a public health problem [@pone.0097526-Nagajyothi1], [@pone.0097526-Machado1]. Over 10 million people are infected in Latin America and more than 100 million individuals live at risk of infection by blood transfusion, congenital, or oral transmission [@pone.0097526-Barbosa1]. Forty years after its introduction, benznidazole and nifurtimox continue to be the first choice of treatment for Chagas disease. However, chemotherapy based on nitroheterocyclic compounds has a limited efficacy for patients in the chronic phase of infection and these drugs are highly toxic [@pone.0097526-Diniz1],[@pone.0097526-Clayton1]. Little progress has been made toward the treatment of infected individuals and the development of more efficient drugs to treat Chagas disease patients remains urgent. Considering the resistance of some parasites to chemotherapy, the introduction of vaccines against *T. cruzi* could be another option [@pone.0097526-Barbosa1], [@pone.0097526-Buckner1]. *T. cruzi* is capable of resisting high doses of gamma radiation, enduring up to 1.5 kGy. As a direct biological effect, gamma radiation causes double-strand breaks (DSB) in the parasite DNA. However, 48 hours after irradiation, it is possible to see the chromosomal bands already restored. The parasite growth arrests for up to 120 hours, returning to the normal rate after this period [@pone.0097526-Grynberg1], [@pone.0097526-RegisDaSilva1]. This extraordinary recovery might be due to a very efficient DNA repair system. Homologous recombination is required to repair DNA DSBs and the involvement of the TcRAD51 protein in this process was evaluated by our group elsewhere. The overexpression of TcRAD51 ensures a more effective DSB DNA repair and a greater resistance to DNA damage in *T. cruzi* [@pone.0097526-PassosSilva1]. Oxidative stress is another effect of ionizing radiation due to the production of hydroxyl radicals (OH^•^), superoxide (O~2~ ^•^), and hydrogen peroxide (H~2~O~2~), directly from radiolysis of water. These products are commonly called reactive oxygen species (ROS) [@pone.0097526-Stadtman1]. Once the DNA molecule is intimately associated with water, the production of OH^•^ results in damages that include, apart from DSBs, oxidation of nitrogenous bases and sugar [@pone.0097526-Hutchinson1], [@pone.0097526-Riley1]. Approximately 75--80% of the biological damage caused by this type of radiation is mediated by OH^•^ formation. Such radicals are capable of reacting with most biologically relevant molecules. Each amino acid reacts differently with OH^•^ and the precise mechanisms of reaction are poorly understood [@pone.0097526-Anitori1]. Another organism that is extremely resistant to ionizing radiation is the bacterium *Deinococcus radiodurans*, which can withstand radiation doses of up to 15 kGy [@pone.0097526-Appukuttan1]. *D. radiodurans* presents a very robust DNA repair apparatus; nevertheless, the biological responses to genomic lesions depend on its proteome integrity. Considering that ionizing radiation also induces protein damage through oxidative stress, a protected functional proteome ensures an efficient cell recovery from this type of stress [@pone.0097526-Daly1]. Using the classical proteomic approach of two-dimensional differential gel electrophoresis (2D-DIGE) coupled with mass spectrometry (MS), Basu & Apte observed in a time-course analysis that some classes of proteins have a strong influence on stress responses. These proteins are mainly involved in processes such as DNA damage repair, protein synthesis and folding, and responses to oxidative stress [@pone.0097526-Basu1]. Proteome *versus* transcriptome analyses have been highly recommended for studies with tripanosomatids, as they have very peculiar molecular features concerning their gene expression control. As a kinetoplastid, *T. cruzi* transcription is polycistronic and gene regulation occurs mainly post-transcriptionally, with mature mRNAs being generated by trans-splicing and polyadenylation [@pone.0097526-Jger1], [@pone.0097526-Matthews1]. The processing and stabilization of mRNAs are extremely important in trypanosomatid gene regulation [@pone.0097526-DiNoia1], [@pone.0097526-DOrso1]. Furthermore, other dynamic control mechanisms, such as post-translational modifications, are fundamental in the regulation of gene expression and need to be better characterized in these organisms [@pone.0097526-Martnezcalvillo1]--[@pone.0097526-Zinoviev1]. A time-course microarray study previously carried out by our group analyzed the *T. cruzi* gene expression in response to gamma radiation [@pone.0097526-Grynberg1]. Among the 273 differentially expressed genes, 160 were upregulated and 113 were downregulated. The majority of the genes with assigned functions was downregulated. Translation, protein metabolic processes, and the generation of precursor metabolites and energy pathways were affected. Four mitochondrial genes and Retrotransposon Hot Spot genes were upregulated; likewise, the tyrosyl-DNA phosphodiesterase 1, a gene involved in DNA DSB repair, was also induced [@pone.0097526-Grynberg1]. Taking into account the *T. cruzi* gene expression peculiarities, analyses of proteome changes after irradiation in different time points may contribute to the understanding of the parasite response to such stress. In this work, we performed quantitative proteomic analyses using 2D-DIGE to ascertain the parasite response to ionizing irradiation. A total of 543 protein spots were found to be differentially expressed considering all analyzed time points and 53 different proteins were identified by tandem mass spectrometry (MS/MS). The great majority of the identified proteins was represented by several isoforms, suggesting that post-transcriptional and/or post-translational modifications are occurring as a consequence of gamma radiation exposure. Overexpression of tryparedoxin after irradiation was also observed, indicating that the parasite may be responding to the oxidative stress caused by irradiation. We also compared the time-course microarray and proteomic analyses. Although some of the protein expression patterns confirmed the microarray results, the correlation between mRNA and protein levels of the genes identified in both studies was extremely poor. In addition, treatment of the parasites with translation inhibitors showed that the synthesis of proteins putatively involved in the parasite response to stress is essential for its recovery from such a harmful stress. Materials and Methods {#s2} ===================== Cell Culture and Gamma Irradiation {#s2a} ---------------------------------- In this work, we used *T. cruzi* epimastigote forms of the CL Brener strain, which were isolated and characterized by Brener & Chiari [@pone.0097526-Brener1]. Clones have been maintained as frozen stocks at Universidade Federal de Minas Gerais. Parasites were grown at 28°C in liver infusion tryptose (LIT) medium pH 7.3, supplemented with 10% fetal bovine serum, streptomycin sulfate (0.2 g/L), and penicillin (200,000 units/L). Cultures in the exponential growth phase (2×10^7^ cells/mL) were exposed for 20 minutes to 500 Gy of gamma radiation (1,578 Gy/h) in a cobalt (60 Co) irradiator (Centro de Desenvolvimento da Tecnologia Nuclear -- CDTN, Belo Horizonte, Brazil). Cells were counted daily after irradiation to generate the growth curve. Cycloheximide and Puromycin Treatments {#s2b} -------------------------------------- Parasites exposed or not exposed to 500 Gy of gamma radiation were treated with cycloheximide (Calbiochem) 50 µg/mL for 15 minutes or with puromycin (Sigma) 25 µg/mL for 1 hour. Both drugs were added to the parasite cultures 4 hours after irradiation. Parasites were washed twice in phosphate buffered saline (137 mM NaCl, 4 mM Na~2~HPO~4~, 1.7 mM KH~2~PO~4~, and 2.7 mM KCl), the LIT medium was replaced, and the cells were counted. Protein Extract Preparation and DIGE Labeling {#s2c} --------------------------------------------- Protein extracts were obtained, simultaneously, in triplicate for each condition: non-irradiated control (NI), 4, 24, and 96 hours after irradiation. Parasites (2×10^9^ cells) were washed twice with LIT medium followed by centrifugation at 1,500 g for 5 minutes at 4°C. Each pellet was resuspended in 200 µL of lysis buffer (8 M urea, 2 M thiourea, 4% CHAPS, 10 mM Tris base) and a protease inhibitor mix (GE Healthcare, USA). Samples were mixed on vortex every 30 minutes during 2 hours of incubation at room temperature and subsequently centrifuged at 14,000 g for 30 minutes. The supernatants were aliquoted and stored at −70°C for further use. For all samples, protein concentration was determined using the 2D Quant kit (GE Healthcare, USA), according to manufacturer\'s instructions. Before labeling, samples had their pH adjusted to 8.5 with NaOH 0.05 M (as recommended by the manufacturer\'s protocol). To reduce biological variation, a pool of protein extracts obtained from triplicates was used. A total of 50 µg of protein from each pool (NI, 4, 24, and 96 hours after irradiation) was labeled with CyDye DIGE Fluor Minimal Labeling Kit (GE Healthcare, USA). The dye swap strategy was used to avoid label bias, where each sample was labeled with 400 pmol of either Cy3 or Cy5. A mixture of all protein extracts (12.5 µg of each pool sample) was labeled with Cy2 as the internal control. Reactions were carried out on ice for 30 minutes in the dark and then stopped by the addition of 10 mM lysine. Two-Dimensional Gel Electrophoresis {#s2d} ----------------------------------- ### First dimension {#s2d1} The isoelectric focusing (IEF) was performed using Immobiline Dry Strips (GE Healthcare, USA) 18 cm in size, with a pH ranging from 4--7. Strips were loaded with 50 µg of protein per CyDye (total of 150 µg) and sample buffer containing 8 M urea, 2 M thiourea, 4% CHAPS, 1% dithiothreitol (DTT), 0.002% bromophenol blue, and 1% IPG buffer (pH 4--7; GE Healthcare, USA). Passive rehydration followed overnight, at room temperature, in a strip holder (GE Healthcare, USA). The IEF protocol used in the Ettan IPGphor3 (GE Healthcare, USA) instrument was as follows: 50 µA per strip, 20°C, steps 1 to 5: 0.2 kV for 12 hours, 0.5 kV for 2 hours; 1 kV for 1.5 hour, 8 kV for 2 hours, 8 kV gradually raising to 40 kV, accumulating approximately 60 kV in total. Focused IPG strips were equilibrated for 15 minutes in an equilibration solution (50 mM Tris-HCl pH 8.8, 6 M urea, 30% glycerol, 2% SDS, 0.002% bromophenol blue and 125 mM DTT) and then alkylated for an additional 15 minutes in an equilibration solution containing 13.5 mM iodoacetamide instead of DTT. ### Second dimension {#s2d2} Equilibrated strips were briefly washed in 1x running buffer (25 mM Tris, 192 mM glycine, and 0.2% SDS) and placed on top of 12% acrylamide/bis-acrylamide gels, overlaid with a 0.5% agarose solution. Protein separation was carried out at 10°C, in an Ettan Dalt Six Electrophoresis System (GE Healthcare, USA), 45 mA per gel, until the dye front reached the bottom of the gel. Labeled proteins in each gel were visualized using the Typhoon FLA 9000 scanner (GE Healthcare, USA) at 100 µM image resolution with excitation/emission wavelengths for Cy3 (532/580 nm), Cy5 (633/670 nm), and Cy2 (488/520 nm). Gel images were uploaded and cropped using Image Loader Software (GE Healthcare, USA), then imported to DeCyder 2D software, version 7.0 (GE Healthcare, USA). DIGE Data Analysis {#s2e} ------------------ For spot detection, the Differential In-gel Analysis (DIA) module of DeCyder 2D software, version 7.0 (GE Healthcare, USA), was used. The DIA co-detection algorithm exploits the identical spot patterns from multiple samples in the same gel. After the removal of some artifacts from the gels, spot quantification was performed automatically by normalizing the spot volumes against the internal control. The following steps were performed in the Biological Variation Analysis module, which uses images processed in DIA and matches spots across gels. One-way ANOVA and Student\'s t-test were applied to evaluate differential protein expression levels between the groups of study. Spots classified as significantly differentially expressed were manually inspected. Abnormal spots were excluded from the analysis when necessary and gels were re-matched. Trypsin in-Gel Digestion, Mass Spectrometry, and Protein Identification {#s2f} ----------------------------------------------------------------------- Differentially expressed protein spots were excised and trypsin in-gel digestion was carried out overnight at 37°C with 20 ng/µL of trypsin (Promega, Sequencing Grade Modified Trypsin, USA), diluted in 25 mM ammonium bicarbonate. After trypsin digestion, peptides were extracted from the gel by washing twice with 30 µL of 50% acetonitrile and 5% formic acid solution and shaking for 15 minutes. Peptides were then concentrated (Eppendorf Concentrator 5301) to 10 µL and desalted using Zip-Tip (C18 resin, P10, Millipore Corporation, USA). Once the peptides were eluted (50% acetonitrile/0.1% trifluoroacetic acid) from columns, 0.5 µL of each sample was mixed with 0.25 µL of a saturated matrix solution \[10 mg/mL α-cyano-4-hydroxycinnamic acid (Aldrich, USA) in 50% acetonitrile/0.1% trifluoroacetic acid\]. Samples were spotted on the MTP AnchorChip 600/384 (Bruker Daltonics) and let to dry at room temperature. Raw data for the identification of proteins were obtained with the MALDI-TOF-TOF AutoFlex III (Bruker Daltonics, USA) instrument (Laboratório Multiusário de Biomoléculas, Departamento de Bioquímica e Imunologia, UFMG, Brazil) in the positive/reflector mode controlled by FlexControl software. Instrument calibration was achieved by using peptide calibration standard II (Bruker Daltonics) as a reference. Trypsin and keratin contamination peaks were excluded from the peak lists used for data base searching. Each spectrum was produced by accumulating data from 200 consecutive laser shots. MS/MS spectra were searched against the non-redundant protein sequence database from the National Center for Biotechnology Information (<http://www.ncbi.nlm.nih.gov>) using the MASCOT software (version 2.1) MS/MS ion search tool (<http://www.matrixscience.com>). The search parameters were as follows: no restrictions on protein molecular weight, two tryptic miss-cleavages allowed, and variable modifications of methionine (oxidation), cysteine (carbamidomethylation), and pyroglutamate formation at N-terminal glutamine of peptides. The mass tolerance for the peptides in the searches was 0.6 Da for MS spectra and 0.4 Da for MS/MS spectra. Peptides were considered to be identified when the scoring value exceeded the identity or extensive homology threshold value calculated by the MASCOT software (p\<0.05). Manual Curation and Statistical Analysis {#s2g} ---------------------------------------- Peptide sequences obtained from MASCOT were aligned to the *T. cruzi* annotated genome using the BLAST tool from TriTrypDB (<http://www.tritrypdb.org>). Protein annotation was reassigned particularly when partial sequences were chosen by MASCOT and full-length sequences were available at the TriTrypDB. Once a final annotated and curated set of upregulated and downregulated spots was available, it was possible to assess the protein species by their expected and observed weights (retrieved from the TriTrypDB and calculated from the position in the 2D-DIGE, respectively). Statistical analyses were performed using R in-house scripts with built-in statistical functions. A linear model was applied to test the correlation between molecular weight and fold-change. The Wilcoxon test was used to evaluate the presence of significant differences between 1) the observed molecular weights of upregulated and downregulated protein spots and 2) the observed and expected molecular weights from upregulated and downregulated protein spots. The final set of proteins was further manually annotated according to biological function and grouped into different functional classes based on literature data describing each protein and its molecular role. Results and Discussion {#s3} ====================== The Effects of Protein Synthesis Inhibition on the Growth of *T. cruzi* Epimastigote Cells Exposed to Gamma Radiation {#s3a} --------------------------------------------------------------------------------------------------------------------- Normal growth of epimatigote cells was affected by protein synthesis inhibition (using 50 µg/mL cycloheximide or 25 µg/mL puromycin) and by ionizing radiation treatment (500 Gy), as shown in [Figure 1](#pone-0097526-g001){ref-type="fig"}. However, irradiation promoted a more drastic growth arrest that persisted for approximately 96 hours; after this period, the parasites resumed normal growth, reaching the stationary phase 216 to 240 hours after irradiation ([Figure 1](#pone-0097526-g001){ref-type="fig"}). The treatment of NI cells with cycloheximide ([Figure 1A](#pone-0097526-g001){ref-type="fig"}) or puromycin ([Figure 1B](#pone-0097526-g001){ref-type="fig"}) retarded the cell growth by at least 24 hours when compared with non-treated cells, but did not lead to parasite death. Conversely, the combination of cycloheximide treatment and gamma radiation was lethal to 40% of the parasites. The remaining parasites resumed growth only 270 hours after irradiation, reaching the stationary phase 408 hours after irradiation ([Figure 1A](#pone-0097526-g001){ref-type="fig"}). For puromycin, a similar effect was observed, but treated cells resumed normal growth earlier when compared with cycloheximide-treated parasites ([Figure 1B](#pone-0097526-g001){ref-type="fig"}) and, in this case, no parasite death was detected. {#pone-0097526-g001} Analysis of the Proteome Profile of *T. cruzi* Epimastigote Cells Exposed to Gamma Radiation {#s3b} -------------------------------------------------------------------------------------------- Since we have verified that newly synthesized proteins have an impact on parasite recovery from irradiation stress, we decided to analyze time-course *T. cruzi* changes in the proteome induced by irradiation. Protein extracts were obtained from control NI cells and 4, 24, and 96 hours after irradiation. No significant losses in the total protein content and integrity were observed by 1D-gel electrophoresis ([Figure S1](#pone.0097526.s001){ref-type="supplementary-material"}). Using the 2D-DIGE approach, six gels were produced following the experimental design specified in [Table 1](#pone-0097526-t001){ref-type="table"}. This technique was chosen due to its greater sensitivity, reduced gel-to-gel variation, and its capacity for quantitative measurements of the relative abundance of each protein in a complex sample [@pone.0097526-Marouga1]. [Figure 2](#pone-0097526-g002){ref-type="fig"} illustrates 2D-DIGE gels at all time points. An average of 2,186±140 spots was found when compared with the master gel. From those, 543 presented altered expressions after irradiation, considering all time points (one-way ANOVA, p\<0.01) and 215 were identified by peptide mass fingerprint, corresponding to 53 different proteins ([Table 2](#pone-0097526-t002){ref-type="table"}). Almost half of these proteins (26) were represented by more than one spot in the 2D gel (ranging from 2--12 spots per protein), indicating the presence of several isoforms for the same protein. These results suggest that post-translational modifications or protein processing are occurring during the response to gamma radiation stress. We have manually annotated the function of all 53 identified proteins via a literature search. Proteins were then manually assigned to 15 different classes according to their biological function ([Figure S3](#pone.0097526.s003){ref-type="supplementary-material"}). {ref-type="table"}) showing -- in triplicate -- parasite proteins from each time point, labeled either with Cy3 (green) or Cy5 (red). Proteins were separated in the first dimension along a pH gradient (pH 4--7, 18 cm Immobiline DryStrip (GE Healthcare, USA), and in the second dimension in a 12% polyacrylamide gel. The molecular weight marker (MW) is indicated in kDa.](pone.0097526.g002){#pone-0097526-g002} 10.1371/journal.pone.0097526.t001 ###### Experimental design. {#pone-0097526-t001-1} Gel NI 4 h 24 h 96 h Pool ----- ----- ----- ------ ------ ------ 1 Cy3 Cy5 Cy2 2 Cy3 Cy5 Cy2 3 Cy3 Cy5 Cy2 4 Cy5 Cy3 Cy2 5 Cy3 Cy5 Cy2 6 Cy3 Cy5 Cy2 Each two-dimensional gel was loaded with 50 µg of total protein extract per sample, labeled either with Cy3 or Cy5. The internal control (a pool containing 50 µg of all time point proteins: NI, 4, 24, and 96 hours after irradiation) was labeled with Cy2. 10.1371/journal.pone.0097526.t002 ###### Protein data for the 53 proteins identified in this study. {#pone-0097526-t002-2} Observed/Expected Fold-change Mascot MS/MS ion search --------------------------------------------------------------------------------------------------------- --------------------------------------------------- ----- ------------------- ------------- ------- ------------------------- -------- ---------- ---- ----- ----- 14-3-3 protein; putative 511167.90 153 5.00/4.78 24.3/29.1 −1.35 1.08 −1.16 5.61E--5 2 12% 73 40S ribosomal protein S12, putative 508551.20 195 4.82/4.78 10.9/15.9 1.45 3.52 2.92 1.32E--5 2 24% 110 Actin, putative 510571.39 or 510127.79 or 510571.30 66 5.75/5.46 45.9/41.2 −1.64 −14.92 −12.65 8.20E--9 3 10% 136 Alpha tubulin, putative 411235.9 69 6.07/4.7 44.7/49.8 −1.01 −12.65 −10.53 2.00E--7 5 16% 248 99 5/4.7 55.6/49.8 −1.94 −31.15 −48.82 2.97E--8 11 42% 89 100 5.06/4.7 55.2/49.8 −2.11 −47.2 −45.08 1.34E--9 5 15% 275 136 5.29/4.7 29.9/49.8 1.54 −2.02 −1.68 8.44E--9 4 14% 247 138 5.47/4.7 30.4/49.8 1.09 −2.34 −2.21 8.55E--7 3 9% 155 139 5.45/4.7 28.9/49.8 1.36 −1.95 −1.63 6.52E--8 4 13% 237 Aminopeptidase, putative,metallo-peptidase, Clan MF, Family M17, putative 508799.240 35 6.22/6.44 58.9/55.9 −1.55 −−1.53 −1.73 5.40E--5 1 2% 52 ATPase beta subunit 509233.180 26 5.15/5.07 58.6/55.7 −2.42 −21.04 −18.44 7.26E--7 5 18% 253 Beta tubulin, putative 506563.40 107 5.58/4.43 45.0/49.7 −1.36 −4.9 −5.17 1.02E--7 3 7% 156 129 5.30/4.43 33.3/49.7 −1.01 −4.35 −3.99 1.53E--7 5 13% 242 130 5.15/4.43 33.8/49.7 −1.31 −10.73 −9.26 4.56E--8 6 17% 438 171 4.71/4.43 28.6/49.7 1.8 1.69 2.07 1.49E--8 7 22% 73 173 4.65/4.43 25.5/49.7 2.53 6.33 5.29 8.44E--9 4 11% 237 174 4.58/4.43 25.5/49.7 1.16 2.59 2.43 1.17E--6 4 11% 290 176 4.47/4.43 25.0/49.7 1.7 3.58 3.52 2.60E--7 3 9% 216 183 4.34/4.43 19.4/49.7 1.15 2.38 1.93 1.27E--5 2 6% 101 184 4.25/4.43 19.4/49.7 1.56 5.97 5.14 3.49E--8 2 6% 116 185 4.25/4.43 21.2/49.7 1.67 4.39 4.22 3.61E--6 2 6% 133 Calreticulin, putative 510685.10 81 6.37/4.49 42.2/46.2 −1.07 1 1.52 1.81E--5 2 6% 89 Chaperonin containing t-complex protein, putative 511725.250 12 5.02/4.80 69.7/59.2 −1.95 −12.81 −16.4 4.12E--7 3 9% 113 Chaperonin HSP60; mitochondrial precursor; GroEL protein; heat shock protein 60 (HSP60) 507641.290 or 507641.300 or 510187.551 17 5.44/5.14 66.2/59.2 −4.9 −19.19 −19.86 1.27E--8 9 24% 470 18 5.55/5.14 66.0/59.2 −4.07 −9.24 −10.18 1.75E--8 3 20% 111 20 5.65/5.14 65.8/59.2 −3.99 −8.83 −9.9 1.25E--7 7 18% 354 23 5.22/5.14 62.3/59.2 −2.03 −8.02 −9.51 6.42E--8 7 18% 191 24 5.13/5.14 62.2/59.2 −1.72 −11.52 −13.96 9.42E--8 11 31% 533 25 5.21/5.14 61.5/59.2 −1.78 −16.99 −15.1 1.61E--7 6 16% 205 28 5.29/5.14 62.0/59.2 −2.16 −9.12 −9.25 3.25E--8 5 14% 195 88 4.68/5.14 51.2/59.2 −1.06 −3.44 −2.79 9.76E--7 3 18% 112 89 4.83/5.14 51.8/59.2 −1.28 −7.85 −7.58 3.21E--7 3 18% 150 106 5.5/5.14 43.1/59.2 1.23 −5.57 −4.86 2.11E--7 6 15% 177 131 5.03/5.14 34.0/59.2 −1.76 −9.8 −7.3 3.45E--8 1 5% 76 162 5.66/5.14 20.3/59.2 1.97 1.81 1.97 4.31E--7 2 10% 62 Chaperonin; Tcomplex-protein 1; theta subunit; putative 506247.50 16 5.42/5.12 68.9/58.3 −2.42 −4.81 −5.44 6.21E--8 3 7% 138 Cystathionine beta-synthase, cysteine synthase, serine sulfhydrylase (CBS) 508177.120 or 506905.50 or 78 6.83/7.14 45.0/47.0 −1.1 −2.96 −2.21 1.61E--7 4 13% 206 508175.360 or 511691.10 80 6.37/7.14 45.3/47.0 −1.98 −10.23 −12.17 1.75E--8 2 7% 88 Cytochrome c oxidase subunit IV; putative 506529.360 or 510889.50 124 5.51/5.96 36.0/38.9 −1.01 −1.92 −1.58 5.98E--7 2 7% 105 Cytochrome c oxidase subunit V, putative 510565.30 or 508503.20 200 5.5/6.4 14.8/22.2 1.99 6.2 5.12 1.53E--7 2 14% 69 D-isomer specific2-hydroxyacid dehydrogenase-protein 510099.120 119 6.72/6.41 35.2/38.5 −1.46 −6.4 −5.31 6.69E--7 12 36% 103 197 5.26/6.41 11.2/38.5 1.13 −4.38 −3.75 9.76E--9 6 20% 321 Dihydrolipoamide acetyltransferase precursor 509717.20 and 510105.170 0 5.75/6.39 62.2/49.6 −1.32 −7.65 −5.03 1.75E--8 1 3% 61 33 5.91/6.68 62.1/49.6 −1.49 −7.45 −6.79 4.65E--9 4 14% 167 Dihydrolipoyl dehydrogenase; putative (GCVL-2) 507089.270 or 511025.110 73 6.69/7.4 52.5/54.9 1.01 −2.41 −2.34 2.36E--7 3 7% 98 Dipeptidyl-peptidase 508601.141 or 509205.120 29 5.40/5.60 62.4/74.4 −2.19 −5.73 −4.96 5.14E--8 2 3% 58 30 5.48/5.63 62.3/74.4 −1.9 −7.12 −8.88 1.75E--8 3 6% 139 Drug resistance protein 444777.10 123 5.54/4.05 37.2/50.3 −1.18 1.68 2.34 5.72E--6 1 5% 26 Elongation factor 2, putative 510963.90 36 6.36/5.86 55.8/94.2 −1.82 −6.16 −5.58 5.78E--8 4 6% 197 49 5.95/5.86 54.6/94.2 −1.99 −7.15 −6.91 2.60E--7 3 4% 130 50 5.99/5.86 54.6/94.2 −1.93 −11.21 −11.53 3.72E--9 4 6% 146 65 5.78/5.86 50.2/94.2 −1.4 −12.96 −9.24 3.48E--7 6 8% 325 125 5.34/5.86 36.0/94.2 −1.2 −7.53 −7.72 1.34E--9 5 9% 293 112 5.96/5.86 38.0/94.2 1.11 −2.38 −1.3 6.35E--4 4 6% 151 137 5.38/5.86 30.7/94.2 1.83 1.32 1.98 2.44E--7 3 2% 92 Enolase 504105.140 72 6.54/6.2 50.6/46.4 −2.63 −5.28 −6.37 6.94E--8 2 7% 73 Eukaryotic translation initiation factor 6 (elF-6); putative 506679.70 168 5.04/6.09 20.7/33.2 1.55 1.31 1.27 3.90E--5 2 9% 127 Glucose-regulated protein 78, putative 506585.40 2 5.19/4.82 76.9/71.3 −2.45 −23.64 −16.54 1.02E--7 4 12% 198 13 4.98/4.82 67.1/71.3 1.21 −2.77 −2.68 1.91E--6 2 4% 88 95 4.72/4.82 45.7/71.3 −1.08 −4.09 −3.33 7.81E--7 2 3% 74 96 4.58/4.82 45.4/71.3 1.52 1.68 1.52 4.25E--5 4 9% 297 Glutamamyl caboxypeptidase; putative 507689.40 or 507657.20 or 507657.10 70 6.18/6.51 47.6/43.4 −1.22 −2.15 −2.41 1.66E--6 2 6% 92 76 6.53/6.51 47.0/43.4 −1.28 −2.09 −2.19 5.13E--7 2 6% 110 77 6.59/6.51 45.5/43.4 −1.23 −2.99 −2.31 1.36E--7 3 9% 129 Glutamate dehydrogenase 508111.30 212 6.72/8.05 15.9/45.0 1.62 1.64 2 6.63E--6 2 7% 77 213 6.79/8.05 15.9/45/0 1.88 2.39 2.64 4.83E--8 2 7% 110 214 6.78/8.05 15.1/45.0 1.86 3.66 4.03 6.54E--9 4 13% 173 Glycerate kinase, putative 508741.170 159 6.49/8.21 20.7/56.1 1.37 −1.87 −2.19 1.43E--6 1 3% 37 Heat-shock protein 70kDa, putative 509543.50 and 511257.10 1 5.14/4.55 76.1/70.0 −2.15 −9.69 −10.51 3.28E--8 2 11% 70 90 4.90/4.55 51.8/70.0 −1.16 −4.64 −5.4 2.91E--7 2 11% 79 91 4.98/4.60 52.6/70.0 −1.12 −4.52 −3.7 2.44E--7 4 18% 205 Heat-shock protein 70kDa, putative 506135.9 155 5.63/6.56 23.4/30.2 1.7 2.79 2.2 5.79E--5 2 10% 88 Heat-shock protein 70kDa, putative 511211.160 7 5.55/5.85 72.7/70.9 −1.34 −4.53 −3.34 1.15E--7 4 15% 191 92 5.00/5.85 47.8/70.9 −1.11 −5.67 −5.23 8.47E--8 2 11% 67 101 5.13/5.85 44.1/70.9 1.3 −3.49 −3.28 4.17E--9 2 4% 67 105 5.50/5.85 44.9/70.9 −1.49 −4.58 −3.7 1.75E--7 2 4% 78 156 5.86/5.85 23.1/70.9 1.87 2.21 1.66 2.44E--7 4 19% 181 160 6.72/5.85 23.0/70.9 1.08 −2.66 −2.99 2.96E--8 13 48% 107 164 5.32/5.85 22.1/70.9 1.65 1.87 1.5 8.47E7 4 14% 214 175 4.56/5.85 24.1/70.9 1.23 1.34 1.38 8.68E--5 1 2% 66 177 4.65/5.85 20.8/70.9 2.01 5.22 4.11 5.33E−8 1 2% 72 Heat shock 70 kDa protein, mitochondrial precursor, putative 507029.30 8 5.65/5.71 72.4/71.0 −1.39 −4.87 −3.73 9.36E--9 2 4% 81 9 5.77/5.71 72.7/71.0 −1.42 −5.3 −4 9.88E--9 6 14% 293 10 5.90/5.71 73.1/71.0 −1.32 −3.83 −3.08 7.83E--8 3 7% 85 11 5.87/5.71 69.7/71.0 −1.05 −1.87 −1.94 1.55E--5 4 10% 152 19 5.60/5.71 67.4/71.0 1.18 −1.4 −1.53 1.17E--6 5 12% 285 21 5.73/5.71 67.7/71.0 1.26 −1.42 −1.58 1.70E--7 4 10% 109 Heat-shock protein 85kDa, putative 509643.130 or 507713.30 or 509105.140 93 4.89/4.79 47.6/80.7 −1.44 −5.4 −4.32 7.81E--7 2 3% 87 97 4.90/4.79 43.2/80.7 1.43 −1.5 −1.15 3.25E--7 3 5% 114 98 4.95/4.79 43.0/80.7 1.02 −3.01 −3.37 1.49E--8 2 3% 145 104 5.30/4.79 41.5/80.7 −1.63 −16.15 −18.55 2.26E--8 3 5% 143 126 5.40/4.79 33.5/80.7 1.2 −3.77 −3.74 4.34E--8 2 3% 89 134 4.62/4.79 40.6/80.7 1.39 2.58 2.18 7.35E--6 4 6% 258 148 4.95/4.79 28.8/80.7 1.76 2.78 2.47 3.49E--8 1 1% 100 149 4.86/4.79 28.9/80.7 1.67 3.55 3.23 6.89E--8 1 1% 62 Hypothetical protein, conserved 505989.110 182 4.46/4.50 18.4/22.2 −1.25 3.47 3.51 1.92E--5 2 11% 62 Hypothetical protein, conserved 506605.120 or 511239.110 202 5.72/4.99 15.1/28.6 2.75 7.08 7.83 1.34E--9 4 22% 174 203 5.67/4.99 14.2/28.6 1.87 3.96 3.33 3.66E--6 9 41% 92 204 5.86/4.99 14.1/28.6 1.48 2.12 2.78 1.88E--7 6 31% 326 Hypothetical protein 508817.20 or 503801.70 154 5.38/8.58 23.5/66.7 1.78 4.80 4.29 3.96E--8 1 1% 17 Nucleoside phosphorylase, putative 508989.9 and 509569.100 121 6.90/6.42 34.2/37.0 −1.18 −5.59 −4.38 2.96E--8 3 16% 190 118 6.38/6.42 35.6/37.0 −1.12 −3.17 −2.56 1.17E--8 1 4% 23 Oligopeptidase B, putative 503995.50 47 5.86/6.1 55.2/80.8 −1.71 −5.94 −7.55 9.76E--9 2 3% 68 63 5.86/6.1 52.0/80.9 −1.12 −2.26 −2.47 8.01E--7 2 3% 69 Paraflagellar rod protein 3 509617.20 60 6.09/5.96 56.1/68.6 1.21 −3.51 −4.33 6.89E--8 1 2% 24 Peptidase M20/M25/M40 510257.80 39 5.57/5.19 55.4/51.2 −1.74 −5.55 −7.07 8.47E--7 2 6% 66 40 5.50/5.19 54.4/51.2 −1.64 −5.35 −9.41 2.26E--8 1 3% 31 Peroxiredoxin; tryparedoxin peroxidase 509499.14 189 5.16/7.92 18.1/25.5 1.11 1.42 1.73 7.81E--7 3 15% 174 Phosphoglycerate kinase, putative or 3-phosphoglycerate kinase, glycosomal (PGKA) 511419.40 or 505999.90 or 511419.50 or 505999.100 74 6.76/7.4 51.9/54.90 −2.99 −5.56 −5.35 2.46E--7 1 3% 78 Prostaglandin F2 alpha synthase (TcPGFS) 508461.80 14 5.11/6.43 68.1/42.2 −1.19 −6.45 −7.01 1.59E--6 4 14% 169 111 5.91/6.43 40.0/42.2 −1.31 −11.76 −7.75 1.74E--8 5 17% 284 113 6.10/6.43 40.0/42.2 −1.11 −10.19 −6.05 1.75E--8 8 38% 116 114 6.09/6.43 38.7/42.2 1.12 −6.24 −5.06 3.12E--7 4 13% 188 144 6.15/6.43 26.7/42.2 1.49 −1.65 −1.39 5.49E--8 4 13% 217 161 5.52/6.43 22.1/42.2 1.74 1.88 1.99 6.34E--8 9 31% 92 Protein disulfide isomerase 506247.10 or 507611.370 180 4.42/4.6 20.9/53.5 1.46 3.7 3.6 2.44E--7 2 4% 44 Pyruvate dehydrogenase E1 beta subunit; putative 510091.80 132 5.03/5.02 30.9/37.8 1.08 −2.21 −1.88 1.27E--6 5 20% 191 133 4.62/5.02 40.7/37.8 1.02 −4.58 −3.83 4.13E--7 2 6% 108 Pyruvate kinase 2, putative 507993.390 or 511281.60 68 5.97/7.44 46.6/54.6 −1.3 −4.94 −4.8 1.73E--7 1 2% 29 Pyruvate phosphate dikinase 510101.140 194 5.00/8.27 13.5/100.8 1.83 10.51 7.42 7.88E--9 2 3% 94 Receptor for activated C kinase 1, putative 511211.120 or 511211.130 122 5.93/6.04 35.4/35.0 −1.22 −6.66 −6.06 2.31E--9 3 11% 122 S-adenosylhomocysteine hydrolase 511229.50 or 511589.200 193 5.25/6.64 12.9/48.4 1.41 −1.23 −1.1 3.74E--6 2 7% 80 Seryl-tRNA synthetase 511163.1 or 506777.80 140 5.38/5.41 28.9/25.7 1.4 −1.07 1.28 3.36E--6 3 19% 97 succinyl-CoA ligase \[GDP-forming\] beta-chain, putative 507767.10 150 4.92/5.58 26.2/34.5 1.63 2.03 2.25 1.08E--6 2 7% 87 Thiol−dependent reductase 1; putative; thiol transferase; putative; glutathione s-transferase; putative 509105.70 or 503419.30 62 6.00/5.83 51.9/50.7 −1.18 −2.07 −1.56 4.56E--6 2 6% 86 158 6.21/5.83 21.5/50.7 −1.32 −2.17 −2.64 4.48E--6 1 3% 32 Trans-sialidase 509927.10 186 4.18/6.67 21.0/54.7 1.67 4.7 4.71 6.10E--9 1 4% 25 Tryparedoxin peroxidase 487507.10 or 509445.10 or 504839.28 or 507259.10 210 6.24/6.75 11.5/22/4 2.17 3.18 3.42 8.05E--5 1 5% 33 211 6.66/6.75 17.0/22.4 −1.52 −3.96 −1.82 3.25E--8 3 22% 105 215 6.75/6.75 11.8/22.4 1.44 −1.3 1.25 1.83E--6 3 19% 109 Tyrosine aminotransferase 510187.20 and 510187.50 or 510187.40 or 510187.30 64 5.83/7.2 50.6/46.1 −1.22 −1.77 −1.51 5.02E--5 1 2% 58 79 6.25/6.14 44.9/46.1 −2.36 −31.03 −28.73 1.34E--9 4 16% 160 115 6.21/6.14 38.9/46.1 −1.05 −7.27 −7.38 9.64E--9 4 11% 110 116 6.18/6.14 37.0/46.1 1.09 −5.43 −4.73 8.44E--9 2 14% 128 Vacuolar ATP synthase subunit B 506025.50 or 511209.10 37 5.71/5.29 59.1/55.5 −1.69 −3.54 −4.01 2.44E--7 6 21% 207 Additionally, the Student\'s t-test (p\<0.01) was applied to verify which proteins were differentially expressed in each time point when compared with the NI sample. The overall and time-specific number of downregulated protein spots was higher than the number of upregulated ones ([Figure 3A](#pone-0097526-g003){ref-type="fig"}). These findings are different from those described previously in our microarray study. Twenty-four hours post-irradiation, the number of downregulated genes decreases drastically, reaching only 6 down-expressed transcripts 96 hours after irradiation, while the number of upregulated genes increases [@pone.0097526-Grynberg1]. A linear regression analysis between mRNA and protein levels from genes concomitantly identified in both studies was carried out for each time point. The correlation was extremely poor at all time points, starting with multiple R2 = 0.064 at 4 hours and reaching R2 = 0.27 and 0.24 at 24 and 96 hours post-irradiation, respectively (data not shown). Although a very low correlation was obtained, the result is in agreement with other studies performed in both prokaryotes and eukaryotes using classical methodologies such as microarray, Serial Analysis of Gene Expression (SAGE) and RNA-Seq for transcriptomic expression data, and 2-DE, Multi-dimensional protein identification technology, and MS for proteomics data [@pone.0097526-Gygi1]--[@pone.0097526-Butter1]. This analysis reinforces the idea that transcriptomic and proteomic approaches are complementary, not confirmatory [@pone.0097526-Nie1]. {#pone-0097526-g003} Moreover, changes in the *T. cruzi* proteome are more evident 24 hours after exposure to gamma radiation ([Figure 3](#pone-0097526-g003){ref-type="fig"}). This scenario suggests that epimastigote cells present an immediate but subtle response to gamma radiation characterized by 12 induced and 21 repressed protein spots 4 hours after irradiation ([Figure 3A](#pone-0097526-g003){ref-type="fig"}). Between 4 and 24 hours after irradiation, a more intense response to stress was observed and most of the induced and repressed protein spots were still significantly altered until 96 hours (428 spots; [Figure 3B](#pone-0097526-g003){ref-type="fig"}). This finding indicates a sustained alteration in the abundance of specific *T. cruzi* proteins 24 hours after gamma radiation exposure. When analyzing the *T. cruzi* proteome 24 hours after irradiation, we found that, from the 59 exclusive spots, approximately 66% were repressed and 34% were induced. However, the majority of the 23 exclusive spots found 96 hours after irradiation were induced (approximately 61%). Exposure to Gamma Rays Increases the Levels of Shorter and/or Processed Proteins in Epimastigote Cells {#s3c} ------------------------------------------------------------------------------------------------------ When analyzing the set of upregulated proteins (especially 24 and 96 hours after irradiation), we observed a tendency for the overexpression of shorter molecules to the detriment of longer ones. The upregulated protein spots (red-colored dots) are mainly at the lower part of the gel (lower molecular weight), while the downregulated protein spots (green-colored dots) are more sparsely distributed across the gel ([Figure 4A](#pone-0097526-g004){ref-type="fig"}). In addition, low molecular weight protein spots tended to have larger fold changes when compared with those with molecular weights close to the expected value ([Figure 4B](#pone-0097526-g004){ref-type="fig"}). The Wilcoxon test was applied and confirmed that the median values of the molecular weight of downregulated and upregulated protein spots were different for each time point (p\<1e-09; median values of 55.45/19.39, 45.64/19.38, and 46.51/19.42 for 4, 24, and 96 hours, respectively; [Figure 5A](#pone-0097526-g005){ref-type="fig"}). {#pone-0097526-g004} {#pone-0097526-g005} When we considered the expected molecular weight of the full-length isoform (predicted size obtained from the TriTryDB website) of both upregulated and downregulated proteins, we noticed a decrease in protein size in the former case and an increase in the latter case, thus showing the emergence of lower molecular weight protein isoforms after irradiation ([Figure 5B](#pone-0097526-g005){ref-type="fig"}). We then decided to confirm if the observed molecular weight of these proteins in the 2D gels was in agreement with their expected molecular weight. In the case of upregulated proteins, the observed molecular weight was significantly lower than expected ([Figure 5B](#pone-0097526-g005){ref-type="fig"}), indicating that these proteins might be processed, yielding shorter polypeptides. It is important to note that this result does not seem to be a consequence of protein degradation, since clear spots can be observed in the 2D gel, indicating the presence of a large amount of identical polypeptides in this region of the gel. This would not be the case if proteins were degraded, considering that in this situation peptides of variable size would be generated and no clear spot would be observed in the gel. These results may indicate the emergence of new protein isoforms, as the result of protein processing, alternative splicing of mRNAs, and/or alternative translational start/stop sites after irradiation. Alternative splicing of transcripts has the potential to expand the repertoire of proteins. Recent studies have estimated that all multi-exonic human genes are able to produce at least two alternatively spliced mRNA transcripts by alternative splicing, generating different proteins isoforms with altered structures and biological functions [@pone.0097526-Ezkurdia1]. In trypanosomatids, mature mRNAs are generated after two processing events: trans-splicing to add the spliced leader (SL) sequence to the 5′ end of transcripts and subsequent polyadenylation [@pone.0097526-Jger1]. A genome-wide analysis comparing the SL addition site along the developmental cycle of the parasite suggests that alternative trans-splicing plays an important role in differential gene expression [@pone.0097526-Michaeli1]. The occurrence of alternative trans-splicing could be an explanation for the presence of so many different isoforms in *T. cruzi* after radiation response. A similar event has already been described in *D. radiodurans*, since different isoforms of the single-strand binding (SSB) protein were produced after ionizing radiation stress induction. SSB proteins are vital for cell survival due to their involvement in processes such as DNA replication, recombination, and repair. The SSB protein spots in the gel followed a dynamic pattern of appearance, indicating a progressive processing of the C-terminal acidic tail, perhaps upon its interaction with ssDNA. The observed isoelectric point (pI) and molecular weight of deinococcal SSB isoforms were in agreement with the *in silico*-predicted pI and molecular weight of the SSB proteins shortened from the C-terminal end [@pone.0097526-Basu1]. Intriguingly, in most of the observed processed proteins, the identified peptide sequences were the same or nearly the same in all sequenced protein spots and, therefore, it was impossible to define the actual outcome of the protein processing. As a particular case of study, the protein annotated as prostaglandin F2 alpha synthase, which is similar to NADH-flavinoxidoreductase, is processed to a total of six different forms ([Figure S2](#pone.0097526.s002){ref-type="supplementary-material"}). While the expected molecular weight for the annotated sequence is of 42 kDa, only two isoforms are nearly this size (∼40 kDa) and were, in fact, the most downregulated isoforms. A third isoform has a predicted molecular mass of 68 kDa, greatly exceeding the expected protein size. As all of the MS/MS-identified peptides were mapped to the C-terminal portion of the isoforms, there is no information to characterize the N-terminal of this enlarged protein naturally present in the NI parasites and downregulated after exposure to gamma radiation. A smaller (29 kDa) protein is expressed in approximately equal levels before and after radiation exposure, while an even smaller (22 kDa) protein species is exclusively present in irradiated cells. This is an interesting example, representative of multiple cases, in which we have observed the emergence of shorter isoforms of a same protein after epimastigote irradiation. The list of processed proteins expressing shorter isoforms after irradiation includes alpha and beta-tubulin, D-isomer specific 2-hydroxyacid dehydrogenase-protein, elongation factor 2, glycerate kinase, pyruvate dehydrogenase E1 beta subunit, tyrosine aminotransferase, and several heat shock proteins (HSPs), such as HSP60, DnaK, HSP70s, and glucose-regulated protein 78. Apart from the previously discussed SSB proteins in *D. radiodurans*, very few references in the scientific literature mention the presence of shorter protein fragments after radiation exposure in any organism. Interestingly, Parodi-Talice and collaborators [@pone.0097526-ParodiTalice1] observed a similar pattern in *T. cruzi* for the proteins glutamate dehydrogenase (GluDH), HSP70, and alpha and beta-tubulins, where lower molecular weight isoforms were differentially expressed during metacyclogenesis when compared with isoforms with the predicted molecular weight. The transformation of epimastigotes into metacyclic trypomastigotes is a complex process of differentiation, requiring a controlled production of various proteins [@pone.0097526-ParodiTalice1]. Similarly, a quantitative time-course proteome analysis for the schizont-stage of *Plasmodium falciparum* (34 to 46 hours after invasion) demonstrated that actin-I, enolase, HSPs, and eukaryotic initiation factor 4A and 5A presented more than one isoform. The isoforms also showed different expression patterns at the different time points analyzed. *P. falciparum* is characterized by a complex life cycle, undergoing extensive morphological and metabolic changes, which reflects its capacity to survive in different host environments [@pone.0097526-Foth1]. According to the authors, post-translational modifications may be a very important strategy for the parasites to control gene expression during differentiation [@pone.0097526-ParodiTalice1], [@pone.0097526-Foth1]. Differentially Expressed Proteins after Gamma Radiation Exposure {#s3d} ---------------------------------------------------------------- Regarding the differentially expressed proteins, many of the listed proteins in [Table 2](#pone-0097526-t002){ref-type="table"} and [Figure S3](#pone.0097526.s003){ref-type="supplementary-material"} are related to the protein synthesis process that seems to be upregulated, except for some protein spots of the elongation factor 2 that show a reduction in their levels. This may be a response to compensate for the processing of proteins that occurs after irradiation. This response may also enhance the synthesis of specific proteins that will possibly play a role in the stress response. The results obtained from the analyses of translation inhibition and proteomic profile after irradiation place *de novo* protein synthesis as an important cellular response to gamma radiation. The same pattern is observed in *D. radiodurans*, where proteins related to translation/folding displayed either enhanced or *de novo* expression in the first hour of post-irradiation recovery. Proteins involved with DNA repair and oxidative stress alleviation were also induced in *D. radiodurans* under ionizing radiation stress [@pone.0097526-Basu1]. Proteins involved in protein folding processes, such as chaperones, are mostly downregulated post-irradiation ([Figure S3](#pone.0097526.s003){ref-type="supplementary-material"}). This represents an unexpected result, since these proteins are classically involved with stress response by stabilizing newly synthesized protein molecules. Nevertheless, this result is in agreement with transcriptomic data observed in microarray experiments [@pone.0097526-Grynberg1]. It is worth noting that, although HSPs are mostly downregulated, processed forms of these molecules are upregulated and may even be functional. Interestingly, the two chaperones localized in the endoplasmic reticulum (calreticulin and protein disulfide isomerase) are upregulated after gamma radiation exposure, which may indicate an important role of this compartment in the ionizing radiation stress response, suggesting the existence of an unfolded protein response-like in this condition [@pone.0097526-Conte1]. Another unexpected result is the downregulation of proteins involved in the ATP metabolism (namely the beta subunit of ATP synthase and the subunit IV of cytochrome c oxidase), although another member of this class is upregulated (cytochrome c oxidase subunit V). The outcome of this result is not clear and a more in-depth study of the cell energy metabolism would be important. Perhaps the most remarkable observation in the post-irradiation proteome investigated here is the putative decline in the activity of the glycolytic and amino acid metabolism pathways. Several important enzymes of glycolysis were downregulated after gamma radiation exposure. Accordingly, the only enzyme (pyruvate phosphate dikinase) from gluconeogenesis listed here was upregulated. Most enzymes involved in the amino acid metabolism were also downregulated, but shorter isoforms of the GluDH were upregulated after irradiation. They consist of three isoforms with experimental molecular weights (15 kDa) lower than the predicted values (45 kDa), suggesting once again the occurrence of post-transcriptional modifications/processing of important metabolic enzymes during the stress response. GluDH catalyzes the NAD- and/or NADP-dependent reversible deamination of L-glutamate to form alpha-ketoglutarate and is essential for the metabolism of amino nitrogen in organisms ranging from bacteria to mammals [@pone.0097526-BenachenhouLafha1]. *T. cruzi* has a metabolism that is largely based on the consumption of amino acids, mainly, proline, aspartate, and glutamate, which constitute the main carbon and energy sources of the epimastigote forms. In *T. cruzi*, GluDH has NADP-specific activity [@pone.0097526-Barderi1], indicating that it may serve as a pentose-phosphate shunt-independent source of NADPH in these parasites. Taken together, these results suggest that the parasite experiences an overall reduction on its energy metabolism as a consequence of its growth arrest after irradiation. We have identified four proteins classified as redox sensors in this study. While two of these are downregulated (both oxidoreductases), the other two are upregulated and these are both tryparedoxins, which efficiently reduce hydrogen peroxide [@pone.0097526-KrauthSiegel1]. Throughout its life cycle, *T. cruzi* is exposed to various stresses in different environments: the invertebrate (triatomine bugs) and the vertebrate hosts. One of the most deleterious consequences of oxidative stress may be the formation of DNA lesions. Guanine is the most susceptible base to oxidation, due to its low redox potential, and the 7,8-dihydro-8-oxoguanine (8-oxoG) is the most common lesion. When 8-oxoG is inserted during DNA replication, it can generate double-strand breaks, which makes this lesion severely deleterious. Recently Aguiar *et al*., 2013, demonstrated that parasites overexpressin MutT are more resistant to the oxidative stress caused by hydrogen peroxide (H~2~O~2~) treatment. The MutT enzyme product, 8-oxod-GMP, can generate an oxidative stress signal, enabling the cells to overcome this stress. MutT hydrolyses 8-oxo-dGTP in the nucleotide pool, returning it to the monophosphate form so that it cannot be incorporated into DNA by polymerases. Parasites overexpressing heterologous MutT also increase the levels of cytosolic and mitochondrial peroxidases (TcCpx and TcMPx) after H~2~O~2~ treatment. Taking this into account and also that parasites subject to gamma radiation experience oxidative stress and increase the levels of some antioxidant enzymes not immediately after irradiation, but later after irradiation, we could suggest that *T. cruzi* does not respond directly to ROS production as a consequence of irradiation, but to 8-oxo-dGMP that is generated subsequently. The nucleotide 8-oxo-dGMP, or another secondary metabolite generated from this process, could be acting as a second messenger to the cell and indicating the presence of oxidative stress. Recently, Krisko & Radman proposed a new paradigm when a cell is subject to ionizing radiation: the proteome rather than the genome is the primary target in radiation-induced cell death. This paradigm has been supported by several experimental evaluations showing that *D. radiodurans* has a way of protecting its proteins from oxidative damage [@pone.0097526-Krisko1]. Indeed, a strong correlation between intracellular Mn/Fe concentration ratios and bacterial resistance to radiation has been shown, in which the most resistant bacteria tolerates 300 times more Mn^2+^ and three times less Fe^2+^ than the most radiation-sensitive bacteria [@pone.0097526-Daly1]. Manganese ions prevent the formation of iron-dependent ROS through the Fenton reaction, acting as chemical antioxidant protectors. Furthermore, measurements of protein carbonyl groups in *D. radiodurans* revealed that Mn^2+^ accumulation prevented protein oxidation; these results were also observed in other radioresistant bacteria [@pone.0097526-Daly2]--[@pone.0097526-Robinson1]. Furthermore, the level of oxidative protein damage caused during irradiation controls the survival of many organisms (*Bdelloid rotifers*, a class of freshwater invertebrates, *Caenorhabditis elegans,* the bacteria *D. radiodurans,* and *Halobacterium salinarum*), which are extremely resistant to ionizing radiation [@pone.0097526-Robinson1]--[@pone.0097526-Zhang1]. The principal factor responsible for this extraordinary radioresistance is their great antioxidant protection of their cellular constituents, including those required for DSB repair, allowing them to recover from stress and continue reproduction [@pone.0097526-Krisko3]. An important finding of this study is the significant upregulation of three hypothetical proteins after gamma radiation. This may indicate a role for species-specific proteins in the response to stress after ionizing radiation exposure, since these most likely represent proteins with no homologues in other species. Similarly, in an initial *D. radiodurans* proteome study, hypothetical proteins were identified and further proved to be crucial for the response to radiation in this bacterial species [@pone.0097526-Zhang1]. Sghaier and collaborators have recently published a study on the amino acid composition of proteins from radiation-resistant bacteria [@pone.0097526-Sghaier1]. The authors report that such proteins bear more small amino acids and fewer aromatic rings. We have also assessed the amino acid composition of *T. cruzi* proteins in a slightly different perspective. Amino acid counts were performed for upregulated *T. cruzi* proteins after gamma radiation exposure (we have considered as upregulated the proteins that were more abundant than in NI cells at least in one time point) and for the orthologues in *T. brucei* of *T. cruzi* upregulated proteins. In both cases, amino acid counts were normalized by the count performed in the set of all annotated proteins of the respective *Trypanosoma sp*. The hypothesis was that proteins with important roles after irradiation in *T. cruzi* would have an amino acid composition different than what is observed in the set of all *T. cruzi* proteins and in the respective orthologues in *T. brucei* (which is not radio-resistant). When we compared *T. cruzi* proteins that were upregulated after radiation exposure with the entire set of annotated proteins from this parasite, we observed that the former have in general fewer polar, hydrophobic, and small amino acids (although some amino acids in these classes are more frequent). In addition, upregulated proteins have fewer aromatic amino acids (except for tyrosine, which is more frequent) and less sulfur-containing cysteine residues. Conclusions {#s4} =========== Using 2D-DIGE and MS, we have identified 543 protein spots differentially expressed after gamma radiation exposure. The presence of multiple isoforms was observed for more than half of the identified proteins, most of which are shorter than the annotated protein size in the *T. cruzi* genome. Additionally, there was a strong correlation for lower molecular weight peptide spots to be overexpressed. This result could be explained by *de novo* protein synthesis of different isoforms, protein processing, and/or modification events subsequent to radiation exposure. This observation indicates that post-translational control of gene expression have an important role in the parasite response to gamma radiation stress. The inhibition of protein synthesis in face of gamma radiation was shown to have a significant effect decreasing parasite growth and survival rates, highlighting the importance of active translation for parasite recovery after exposure to ionizing radiation. We have annotated all 53 proteins identified by MS according to their biological roles. Several proteins were represented by multiple spots, and most of them had molecular weights lower than predicted. As a consequence of this observation, we cannot precisely state which biological processes are upregulated *versus* downregulated, since the different protein isoforms may not function in the same way as the full-length protein. Nevertheless, some tendencies could be observed in this study, including changes in the following biological processes: upregulation of the protein synthesis process, downregulation of protein folding (except for the upregulation of two endoplasmic reticulum chaperones), downregulation of the ATP generation pathway, glycolysis, and amino acid metabolism, and the upregulation of two tryparedoxins (which reduce hydrogen peroxide in response to oxidative stress). Finally, taking into account the translation inhibition results obtained herein, together with the observed proteomic profile after irradiation, we can conclude that *de novo* protein synthesis is an essential cellular response to gamma radiation. Supporting Information {#s5} ====================== ###### Electrophoretic analysis of total protein extracts of irradiated and NI epimastigote cells. Total protein extracts were obtained for each time point NI, 4, 24, and 96 hours after irradiation. Samples were subjected to 12% SDS-PAGE and stained with coomassie blue. (TIF) ###### Click here for additional data file. ###### Differentially expressed isoforms of prostaglandin F2 alpha synthase. The upregulated protein spot (161) shows a lower molecular weight when compared with the downregulated proteins spots (14, 111, 113, 114, and 144). (TIF) ###### Click here for additional data file. ###### Time point expression of protein spots. (PDF) ###### Click here for additional data file. The authors would like to thank Neuza Antunes Rodrigues for technical assistance and Dr. Adriano Pimenta for making available the use of the mass spectrometry facility at Departamento de Bioquímica e Imunologia, UFMG. The authors also acknowledge the Laboratory of Gamma Irradiation, CDTN and Dr. Márcio Tadeu Pereira for kindly providing the access to the Co irradiator. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: HGSV PG SFP AMM CRM HMA GRF. Performed the experiments: HGSV PG MB SFP HOH. Analyzed the data: HGSV PG MB SFP AMM CRM HMA GRF. Contributed reagents/materials/analysis tools: HGSV AMM CRM HMA GRF. Wrote the paper: HGSV PG MB SFP GRF.

Introduction {#S0001} ============ Liver cancer is considered to be the fourth most lethal cancer globally, and hepatocellular carcinoma (HCC) accounts for 75--85% of liver cancer cases.[@CIT0001] In addition to the high mortality rate, the prognosis and treatment of HCC are suboptimal, most of the patients reach malignancy within a year of initial diagnosis.[@CIT0002] The survival statistics of the American cancer society from 2008 to 2014 showed that the the overall 5-year survival rate was 18% for liver cancer patients, but the 5-year survival rate for patients with distant metastasis was only 2%. In great contrast, among early-stage HCC patients who were diagnosed and treated before extrahepatic metastasis, the 5-year survival rate would be increased to 31%. To improve HCC early diagnosis rate, HCC biomarkers with higher sensitivity and specificity are required. Postoperative monitoring, which aims to evaluate disease progression and predict cancer recurrence, also heavily relies on the exploration of HCC biomarkers. Recently, targeted therapy, immune checkpoint blockade therapy, and combinational therapy showed promising efficacy in clinical trials. Biomarkers also play an important role in the design of personalized treatment plans. In the new era of genomic oncology, genetic biomarkers are becoming the core of cancer biomarkers. To bring a panoramic view of HCC genetic markers to academic and clinical experts, we conducted a systemic review of these genetic biomarkers for HCC early diagnosis, prognosis, treatment and postoperative monitoring. Etiology And Pathogenesis {#S0002} ========================= The primary risk factors of HCC are chronic hepatitis B and hepatitis C virus infection, alcohol consumption, non-alcoholic fatty liver disease, exposure to dietary aflatoxin, genetic hemochromatosis, and metabolic disorders.[@CIT0003],[@CIT0004] The resulting chronic liver inflammation may develop to severe liver fibrosis and cirrhosis, which were predispositions of HCC. It was reported that up to 90% of HCC cases occurred on the background of liver cirrhosis or fibrosis.[@CIT0005] Increased production of ROS was predicted to cause the accumulation of oxidative stress and DNA instability, which were accompanied by hepatocytes proliferation, telomeres shortening, and chromosomal alterations. These processes were associated with tumor development in fibrosis according to early studies.[@CIT0006],[@CIT0007] Interestingly, each HCC risk factor is involved in differed signaling pathways during carcinogenesis as [Figure 1](#F0001){ref-type="fig"} shows, and the resulting HCC patients often exhibit distinct genomic profiles.Figure 1Signaling pathways affected by the etiological factors of HCC. HBV/HCV infection, alcohol consumption, aflatoxin exposure, NAFLD and metabolic disorders may trigger HCC by manipulating diverse signaling pathways.**Abbreviations:** ADGRB1, adhesion G protein-coupled receptor B1 gene; AKT, protein kinase B; CPT2, carnitine o-palmitoyltransferase 2 gene; ER, endoplasmic reticulum; FAS, fas receptor; KCTD17, potassium channel tetramerization domain containing 17; NF-κB, nuclear factor-κB; NANOG, homeobox protein; PHLPP2, PH domain and leucine-rich repeat protein phosphatase; ROS, reactive oxygen species; STAT3, signal transducer and activator of transcription 3; TLR4, Toll-like receptor 4; TNFα, tumor necrosis factor; TWIST1, twist-related protein 1; UPR, unfolded protein response. Hepatitis B Virus Infection {#S0002-S2001} --------------------------- In HBV endemic regions such as Asia-Pacific and sub-Saharan Africa, HBV infection accounts for 75--90% of HCC incidences.[@CIT0008] Once entered the host cell, the HBV DNA transcribes to 4 viral mRNA for 7 HBV proteins,[@CIT0009],[@CIT0010] one of which is the 17kDa polypeptide HBV X (HBx) that regulates cell proliferation and apoptosis by modulating Wnt/β-catenin expression.[@CIT0011] As [Figure 1](#F0001){ref-type="fig"} shows, overexpression of HBx could also activate NF-κB to block tumor necrosis factor-α(TNFα)- and FAS-mediated apoptosis. In addition, both HBV and HCV can cause mitochondrial stress and increase reactive oxygen species (ROS) levels, which triggers endoplasmic reticulum (ER) stress and the unfolded protein response (UPR), leading to autophagy promoted cell survival and virus persistence.[@CIT0012] Given that different risk factors induce HCC through varied mechanisms, the genomic profiles of patients affected by those risk factors may differ. An early study screened biomarkers for HBV induced HCC and identified 7 up-regulated genes in those patients including RPS5, KRT8, CFLAR, ATP5F1, IGFBP2, MAP3K5, and MMP9. The genes regulate diverse cellular processes ranging from protein synthesis to cytoskeleton organization.[@CIT0013] More recent research showed that CCND1,[@CIT0014] BCL2, Mcl-1,[@CIT0015] NFKB1[@CIT0016] and SOCE[@CIT0017] were also up-regulated in HBV induced HCC. Additionally, one whole-exome sequencing study reported that TP53, CTNNB1, RB1, AXIN1, SELPLG, and FGF19 appear to be the candidates of driver mutation genes for HBV induced HCC.[@CIT0018] The unique genomic profile of HBV induced HCC can be applied to the early diagnosis of HCC, which attributes to the implement of precise curative treatment that may improve the survival of patients. Hepatitis C Virus Infection {#S0002-S2002} --------------------------- Being the second most common cause of HCC worldwide, HCV infection is responsible for at least 10% of HCC incidences.[@CIT0019] HCV infection is a major cause of HCC in Western countries, Africa and Japan.[@CIT0008],[@CIT0020] The core proteins of HCV such as E2, NS5A, and NS5B were shown to interfere with the E2F1 pathway and RAF/MAPK/ERK kinase pathways ([Figure 1](#F0001){ref-type="fig"}), by which they may modulate cell proliferation and tumor development.[@CIT0021] The other HCV produced protein, NV5a, was reported to inhibit the p53 pathway, which consequently modulates cell cycle.[@CIT0022] Apart from carcinogenesis, data suggest that HCV induced HCC can be distinguished from HBV induced HCC based on the genomic profile of the patients. An early study compared the gene expression pattern of HCV and HBV induced HCC, they concluded that different genes were up-regulated in the two scenarios.[@CIT0013] Unlike the HBV infected population, VIM, ACTB, GAPD, and CD58 were up-regulated in HCV induced HCC cases. A later study identified 40 up-regulated genes in HCV triggered HCC cases compared to the controls, including RYBP, ATP1B3, TMC, ZNF567, GPR108, CD19.[@CIT0023] These studies identified potential biomarkers for HCV induced HCC, which are crucial to the design of treatment strategy for precision medicine. Alcohol Consumption {#S0002-S2003} ------------------- Alcohol-related cirrhosis is the third most common trigger of HCC worldwide, which appears to correlate with alcohol consumption behavior.[@CIT0024] Alcohol intake may increase the production of iron-induced reactive oxygen species (ROS), which would interfere with DNA repair mechanisms ([Figure 1](#F0001){ref-type="fig"}). Moreover, acetaldehyde is formed during ethanol metabolization, the accumulation of acetaldehyde has negative effects on DNA and proteins.[@CIT0025] Patients carrying alcohol-related HCC also exhibit unique genetic profiles. It was reported that mutations in TRET promoter, CTNNB1, ARID1A are more common in alcohol-related HCC incidences.[@CIT0026] A recent study identified 5 up-regulated genes (CSMD1, MAGEA3, MAGEA6, CSAG1, and CSAG3) and 4 down-regulated genes (CD5L,UROC1, IGF2, and SLC22A10) that were associated with alcohol-related HCC.[@CIT0027] Exposure To Aflatoxin {#S0002-S2004} --------------------- Exposure to aflatoxin B1 (AFB1) is identified as a risk factor for HCC, which is propagated via food contaminations. AFB1 and HBV were proposed to have synergistic interaction.[@CIT0028] Although the mechanism remains unclear, HBV infection seems to sensitize hepatocytes to the carcinogenic effects of AFB1.[@CIT0029] AFB1 may trigger carcinogenesis via the transformation to aflatoxin-8, 9-exo epoxide, which interacts with the p53 tumor suppressor gene ([Figure 1](#F0001){ref-type="fig"}), and facilitates the mutation at R249S.[@CIT0029],[@CIT0030] Compared to other risk factors, the biomarker for AFB associated HCC was not extensively studied. In addition to TP53, One recent study claimed that frequent mutation in the adhesion G protein coupled receptor B1 (ADGRB1), AXINI and TERT were observed in AFB associated HCC cases.[@CIT0031] Non-Alcoholic Fatty Liver Disease And Non-Alcoholic Steatohepatitis {#S0002-S2005} ------------------------------------------------------------------- Non-alcoholic fatty liver disease (NAFLD) is one of the most common risk factors of chronic liver disease in the US, it is frequently associated with cirrhosis, which might lead to HCC.[@CIT0032] NAFLD represents a variety of chronic liver diseases ranging from hepatic steatosis to the progressed and inflammatory form non-alcoholic steatohepatitis (NASH). In particular, NASH-related HCC incidence increased by 63% from 2002 to 2012 in the US.[@CIT0033] A study reported that NAFLD and alcohol consumption contributed to steatohepatitis.[@CIT0034] Together with HCV-NS5A stimulated TLR4-NANOG and the leptin-phosphorylate STAT3 signaling pathways, NAFLD results in HCC by up regulating TWIST1 in tumor-initiating stem-like cells. NAFLD may alter the expression of KCTD17, PHLPP2, and CPT2, which promotes hepatic steatosis, NASH, and hepatocarcinogenesis.[@CIT0035],[@CIT0036] Similarly, The missense mutation of the PNPLA3 gene was reported to promote NAFLD/NASH-related HCC.[@CIT0037] Moreover, one study identified 41 candidate genetic markers for NAFLD associated HCC including Sav1, Son, Slc25a17, Fbxo11, Myo10, and Pten, the mutation seems to promote apoptosis and fibrogenesis.[@CIT0038] Staging Systems {#S0003} =============== The severity of cancer is classified by cancer staging systems. A number of systems have been proposed to predict the prognosis of HCC patients, they considered different variables and were tested in varied populations. Therefore, none of the staging schemes has been universally applied. The commonly adopted staging and scoring system for HCC prognosis are the tumor, node, metastasis (TNM) staging, the Okuda stage, the Barcelona Clinic Liver Cancer (BCLC) systems, the Cancer of the Liver Italian Program (CLIP) score, the Japan Integrated Staging (JIS) scores, the Chinese University Prognostic Index (CUPI) scores, the French scores, and the albumin-bilirubin (ALBI) grading system ([Table 1](#T0001){ref-type="table"}).Table 1Staging And Scoring System Of HCCNameCountryStagesReferenceTNMFranceStage I, II, III[@CIT0118]OkudaJapanScore A, B, C[@CIT0119]BCLCSpainScore 0-7[@CIT0039]CLIPItalyStage 0, A-D[@CIT0120]JISJapanStage I-IV[@CIT0121]CUPIHong KongScores Low-High risk[@CIT0122]ALBIJapanGrade 1,2,3[@CIT0123][^1] Different staging and scoring system were selected based on the clinical and scientific requirement. For instance, BCLC scoring was applied in clinical trials evaluating sorafenib, and sorafenib is recommended to be the treatment option for BCLC grade C.[@CIT0039],[@CIT0040] It is generally accepted that diagnosis at the early stage could significantly improve the survival of patients by allowing the implement of curative treatment. Furthermore, the design of treatment strategy is based on the understanding of the cause and stage of each individual. In that sense, diagnosis methods for precision medication should be developed for the identification of HCC at very early stage. Early Diagnosis Of HCC {#S0004} ====================== Diagnosis at an early stage is the key to HCC patient survival. Currently, the most widely used biomarker for HCC diagnosis is serum alfa-fetoprotein (AFP), but its sensitivity and specificity are both around 50%.[@CIT0041] It is worth noting that liquid biopsy has been extensively developed and put into clinical practice over the past decades. It mainly detects circulating tumor DNA (ctDNA), circulating tumor cell (CTC), exosomes, and circulating tumor RNA (ctRNA) in body liquid including plasma ([Figure 2](#F0002){ref-type="fig"}), urine, and cerebrospinal fluid. Among them, ctDNA is the most widely applied genetic biomarker. It is derived from tumor tissue, and carries somatic mutations, CNVs, DNA methylations, viral sequences, and physical characteristics associated with carcinogenesis.Figure 2Genetic biomarkers for HCC early diagnosis. Characteristics of the circulating genetic materials can be applied in HCC early diagnosis including somatic mutations, DNA methylation, exosome, micro RNA (miRNA), long non-coding RNA (lncRNA), and physical characteristics of circulating tumor DNA (ctDNA). Somatic Mutations {#S0004-S2001} ----------------- The genomic landscape of HCC has been revealed by multiple genome sequencing studies. The three most frequently mutated genes are TERT (40--60%), TP53 (31%), and CTNNB1 (27%). One systematic analysis summarized frequently mutated genes reported, including the tumor suppressor genes AXIN1 (8%) and RB1 (4%), the chromatin remodeling genes ARID1A (7%), ARID2 (5%) and BAP1 (5%), the cellular anti-oxidant defense genes NFE2L2 (3%) and its interactor KEAP1 (5%), Albumin (13%) and 10% APOB mutations.[@CIT0042] These mutational features were applied in the design of sequencing panels for HCC early screening. One research group developed a liquid biopsy assay named hepatocellular carcinoma screen (HCCscreen), which could detect HCC in asymptomatic HBsAg-seropositive individuals. The assay simultaneously assesses the ctDNA gene status of TERT, TP53, CTNNB1, and AXIN1, and the level of serum AFP and DCP, as well as the HBV integration profile.In the training cohort, the assay effectively distinguished HCC patients from non-HCC individuals, and the sensitivity and specificity of the assay were 85% and 93% correspondingly. In the validation cohort, the assay showed 100% sensitivity and 94% specificity. Notably, while the training cohort recruited individuals who had liver nodules and/or elevated serum AFP levels, the validation cohort only enrolled individuals with normal serum AFP levels and liver ultrasonography results.[@CIT0043] Another study reported that the TP53 mutation at codon 249 could be used as a biomarker to identify HCC caused by aflatoxin exposure or HBV infection. Its sensitivity and specificity reached 40% and 88% respectively.[@CIT0044] These results were encouraging in the sense that it provided a method for HCC screening at a very early, or even asymptomatic stage. DNA Methylations {#S0004-S2002} ---------------- The alteration of DNA methylation status is an early event in carcinogenesis, so it is also considered as a potential biomarker for HCC early detection. One study developed an HCC-specific methylation marker panel and built a diagnostic prediction model with ten markers. The sensitivity and specificity of the model were 85.7% and 94.3% in the training cohort (715 HCC patients and 560 non-HCC individuals), and 83.3% and 90.5% in the validation cohort (383 HCC patients and 275 non-HCC individuals). Furthermore, the model could effectively distinguish HCC induced by varied risk factors.[@CIT0045] Another study performed by Abderrahim Oussalah et al evaluated the accuracy of a PCR-based cfDNA assay for SEPT9 promoter methylation. The area under the receiver operating characteristic curve was 0.944, showing it could be served as a potential biomarker for HCC diagnosis.[@CIT0046] Physical Characteristics {#S0004-S2003} ------------------------ The physical characteristics of ctDNA are different from non-tumor-derived cfDNA in the aspects of size profiles and preferred end coordinates. Dennis Lo et al performed a detailed analysis of the size profiles of plasma DNA in patients with HBV induced HCC and healthy controls. The results showed that aberrantly short or long DNA molecules existed in the plasma of HCC patients. The short DNAs preferentially carried the tumor-associated copy number aberrations.[@CIT0047] Moreover, they revealed specific end coordinates of tumor-associated plasma DNA.[@CIT0048] By calculating the ratio of sequence with tumor-associated DNA ends versus sequence without the characteristic, the possibility of carrying HCC can be estimated, and the area under the receiver operating curve was 0.88. CTCs, Exosomes And ncRNAs {#S0004-S2004} ------------------------- CTCs are derived from advanced tumors, they may promote tumor metastasis. In a meta-analysis study, the pooled sensitivity and specificity of CTC for the detection of tumor were 67% and 98% respectively.[@CIT0049] Similarly, tumor-derived exosomes were shown to deliver oncogenes, pathogens and microRNAs. It was reported that a number of exosomal proteins, RNAs and miRNAs may serve as biomarkers for HCC including miR-122, miR-21, and miR192.[@CIT0050] Consistently, Huang et al performed a systematic review and meta-analysis of published studies to evaluate the utility of microRNAs (miRNAs) in the early diagnosis of HCC. They analyzed 50 studies that included 3423 cases of HCC, 2403 chronic hepatic disease (CH) patients, and 1887 healthy controls. It showed that miRNA could be used as a marker to discriminate HCC patients from healthy individuals, and the sensitivity and specificity were 75.8% and 75.0% correspondingly.[@CIT0051] LncRNAs (long non-coding RNAs) are non-coding mRNA-like trasncripts that are longer than 200 nucleotides. Many novel lncRNAs were proposed to be predictive biomarkers for cancer diagnosis or prognosis, including HULC, HOTAIR, MALAT1, and H19.[@CIT0052] In the case of HCC, lnc-Myd88 was suggested to be a new diagnostic and therapeutic target for HCC because that it could increase Myd88 expression, which in turn activates NF-κB and PI3K/AKT signaling pathways.[@CIT0053] Prognosis And Post-Operational Monitoring Of HCC {#S0005} ================================================ The primary treatment strategy for HCC is to remove the tumor by surgery, but HCC may relapse after the operation. It was reported that the 5-year recurrence rate of HCC after surgery was 74.2% in Japan and Korea.[@CIT0054] Therefore, post-operational monitoring and precise prognosis are necessary for cancer treatment. Genetic sequencing might be an alternative tool for cancer post-operational monitoring. For example, ALB1 mutation was observed in recurrent liver cancer, the time that the mutation can be detected in ctDNA is associated with the time when the HCC relapses.[@CIT0055] A unique mutation of HCK p.V174M was also observed in recurrent HCC and metastatic HCC, which was diminished after surgery but would increase rapidly if HCC relapses.[@CIT0056] Similarly, the IL-28B (rs8099917) TT genotype was interrelated with HCC recurrence.[@CIT0057] Taken those facts together, genetic sequencing may serve as an important tool for the post-operational surveillance of cancer. Genetic testing also serves as a useful tool for HCC prognosis evaluation. A number of studies have identified biomarkers associated with the prognosis of HCC, such as miR-203,[@CIT0058] ATXN7,[@CIT0059] and Alpha-1-fucosidase.[@CIT0060] A study demonstrated that the concentration of HMGB1 in blood was linked to prognosis of patients with HCC after receiving sorafenib treatment for 4 weeks.[@CIT0061] In addition, it was reported that the profile of circulating tumor DNA in blood samples of HCC patients could reveal the heterogeneity of tumors and monitor the process of disease in real time.[@CIT0056] Targeted Therapy Of HCC {#S0005-S2001} ----------------------- U.S Food and Drug Administration has approved 7 drugs for HCC treatment ([Table 2](#T0002){ref-type="table"}), which were applied for targeted therapy or immunotherapy. Sorafenib, lenvatinib, regorafenib, ramucirumab and cabozantinib are inhibitors of receptor tyrosine kinase (TK). These drugs can suppress the activities of major TKs including vascular endothelial growth factor receptors (VEGFR1-3), platelet-derived growth factor receptor β (PDGFR-β) and fibroblast growth factor receptors 1--4 (FGFR1-4), thereby inhibiting the growth of tumor cell by antiangiognic effects.[@CIT0062]--[@CIT0065] HCC is majorly supplied by hepatic arteries whereas liver parenchymas are supported primarily by portal vein. In fact, radio-graphically visible tumors rely entirely on the oxygen and nutrients supply via hepatic vasculature.[@CIT0066] Therefore, one of the strategy applied in HCC treatment is to target the angiogenesis pathways.Table 2FDA Approved HCC Drugs And Their Molecular TargetsDrugApplicationApproval TimeTargetReferenceSorafenibFirst-line2007VEGFR1-3 and PDGFR[@CIT0062]LenvatinibFirst-line2018VEGFR1-3,FGFR1-4, PDGFR-α, KIT and RET[@CIT0071]RegorafenibSecond-line2017VEGFR1-3, c-TKITIE-2, PDGFR-β, FGFR-1, RET, c-RAF, BRAF and p38MAP kinase[@CIT0075]CabozantinibSecond-line2019VEGFR, MET, and AXL[@CIT0078]RamucirumabSecond-line2019VEGFR-2[@CIT0097]NivolumabSecond-line2017Human immunoglobulin G4 anti-PD-1 monoclonal antibody[@CIT0124]PembrolizumabSecond-line2018PD-1 check point[@CIT0091][^2] As the first drug that has been shown to increase the overall survival of patients with liver cancer, sorafenib inhibits the serine/threonine kinases that are the crucial components of the Raf/MEK/ERK pathway ([Figure 3](#F0003){ref-type="fig"}). Moreover, the drug could suppress the activity of vascular endothelial growth factor receptors (VEGFR1-3) and platelet-derived growth factor receptor β (PDGFR-β), thereby inhibiting the growth of tumor cell.[@CIT0067] Consistently, a phase III clinical trial named SHARP demonstrated that sorafenib significantly improved the median survival time (mOS) of the patients compared with the placebo (10.7 months vs 7.9 months).[@CIT0068] A trial in the Asia-Pacific region also showed that sorafenib extended the mOS of patients to 6.5 months, whereas the mOS of the placebo group was 4.2 months.[@CIT0069]Figure 3Pathways and molecules inhibited by sorafenib, regorafenib, cabozantinib, ramucirumab and lenvatinib. Red Xs indicate inhibition by sorafenib, blue Xs indicate inhibition by regorafenib, yellow Xs indicate inhibition by cabozantinib, green Xs indicate inhibition by lenvatinib, and purple Xs indicate inhibition by ramucirumab.**Abbreviations:** ERK, extracellular signal-regulated kinase; MEK, mitogen-activated protein kinase kinase; Raf, RAF proto-oncogene serine/threonine-protein kinase; Ras, Ras GTPases; FGFR, fibroblast growth factor receptor; VEGFR, vascular endothelial growth factor receptor; PDGFR, platelet-derived growth factor receptor; RTKs, receptor tyrosine kinase; PDGFRA, platelet-derived growth factor receptor A; EGFR, epidermal growth factor receptor; PI3K, phosphoinositide 3-kinase; Mcl-1, induced myeloid leukemia cell differentiation protein; eIF4E, eukaryotic translation initiation factor 4E. Lenvatinib was discovered by the Tsukuba Research Laboratory in Japan as an angiogenesis inhibitor,[@CIT0070] it is a multiple receptor tyrosine kinase inhibitor that targets vascular endothelial growth factor receptors (VEGFR1-3), fibroblast growth factor receptors (FGFR1-4), PDGRα, KIT and RET.[@CIT0062],[@CIT0063],[@CIT0070] Lenvatinib blocks the VEGF pathway and inhibits angiogenesis, exerting anti-tumor activity ([Figure 3](#F0003){ref-type="fig"}). The drug has been tested in Phase II and Phase III trials for the treatment of advanced HCC. The phase III trial showed that lenvatinib is comparable to sorafenib in terms of OS. In addition, the incidence of fatal adverse events associated with lenvatinib treatment (including liver failure, cerebral hemorrhage, and respiratory failure) was 2%, which is higher compared to sorafenib (1%).[@CIT0071],[@CIT0072] A series of studies showed that lenvatinib could be an alternative first- line treatment for patients with advanced- stage HCC.[@CIT0071] In August 2018, the US FDA approved lenvatinib as a first-line therapy for advanced liver cancer. Regorafenib is a tyrosine kinase inhibitor that shares structural similarity with sorafenib, its pharmacological targets include VEGFR1-3, c-TKITIE-2, PDGFR-β, FGFR-1, RET, c-RAF, BRAF and p38MAP kinase.[@CIT0073] Regorafenib inhibits multiple protein kinases that are involved in tumor angiogenesis and oncogenesis, proliferation, and metastasis.[@CIT0074] A Phase II clinical trial conducted by Bruix et al showed that regorafenib was safe as a second-line treatment for advanced liver cancer.[@CIT0073] The progression free survival (PFS) and mOS of the 36 patients were 4.3 months and 13.8 months, respectively. Another multicenter, phase III clinical trial demonstrated that regorafenib increased survival of advanced liver cancer patients with disease progression after sorafenib treatment. Compared with the placebo group (mOS 7.8 months, mortality rate 20%), regorafenib can extend the median survival to 10.6 months and reduce the mortality rate to 13%.[@CIT0075] Cabozantinib inhibits tyrosine kinases ([Figure 3](#F0003){ref-type="fig"}), including vascular endothelial growth factor receptors 1, 2, and 3, MET, and AXL.[@CIT0065],[@CIT0076] It is a more potent inhibitor of MET, AXL, RET, FLT3, and TIE-2 compared to regorafenib.[@CIT0077] VEGF, MET, and AXL are involved in tumor proliferation and angiogenesis, thereby cabozantinib can inhibit tumor growth. Abou-Alfa et al showed that among patients who have been treated for HCC, the overall survival was 10.2 months in cabozantinib treated group, and the overall survival of the placebo group was 8.0 months.[@CIT0078] Moreover, the median progression-free survival was 5.2 months in the cabozantinib treated group whereas that of the placebo group was 1.9 months. These results showed that cabozantinib could improve the overall survival of patients who were intolerant to or had progressive disease after sorafenib treatment. Biomarkers For Targeted Therapy {#S0005-S2002} ------------------------------- Therapeutic effects of drugs might be estimated based on the genetic profiles of patients. Yeon-Su Lee et al identified a total of 1813 genomic variations associated with sorafenib responsiveness, 708 of which located within regions transcribing genes associated with sorafenib responsiveness. These genes are involved in drug absorption, distribution, and drug metabolism pathway.[@CIT0079] Another study that focused on precision medicine reported that the mutation in RSK2 led to the lasting activation of RAS, which was associated with HCC resistance to sorafenib. Moreover, Teufel et al reported 9 plasma miRNAs and 49 variants in 27 oncogenes or tumor suppressor genes, these miRNA and mutations might be used to identify HCC patients that are most likely to respond to regorafenib.[@CIT0077] In terms of targeted therapy by antibodies, data suggested that the growth of HCC cells harboring FGF or CCND1 amplification were selectively inhibited by the anti-FGF19 antibody 1A6.[@CIT0080] Immunotherapy Of HCC {#S0005-S2003} -------------------- ### Immune Microenvironment Of The Liver {#S0005-S2003-S3001} Various types of immune cells reside in the liver and produce different cytokines and growth factors in response to local stimulation. As the result, the immune cell repertoire built up an immune microenvironment that maintains the balance between immune tolerance and immune activation in the liver. The key components of the immune repertoire are macrophages, dendritic cells, myeloid-derived suppressor cells, natural killer (NK) cells, NK T cells, T cells and B cells. By single cell RNA sequencing, studies have identified subsets of immune cells that may conduct distinct immune functions. Two populations of intrahepatic CD68+ macrophages were observed in liver, one was characterized as inflammatory macrophages while the other was suggested to be tolerogenic.[@CIT0081] Similarly, the T cell and B cell population can be further divided to subsets that expressing various gene markers, including CD3+ γδ T cells and phosphoantigen-reactive γδ T cells, antigen inexperienced B cells as well as plasma B cells. The heterogeneity of intrahepatic immune cell subsets was confirmed by another single cell RNA sequencing study, which discovered CD163 + VSIG4 + Kupffer cells, MS4A1 + CD37 + subset of B cells, and CD56- and CD56 + CD8A + NKT cells.[@CIT0082] Based on the characterization of the immune microenvironment, HCC patients can be subdivided to high, medium or low immunity groups for prognosis prediction. It was reported that patients with high immunity exhibited poorly cytokeratin 19 (CK19)+, and/or Sal‐like protein 4 (SALL4)+ high‐grade HCC, which was associated with significantly better prognosis.[@CIT0083] The medium and low immunity groups were suggested to adapt distinct treatment strategies for better outcomes. Therefore, the efficacy of immunotherapy was partially determined by the immune microenvironment of individuals. Immunotherapy {#S0005-S2004} ------------- Immunotherapy has been extensively studied in the past decades, and has made significant progress in the field of cancer therapy.[@CIT0084] By targeting the immune checkpoint receptors or ligands including cytotoxic T lymphocyte antigen 4 (CTLA-4), programmed cell death protein 1 (PD-1), and programmed cell death 1 ligand 1 (PD-L1), immunotherapy breaks immune tolerance and delays tumor progression.[@CIT0085] CTLA-4 is mainly expressed on the surface of activated T cells and regulates the activation of T lymphocytes at the early stages of the tumor immune cycle ([Figure 4](#F0004){ref-type="fig"}). By binding to ligand B7-1/B7-2, and CTLA-4 suppresses the activity of T cells and facilitates tumor immune escape.[@CIT0086] CTLA-4 is mainly expressed on the surface of activated T cells and regulates the activation of T lymphocytes at the early stages of the tumor immune cycle ([Figure 4](#F0004){ref-type="fig"}). By binding to ligand B7-1/B7-2, and CTLA-4 suppresses the activity of T cells and facilitates tumor immune escape (137). CTLA-4 antibodies can block the interaction between CTLA-4 and B7-1/B7-2, promote the binding of B7 to the stimulatory receptor CD28 and restore the activity of T cells.Figure 4Pathways and molecules targeted by immunotherapy. Immune checkpoint blockade drugs suppress cancer development by inhibiting PD-1 (nivolumab, pembrolizumab), PD-L1 (durvalumab, atezolzumab) and CTLA-4 (Tremelimumab).**Abbreviations:** CTLA-4, cytotoxic lymphocyte associated protein 4; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; MHC, major histocompatibility complex; TCR, T-cell receptor. PD-1 is mainly expressed on the surface of immune cells in peripheral tissues, especially in the tumor microenvironment. It was an unexpected discover when studying T cell apoptosis. PD-1 was later identified as a receptor that negatively regulates the immune response. The PD-1 ligands PD-L1 and PD-L2 were discovered in 2000.[@CIT0087] Studies have shown that PD-1 and PD-L2 can block T cell activation, proliferation and the production of cytokines, such as interferon-γ (IFN-γ), thereby exerting a negative immune-modulatory effect which results in immunosuppression.[@CIT0088] Receptor PD-1 on the surface of T lymphocytes binds to PD-L1 on the surface of tumor cells to attenuate the immune responses and inhibits tumor cell elimination by T lymphocytes in the tumor bed.[@CIT0089] Therefore, blocking the PD-1/PD-L1 interaction can alleviate tumor-induced immunosuppression and restore the body's immune activity. Based on a series of clinical trials (Supplementary Table 1) including CheckMate 057, CheckMate 040,[@CIT0090] KEYNOTE-224,[@CIT0091] and KEYNOTE-042, both nivolumab and pembrolizumab were approved by the FDA for treatment of patients with HCC who have been previously treated with sorafenib. Biomarkers For Immunotherapy {#S0005-S2005} ---------------------------- The efficacy of immunotherapy can be predicted by genetic markers. A recent study has shown that low PD-L1 expression is associated with poor clinical outcomes in respond to immunotherapy. The objective clinical responses were observed in 26% (9/34) of patients with PD-L1 expressed in at least 1% of their tumor cells. In contrast, among patients carrying tumor cells with less than 1% of PD-L1 expression, only 19% (26/140) of them responded to ICB treatment.[@CIT0090] On a larger scale, high tumor mutation burden (TMB) is an emerging biomarker that reflects the sensitivity of patients to immune checkpoint inhibitors. TMB measured by hybrid capture-based NGS interrogating 1.2 Mb of the genome can predict clinical outcomes of anti-PD-1/PD-L1 immunotherapy in many tumor types.[@CIT0092] One study has shown that HCC patients with Wnt-β-catenin pathway mutations responded poorly to immune checkpoint blockers. All patients experienced disease progression after treatment, and their median survival was significantly shorter than the patients without the aforementioned mutations (9.1 months vs 15.2 months).[@CIT0093] Despite the fact that the incidence of high microsatellite instability (MSI-H) in HCC is estimated to be low,[@CIT0094] the FDA has approved pembrolizumab for the treatment of advanced-stage cancers with MSI-H regardless of the origin of the cancer.[@CIT0095] Collectively, these results implicate that specific DNA mutations may serve as biomarkers for ICB response prediction. Clinical Trials For HCC Treatment {#S0006} ================================= Clinical trials have been conducted to assess the performance of targeted therapy and immune therapy for HCC treatment. Several large-scale on-going trials focus on the FDA approved drugs for targeted therapy. Sorafenib and lenvatinib have been approved by the FDA as the first-line treatment for HCC patients whose disease cannot be removed by surgery. Notably, recent phase III trial confirmed that lenvatinib was non-inferior to sorafenib in overall survival in untreated HCC (NCT01761266), the progression free survival (7.4 vs 3.7 months, median) and time to progression (8.9 vs 3.7 months, median) of the lenvatinib group were both better than the sorafenib group.[@CIT0071] The trial (NCT01908426) that enrolled 707 participants has reported that cabozantinib could lead to longer overall survival and progression-free survival than placebo in patients with previously treated advanced HCC.[@CIT0078] The median overall survival and progression-free survival were extended to 10.3 months and 5.2 months respectively, with the objective response rate of 4%. Similarly, a trial involving 1000 participants for regorafenib is also underway (NCT03289273). Interestingly, a monoclonal antibody named ramucirumab was applied in targeted therapy as a vascular endothelial growth factor receptor 2 (VEGFR2) inhibitor. The results showed that ramucirumab significantly improved the median overall survival and progression free survival compared to the placebo group.[@CIT0096],[@CIT0097] At the median follow-up of 7.6 months, the median overall survival and pregression free survival reached 8.5 months and 2.8 months correspondingly. Numerous data were generated from clinical trials with drugs that have not been approved by FDA. Apatinib, a reversible dual tyrosine kinase inhibitor that selectively targets and inhibits HER2 and EGFR,[@CIT0098] is tested against placebo in a phase III clinical trial (NCT02329860). The preliminary results showed that 50% patients survived longer that 11.4 months following apatinib administration.[@CIT0099] The stable disease rate (40.9%) was considerably higher than the progressive disease rate (18.2%). Furthermore, the highly selective FGFR4 inhibitor BLU-554 (NCT02508467), the ATP-competitive cyclin A/CDK2 inhibitor milciclib (NCT03109886) and the dual TORC1/TORC2 inhibitor ATG-008 (NCT03591965) were also tested against HCC. The primary results suggest that BLU-554 might be selective for FGF19 IHC-positive HCC.[@CIT0100] No conclusions have been drawn for milciclib and ATG-008. Nivolumab (Opdivo) and pembrolizumab (Keytruda) were widely applied in clinical trials of immunotherapy. CheckMate-040 (NCT01658878) had confirmed that nivolumab induced active responses in patients with advanced HCC.[@CIT0090] The follow-up of CheckMate 459 was announced on EMSO 2019 congress, which reported that clinical benefits following nivolumab treatment were observed in predefined subgroups including hepatitis infection status, regions (Asia vs non-Asia), and vascular invation/matastasis status.[@CIT0101] However, the Bristol-Myers Squib announced in June 2019 that the randomized Phase 3 study evaluating Opdivo (nivolumab) versus sorafenib as a first-line treatment in patients with unresectable HCC failed to achieve statistical significance for its primary endpoint of overall survival (OS) per the pre-specified analysis. Similarly, the KEYNOTE-224 trial (NCT02702414) reported that pembrolizumab is effective in HCC patients previously treated with sorafenib.[@CIT0091] The primary result of KEYNOTE-240 was released on 2019 ASCO, which showed that pembrolizumab improved progression free survival over placebo. However, the comparisons were not statistically significant according to the prespecified statistical criteria.[@CIT0102] Ipilimumab (Yervoy) is an inhibitor targeting CTLA-4. Concurrent clinical trials primarily focused on the combined effects of ipilimumab and nivolumab. The follow-up of CheckMate-040 reported that the combination obtained meaningful clinical responses, the result of arm A showed that the median overall survival was extended to 23 months.[@CIT0103] Combination Of Treatments {#S0006-S2001} ------------------------- The combined application of targeted therapy, immunotherapy or locoregional therapy may give rise to great clinical outcomes. A trial in the US assessed the combined effect of sorafenib and carotuximab (TRC105), an antibody against an important angiogenic target named endoglin. The primary results indicated that TRC205+ sorafenib may give rise to additional activity regarding the treatment of advanced/metastatic HCC.[@CIT0104] A follow-up trial with the same combination is on-going, the combined effect of target therapy and immunotherapy is very promising and the outcome is expected in 2020 (NCT02560779). Another on-going multi-center study is testing the combined efficacy of nivolumab and a novel transforming growth factor beta receptor I kinase inhibitor named galunisertib. The study aimed to include 100 participants and finish in December 2019 (NCT02423343). Similarly, the effect of lenvatinib in combination with pembrolizumab as first-line treatment for HCC is tested in 750 participants, and the multi-centered, phase III trial would be completed in 2022 (NCT03713593). The efficacy of such combinations were confirmed by the primary result of one clinical trial (NCT03289533), which exhibited that the combined use of avelumab and axitiib reduced the sizes of tumors in more than 68% of patients.[@CIT0105] Regarding the combination of immunotherapies, AstraZeneca sponsored a global phase III trial that aimed to enroll 1310 patients for the assessment of durvalumab plus tremelimumab as first-line HCC treatment (NCT03298451). The trial was designed based on the promising results of a previous study, which reported that the combined treatment of durvalumab and tremelimumab showed positive clinical activity in 70% of uninfected HCC patients at ≥16 weeks follow-up.[@CIT0106] Targeted therapy is also combined with locoregional treatment. One clinical trial (NCT03838796) in China involved 482 HCC patients for the observation of the effect of lenvatinib combined with transcatheter arterial chemoembolization (TACE), and the primary study might be completed in 2021 (NCT03838796). Another clinical trial revealed that TACE plus radiofrequency ablation achieved superior efficacy compared to TACE alone, the median overall survival was improved to 29 months whereas that of the TACE group was 18 months.[@CIT0107] Geographic Distribution Of Clinical Trials In Relation To HCC Incidence Rates {#S0006-S2002} ----------------------------------------------------------------------------- To overview the geographic distribution of clinical trials, 96 on-going clinical trials for HCC treatment were summarized in Supplementary Table 2 and tagged onto the HCC incidence world map, in which the HCC incidence rate was reflected by color intensity ([Figure 5](#F0005){ref-type="fig"}). It appears that the high HCC incidence rates were mostly observed in East Asian and African countries according to the latest global cancer statistics,[@CIT0001] however, the clinical trials for HCC treatments were conducted majorly in developed regions including America and European countries. In addition, China also has an increasing number of trials initiated in recent years. It is not unexpected that most clinical trials were planned in regions with good medical care and research facilities. Nevertheless, regions with high incidence rates would have more potential participants for clinical trials, and it would be beneficial for the patients if they can sign up for clinical trials at accessible locations.Figure 5Geographic distribution of trails in relation to HCC incidence rates. The Age-standardized HCC incidence per 100,000 people of each country or region was reflected by a color intensity map, white corresponds to low incidence rate whereas black represent high incidence rate. The clinical trials for different drugs were labeled by circles and mapped based on their locations. The diameter of the circle correlates to the size of the trial. Data updated on 10th of September 2019. Treatment Strategies Received Varied Clinical Responses In Different Cancer {#S0007} =========================================================================== Being the effective PD-1 inhibitor antibodies, nivolumab and pembrolizumab are widely applied in the treatment of various malignant tumors including lung cancer and melanoma. The results of CheckMate 017 and 057 showed that nivolumab significantly extended the overall survival of squamous and nonsquamous non-small-cell lung cancer (NSCLC) patients while exhibiting lower symptom burden,[@CIT0044],[@CIT0108] which accelerated the approval of nivolumab for the treatment of NSCLC. A 5-year follow-up study reported that the median OS of NSCLC patients received nivolumab second-line treatment was 9.9 months (95% CI, 7.8 to 12.4), with the 5-year survival rate of 16%. Comparing to the low survival rate of patients with metastatic lung cancer, the clinical outcome of nivolumab treatment was considered as a milestone in the advancement of lung cancer treatment.[@CIT0109] FDA approved nivolumab for advanced melanoma treatment in 2014. The recently published result of CheckMate 172 revealed that the median overall survival was 25.3--25.8 months for acral or non-acral cutaneous melanoma after second-line nivolumab treatment, and the survival rates were 57.5% to 59%.[@CIT0110] There is no doubt that nivolumab treatment received positive clinical outcomes that significantly prolonged the overall survival of patients suffering from lung cancer or melanoma. However, the recent press release regarding CheckMate 459, a randomized Phase 3 study evaluating nivolumab versus sorafenib as a first-line treatment in patients with unresectable hepatocellular carcinoma, showed that the primary endpoint of overall survival showed no statistical difference between groups per the pre-specified analysis. The unsatisfying clinical outcomes against HCC were also observed with another PD-1 antibody pembrolizumab. The primary result of KEYNOTE-240 showed that pembrolizumab improved progression free survival over placebo, However, the comparisons between groups were not statistically significant according to the prespecified statistical criteria.[@CIT0102] In great contrast, pembrolizumab achieved meaningful clinical outcomes as second-line treatment in advanced melanoma,[@CIT0111] it also significantly improved the overall survival of lung cancer patients in comparison to chemotherapy.[@CIT0112] It was proposed that the main barrier to successful immunotherapy of HCC is the inherent immunosuppressive function of the liver. The resident immune cells subsets exhibit varied and complex immune functions, which were not fully understood.[@CIT0113] Moreover, the inter- and intra-tumor heterogeneity of HCC immune microenvironment was also believed to be the reason that the efficacy of immunotherapy appeared to be unsatisfactory.[@CIT0114] Apart from immunotherapy, chemotherapies are also limitedly applied in HCC treatment because of the adverse events and toxicities.[@CIT0115] A clinical trial applied doxorubicin and PIAF for HCC chemotherapy, PIAF showed better responsiveness (10.5% vs 20%), however, 82% patients experienced neutropenia and 57% patients suffered from thrombocytopenia. More importantly, the prognosis of the patients remained poor.[@CIT0116] The choice of treatment strategy for HCC has been limited, clinical targets that used in other liver disease was also not effective in HCC clinical trials. For example, the Ligand-activated nuclear receptor peroxisome proliferator-activated receptor alpha (PPARα) was a good clinical target for NASH/NAFLD, but it was not clear yet whether it may benefit HCC treatment.[@CIT0117] Given the facts above, it seems that finding a safe and efficient treatment for HCC still remains a global challenge nowadays. Conclusion {#S0008} ========== HCC patients often exhibit varied genetic profiles, and these differences can be applied in the early screening of HCC and prognosis prediction by genetic markers. In relation to diagnosis, differed treatment strategy can be designed for precision medicine, and potential biomarkers may be utilized to predict responses to drugs. Moreover, genetic markers were involved in the post-operational surveillance of HCC, which provides evidence of tumor reoccurrence at asymptomatic state. Apart from established treatment strategies, clinical trials for the investigation of new treatment plans were conducted globally. Although geographic disparity was observed, these studies enlightened new paths for HCC treatment, which would significantly improve patient survival. The results of the clinical trials suggest that the efficacy of immunotherapy in HCC is not desirable compared to lung cancer or melanoma. The underlying barrier would be the heterogeneity of the immune microenvironment of the liver. Given the above, studies on the immune microenvironment of the liver and the understanding of its heterogeneity are required for the improvement of HCC treatment efficacy. The work received financial support from the Bureau of Industry and Information Technology of Shenzhen, programme grant 20170922151538732. Author Contributions {#S0009} ==================== All authors contributed to data analysis, drafting or revising the article, gave final approval of the version to be published, and agree to be accountable for all aspects of the work. Disclosure {#S0010} ========== Yuling Wu, Jiajia Zhang, Zhichao Fu, Tieshan Feng, Ming Liu, Jie Han, and Shifu Chen are affiliated with HaploX Biotechnology Co. Ltd. The authors report no other conflicts of interest in this work. [^1]: **Abbreviations:** TNM, Tumor, Node, Metastasis staging; BCLC, Barcelona Clinic Liver Cancer systems; CLIP, Cancer of the Liver Italian Program (CLIP) score; JIS, Japan Integrated Staging scores; CUPI, Chinese University Prognostic Index scores; ALBI, albumin-bilirubin grading system. [^2]: **Abbreviations:** FGFR, fibroblast growth factor receptor; VEGFR, vascular endothelial growth factor receptor; PDGFR, platelet-derived growth factor receptor; PD-1, programmed cell death protein 1.

Background ========== The activation of Toll-Like Receptors (TLRs), a family of innate immune receptors, is believed to be an important step in the initiation of the inflammatory response raised against numerous pathogens. TLR3 is a mammalian pattern recognition receptor that recognizes double-stranded (ds) RNA as well as the synthetic ds RNA analog poly-riboinosinic-ribocytidylic acid (poly(I:C)) \[[@B1]\]. Activation of TLR3 by poly(I:C) or by endogenous mRNA ligands, such as those released from necrotic cells \[[@B2]\], induces secretion of pro-inflammatory cytokines and chemokines, a finding that suggests that TLR3 agonists modulate disease outcome during infection-associated inflammation \[[@B3]\]. Thus, long-term activation of TLR3 *in vivo*is thought to occur in the context of viral infection \[[@B4]\] or necrosis associated with inflammation \[[@B2]\]. *In vitro*studies have demonstrated that stimulation of lung epithelial cells with poly(I:C) elicited the secretion of multiple cytokines, chemokines, the induction of transcription factors and increased expression of TLRs \[[@B3]\]. It has also been demonstrated that poly(I:C) enhanced bradykinin- and \[des-Arg^9^\]-bradykinin-induced contractions of tracheal explants *in vitro*, an effect mediated by C-jun-amino-terminal kinase (JNK) and nuclear factor kappa B (NF-kB) signaling pathways \[[@B5]\]. Taken together, these data suggest that TLR3 activation may have a physiological consequence in the lung. Further, these data demonstrate that ligation of TLR3 initiates cascades of phosphorylation and transcriptional activation events that result in the production of numerous inflammatory cytokines that are thought to contribute to innate immunity \[[@B5]\]. Overall, these data suggest that sustained TLR3 activation can be a critical component in the modulation of infection-associated inflammatory diseases. Exacerbations in respiratory diseases such as asthma and chronic obstructive pulmonary disease (COPD) are characterized by the worsening of symptoms and a decline in lung function. Viral infections are associated with respiratory disease exacerbations \[[@B6]\] and may be associated with progression of disease. Secretion of pro-inflammatory cytokines in the lungs following viral infection represents a crucial step in promoting the inflammatory response in various lung diseases \[[@B7],[@B8]\]. A better understanding of the effects of TLR3 activation may provide insight into the mechanisms underlying virally-induced respiratory disease exacerbations. In the current study we examined the effects of TLR3 activation *in vivo*. We sought to induce long term activation of TLR3 to mimic the physiologic disease state associated with virally-induced disease exacerbations. Administration of poly(I:C) to the lungs of mice induced a marked impairment of lung function that was accompanied by the production of pro-inflammatory mediators and inflammatory cell recruitment into the airways. TLR3 appears to play a role in the effects of poly(I:C) since TLR3 KO mice were partially protected. Taken together, our data suggest an important role for TLR3 activation in impairment of lung function. Methods ======= Poly(I:C) induced cytokine secretion in BEAS-2B cells ----------------------------------------------------- The SV-40-transformed normal human bronchial epithelial cell line, BEAS-2B (ATCC, VA) was cultured in LHC-9 media without additional supplements. (Biosource, CA). 1 × 10^6^cells were seeded in collagen type I-coated T75 flasks (BD, NJ) and split every 2--3 days using 0.25% trypsin/ethylenediaminetetraacetic acid (EDTA) (Gibco, CA). Poly(I:C) (Amersham, NJ) was dissolved in phosphate-buffered saline (10 mM phosphate, 150 mM NaCl, pH 7.4; phosphate buffered saline (PBS)) at a concentration of 2 mg/ml and aliquots were stored at -20°C. For poly(I:C) stimulation, cells were incubated at 37°C with different concentrations of poly(I:C). Supernatants were collected after 24 hours and stored at -20°C or assayed immediately for cytokine secretion using a multi-plex bead assay (Biosource, CA) for detection of interferon-alpha (IFNα), interferon-gamma (IFNγ), interleukin-1-beta (IL-1β), interleukin-10 (IL10), interleukin-12p70 (IL12p70), tumor necrosis factor-alpha (TNFα), Chemokine (C-C motif) ligand 3 (CCL3), interleukin-6 (IL-6), interleukin-8 (IL-8), Chemokine (C-C motif) ligand 2 (CCL2), Chemokine (C-C motif) ligand 5 (CCL5), and Chemokine (C-X-C motif) ligand 3 (CXCL10). Limits of detection for the analytes range from 3 -- 20 pg/ml. Sample acquisition and analysis was performed using the Luminex 100S with StarStation software (Applied Cytometry Systems). Administration of Poly(I:C) to the lungs of mice ------------------------------------------------ Female C57BL/6 mice wild-type (WT) (12 weeks old) or female TLR3 knock-out (KO) mice (C57BL/6; 12 weeks old, ACE animals, PA) were anesthetized with isoflurane and different doses (10--100 μg) of poly(I:C) in 50 μl sterile PBS, or PBS alone, were administered intranasally (I.N.) Mice received three administrations of poly(I:C) (or PBS) with a 24 hour rest period between each administration. KO mice were fully backcrossed to C57BL/6 background to at least N10. All animal care was performed according to the Guide for the Care and Use of Laboratory animals and the Institutional Animal Care and Use Committee approved all studies. Whole Body Plethysmography -------------------------- Twenty-four hours following the last poly(I:C) (or PBS) administration, lung function without provocation (baseline) and airway hyperresponsiveness (AHR) to methacholine were measured using whole body plethysmography (BUXCO system). The mice were placed into the whole body plethysmograph chamber and allowed to acclimate for at least 5 minutes. Following baseline readings, mice were exposed to increasing doses of nebulized methacholine (Sigma, MO). The nebulized methacholine was administered for 2 minutes, followed by a 5-minute data collection period, followed by a 10-minute rest period before subsequent increasing-dose methacholine challenges. The increased airflow resistance was measured as Enhanced Pause (Penh) and is represented as the average penh value over the 5-minute recording period. Invasive measures of lung function ---------------------------------- Twenty-four hours following the last poly(I:C) (or PBS) administration, lung function and increased lung resistance in response to methacholine were measured using invasive measures of lung function (BUXCO system). Mice were anesthetized with 50 mg/kg sodium pentobarbital (Nembutal, Abbot Labs, IL). The trachea was cannulated with a 19 gauge cannula and the mouse was connected to a mechanical ventilator, with breath frequency of 120 and stroke volume of 0.3 mL. The mouse was connected to the plethysmograph for lung function measurements. After establishing a stable baseline of lung resistance, methacholine was administered I.V. through the tail vein (240 μg/kg). The peak resistance measured over 3 minutes was recorded. Measurement of lung inflammation -------------------------------- Following lung function measurements, mice were sacrificed by CO~2~asphyxiation and the lungs were cannulated. Bronchoalveolar lavages (BAL) were performed by injecting 1 mL of PBS into the lungs and retrieving the effluent. The lung tissues were removed and frozen. The BALs were centrifuged (1200 rpm, 10 minutes) and the cell-free supernatants were collected and stored at -80°C until analysis. The cell pellet was resuspended in 200 μl PBS for total and differential cell counts using a hemacytometer (on Wright\'s -- Giemsa-stained cytospin preparations). Measurement of proteins in bronchoalveolar lavage samples --------------------------------------------------------- The cell-free supernatants were collected and stored at -80°C until used for analyses. The multiplex assay was performed following the manufacturer\'s protocol and the LINCOplex Multiplex Immunoassay Kit (LINCO Research, St. Charles, MO). Analytes included in the analysis were MIP1α, Granulocyte Macrophage Colony Stimulating Factor (GMCSF), JE, KC, RANTES, IFNγ, IL-1α, IL-1β, Granulocyte Colony Stimulating Factor (GCSF), CXCL10, IL-2, IL-4, IL-5, IL-6, IL-7, IL-9, IL-10, IL-12(p70), IL-13, IL-15, IL-17 and TNFα. Limits of detection for the analytes range from 3 -- 20 pg/ml. Measurement of lung mRNA expression ----------------------------------- Following collection of BAL samples, the right lobes of the lung were removed and placed in Trizol total RNA isolation reagent (Life Technologies, Gaithersburg, MD). RNA was isolated using manufacturer\'s instructions of the Qiagen Rneasy Mini kit (Qiagen, Valencia, CA). Total RNA (2 μg) from pooled groups was then reverse transcribed using the OmniScript RT kit (Qiagen, Valencia, CA) according to the manufacturer\'s protocol. One hundred nanograms of cDNA was then amplified using both the TaqMan^®^Low Density Immune Profiling Array cards (Applied Biosystems, Foster City, CA), or microfluidic cards, and custom Low Density Array cards. Primer-probes with genes of interest were plated in a 384 well format following the manufacturer\'s protocol for Real-Time PCR. Data are normalized to 18s rRNA and represent fold change over PBS treated mice. Histological Analysis --------------------- Following BAL collection, the left lobes were inflated with 10% neutral buffered formalin under constant pressure then immersed in additional fixative, the right lobes were clamped with hemostats and ligated. Tissue was processed by routine methods, oriented so as to provide coronal sections and 5 micron mid-coronal sections cut and stained with hematoxylin and eosin. Morphometric analysis --------------------- A Nikon Eclipse E800 (Nikon Corporation, Tokyo, Japan) microscope was equipped with an Evolution™ MP 5.0 RTV color camera (Media Cybernetics, Inc. Silver Spring, MD). Images were captured and analyzed using Image-Pro Plus software version 5.1 (Media Cybernetics, Inc. Silver Spring, MD). GraphPad Prism version 4.03 (GraphPad Software, Inc. San Diego, CA) was used to interpret, analyze and graph the raw data. SigmaStat Statistical Software version 2.03 (SPSS, Inc. Chicago, IL) was used to perform statistical analysis on the collected data. Using the Auto-Pro tool within the Image-Pro Plus software, custom written macros were used to perform the analysis. Six TLR3 KO mice treated with poly(I:C), six WT mice treated with poly(I:C), four TLR3 KO mice treated with PBS and six WT mice treated with PBS were imaged and analyzed. No imaging or analysis was performed on areas of the lung that were torn, damaged, or folded. Tissue Density -------------- From each lung, five fields were randomly selected and imaged using a 20× objective lens. The total area of the tissue was measured and the ratio of total area of tissue to total area of field calculated. Tissue Cellularity ------------------ From each lung, five fields were randomly selected and imaged using a 20× objective lens. The total area of the nuclei was measured and the ratio of total area of nuclei to total area of field calculated. Airway Cellularity ------------------ From each lung, five airways were chosen and imaged using a 40× objective lens. A line of 100 μm in length was superimposed on the airway at a random location. The number of nuclei within the fixed distance were counted and recorded. Airway Mucosal Height --------------------- From each lung, five airways were chosen and imaged using a 40× objective lens. The image was segmented so as to include only the airway mucosa and the average thickness of the airway mucosa was measured using the curve thickness algorithm built into ImagePro. This algorithm parses the mucosa into 30,000 arc segments, measures the thickness of the mucosa at each arc segment and calculated the average thickness for the mucosa. Statistical analysis -------------------- Specific statistical methods are described in the figure legends. Graphs and summary statistics were also used to assess the results. All statistical tests were 2-sided. Except for where noted, all p-values presented are unadjusted for multiple comparisons. Results ======= Poly(I:C) induces a marked inflammatory response in the lungs of mice --------------------------------------------------------------------- Intranasal administration of three once-daily doses of poly(I:C) resulted in a dose-dependent inflammatory cell influx into the lung. There was a significant increase in total cells in the BAL samples at 50 and 100 μg poly(I:C) compared to PBS treated mice (Figure [1A](#F1){ref-type="fig"}). This increase in total cellularity in the BAL samples was partially due to a significant influx of neutrophils (Figure [1B](#F1){ref-type="fig"}) and mononuclear cells (Figure [1C](#F1){ref-type="fig"}). Due to the robust response at 50 and 100 μg, these doses of poly(I:C) were used in our subsequent studies. {#F1} In an effort to understand the responses to poly(I:C) treatment in the lung at a molecular level, Taqman real-time PCR analyses of the lung tissues was performed. Multiple administrations of poly(I:C) elicited up regulation of a number of pro-inflammatory genes, TLRs and their associated intracellular signaling molecules (Table [1](#T1){ref-type="table"}). TLR genes that were up regulated at the mRNA level as a result of TLR3 stimulation included TLR2, TLR3, TLR7, and TLR9 with approximately 7, 5, 11, and 56 fold increases respectively. In addition there was dramatic increase in CXCL10, TNFα, CCL2, CCL3, and CCL7 gene expression as well as interferon regulatory factor 7 (IRF7), interferon-stimulated transcription factor 3 (ISGF3G), 2\'-5\'-oligoadenylate synthetase 2 (OAS2), and protein kinase-R (PKR.) ###### Poly(I: C) induces up regulation of gene expression of cytokines, chemokines, signaling molecules and TLRs in the lungs of mice. **Cytokines/Chemokines** **Fold Increase** --------------------------- ------------------- CXCL10 357.38 TNFα 78.45 CCL2 76.62 CCL3 30.49 CCL7 48.38 **TLRs** TLR9 55.78 TLR7 10.86 TLR3 5.41 TLR2 6.96 **Transcription Factors** IRF7 22.92 ISGF3G 4.45 Enzymes OAS2 10.76 PKR 9.32 Mice were administered PBS or 100 μg poly(I:C) I.N. every 24 h for three days. 24 h following the last poly(I:C) administration, lungs were lavaged, excised and frozen. RNA was isolated from the tissue and real-Time PCR was then performed. Data are expressed as fold change in mRNA expression over PBS-treated animals and represent pooled cDNA from 6 -- 8 mice. Poly(I:C) administration also induced elevated protein levels of cytokines, chemokines, and growth factors in the lavage including significant increases of IFNγ, IL-1α, IL-6, TNFα, CXCL10, JE, KC, MIP-1α, RANTES, GCSF and GMCSF (Table [2](#T2){ref-type="table"}). There were no changes in IL-1β, IL-2, IL-4, IL-5, IL-7, IL-9, IL-10, IL-12(p70), IL-13, IL-15, or IL-17 (data not shown) among the groups. These data demonstrate that poly(I:C) administered I.N. elicits a cascade of events resulting in the expression and secretion of multiple pro-inflammatory cytokines, and chemokines as well as the up regulation of TLR gene expression. ###### Poly(I: C) induces the secretion of cytokines, chemokines, and growth factors into the airways. Treatment -------- -------------- ---------------------- IFNγ 11.0 +/- 1.6 52.2 +/- 11.2 \*\* IL-1α 16.5 +/- 1.2 21.8 +/- 1.4 \* IL-6 8.8 +/- 1.5 879.0 +/- 171.2 \*\* CXCL10 30.3 +/- 5.9 411.3 +/- 34.9 \*\* JE 11.7 +/- 1.2 798.7 +/- 182.6 \*\* KC 6.2 +/- 1.3 55.4 +/- 6.5 \*\* GCSF 5.2 +/- 0.7 60.6 +/- 6.8 \*\* MIP1α 37.7 +/- 6.3 441.1 +/- 61.6 \*\* RANTES 0.5 +/- 0.04 155.8 +/- 41.6 \*\* TNFα 2.3 +/- 0.33 81.2 +/- 13.7 \*\* GMCSF 19.1 +/- 2.1 33.5 +/- 4.5 \* Mice were administered PBS or 100 μg polyI:C (I.N.) every 24 h for three days. 24 h following the last polyI:C administration, BALs were performed. Analyte levels in BAL were determined. Data are expressed as mean pg/ml ± SEM from 6 -- 8 mice. Statistical significance was determined using the Mann-Whitney test. \* p \< 0.05, \*\*p \< 0.01 when compared to PBS-treated mice. There was no measureable change in the following cytokines (data not shown): IL-1β, IL-2, IL-4, IL-5, IL-7, IL-10, IL-12(p70), IL-9, IL-13, IL-15, or IL-17. Histological analyses of the lungs were performed to better understand the pathology induced by poly(I:C) administration. Representative micrographs from H&E stained lung sections are shown (Figure [2](#F2){ref-type="fig"}). The histology of the control lungs was unremarkable in that the lungs exhibited normal pulmonary architecture and resident cells. The most remarkable changes induced by poly(I:C) were a marked perivascular and a moderate peribronchiolar interstitial inflammatory infiltrate. There were also signs of pulmonary edema as evidenced by a widening of the interstitial space surrounding the airways and vasculature in the poly(I:C) treated mice. The alveolar septa were thickened and contained numerous inflammatory cells, consistent with an interstitial pneumonitis. Few inflammatory cells were observed in the alveolar spaces, but as the bronchoalveolar fluids were collected, most of the cells in the alveoli were probably lost from analysis. The other remarkable changes observed were thickening of the bronchiolar epithelium consistent with hypertrophy. The hypertrophy was accompanied by an increase in the granularity of the cytoplasm of the bronchiolar epithelium, however, there was no evidence for increased mucus production by PAS staining. There was no notable increase in goblet cells. {#F2} The results of the morphometric analysis are shown in Table [3](#T3){ref-type="table"}. Reflecting the increase in interstitial penumonitis there was a 1.7 fold increase in tissue density and a 2 fold increase in overall tissue cellularity. In the small airways, there was a 1.7 fold increase in the mucosal height, reflecting the mucosal hypertrophy and no change in cellularity (data not shown). ###### Morphometric analysis of lungs from WT PBS control and poly(I:C)-treated, and TLR-3 KO PBS control and poly(I:C)-treated mice. ----------------------------------------------------------------------------------------------------- Group Tissue Density\ Tissue Cellularity\ Airway Mucosal Height\ % % μm ------------------------------------ ----------------- --------------------- ------------------------ WT PBS 32 ± 2 8 ± 1 15 ± 1 WT Poly(I:C) 50 ± 5\* 16 ± 2\* 26 ± 4\* Fold Increase (Compared to WT PBS) 1.7 2 1.7 KO PBS 36 ± 7 9 ± 1 18 ± 3 KO Poly(I:C) 49 ± 6\* 14 ± 3\* 19 ± 2\*\* Fold Increase (Compared to KO PBS) 1.4 1.5 NC ----------------------------------------------------------------------------------------------------- NC = No change, \* Different from respective PBS control. \*\* Different from poly(I:C)-treated WT. p \< 0.01 using T-test to compare groups. Poly(I:C) activates BEAS2B epithelial cells ------------------------------------------- The morphometric data identified the induction of mucosal hypertrophy in WT mice following poly(I:C) challenge. To further elucidate the effects of poly(I:C) on epithelial cells, the response of the normal human lung epithelial cell line, BEAS-2B, to poly(I:C) was investigated. Similar to the mouse in vivo data, where analysis was performed 24 hours post final poly(I:C) challenge, BEAS-2B cells responded to a range of poly(I:C) concentrations (16 to 1000 ng/ml) in a dose-dependent manner by secreting a number of cytokines observed in the mouse lungs including IL-6, IL-8, CCL2, CCL5, and CXCL10 (Fig. [3](#F3){ref-type="fig"}), consistent with previous findings \[[@B9]-[@B11]\]. There was no change in response to poly(I:C) in the other analytes included in the multiplex (data not shown), nor was there any obvious change in morphometric parameters of the stimulated cells. {#F3} TLR3 stimulation leads to impairment of pulmonary function ---------------------------------------------------------- In order to investigate the functional consequences of TLR3 ligation, we measured lung function in poly(I:C)-treated mice. Airway hyperresponsiveness to increasing doses of methacholine was measured using whole body plethysmography (WBP) (Figure [4A](#F4){ref-type="fig"}). Poly(I:C)-challenged mice exhibited greater airway hyperresponsiveness to methacholine. Poly(I:C)-challenged mice also exhibited an increase in baseline penh in the absence of provocation, measured using WBP (Figure [4B](#F4){ref-type="fig"}). To confirm the effects of poly(I:C) on lung function, invasive lung function measurements were also performed and the results confirmed those obtained using WBP (Fig [4C](#F4){ref-type="fig"}). {#F4} Poly(I:C)-induced inflammatory cell influx is attenuated in TLR3 KO mice ------------------------------------------------------------------------ In order to elucidate whether the effects induced by poly(I:C) were mediated through TLR3, we treated TLR3 KO and age-matched WT control mice with three repeated doses of 100 μg poly(I:C) I.N. 24 hours after the third dose, mice were euthanized and bronchoalveolar lavage samples were collected. There was a significant increase in total cells, including both neutrophils and mononuclear cells in the bronchoalveolar lavage samples harvested from WT mice administered 3 doses of 100 μg poly(I:C) compared to PBS treated mice (Figure [5A--C](#F5){ref-type="fig"}). In contrast, TLR3 KO mice displayed a reduced influx of inflammatory cells compared to WT mice. The increase in total cells, neutrophils, and mononuclear cells in poly(I:C)-treated WT mice was 18, 70, and 15 fold over PBS treated mice respectively. In contrast, poly(I:C)-treated TLR3 KO mice had increases of 3, 6, and 3 fold in total cells, neutrophils, and mononuclear cells over PBS treated TLR3 KO mice. {#F5} TLR3 KO mice are protected from poly(I:C)-induced bronchial epithelial cell hypertrophy --------------------------------------------------------------------------------------- Representative micrographs from H&E stained lung sections from control and poly(I:C)-treated TLR3 KO mice are shown in Figure [2](#F2){ref-type="fig"}. The histology of the control lungs was largely unremarkable. However, focal eosinophilic mixed inflammatory infiltrates were observed in 2 of 4 TLR3 KO mice examined. The ranges of changes observed in the TLR3 KO mice treated with poly(I:C) was similar to that observed in wild type mice (described above). Perivascular and peribronchiolar interstitial chronic inflammatory infiltrates were present in these mice but were somewhat less extensive. The pulmonary edema and interstitial pneumonitis were modestly attenuated and the bronchiolar epithelial hypertrophy observed in the wild type mice treated with Poly(I:C) was markedly attenuated in the TLR3 KO mice. The attenuation of the effects of poly(I:C) is corroborated by the morphometric analysis (Table [3](#T3){ref-type="table"}). Although there was only a slight change in tissue density in the KO mice compared to WT, the bronchiolar epithelial hypertrophy was decreased substantially. TLR3 KO mice are protected from poly(I:C)-induced changes in lung function at baseline -------------------------------------------------------------------------------------- In order to investigate whether TLR3 plays a role in poly(I:C)-induced lung function impairment, lung function was measured following poly(I:C) treatment of TLR3 KO mice and WT age-matched controls. As shown in Figure [6B](#F6){ref-type="fig"}, TLR3 KO mice were protected from poly(I:C)-induced changes at baseline. The increase in penh observed at baseline following poly(I:C) administration was significantly reduced in TLR3 KO mice. {#F6} Discussion ========== Exacerbations of respiratory diseases such as asthma and COPD are often associated with concomitant respiratory viral infections. Since TLR3 is activated by viral dsRNA, the purpose of the current study was to better understand the functional consequences of TLR3 activation in vivo. Administration of poly(I:C), a synthetic TLR3 ligand, to the lungs of mice induced marked inflammation accompanied by impaired lung function. TLR3 KO mice were partially protected from the effects of poly(I:C) demonstrating the involvement of TLR3. These data provide further support for a role of TLR3 in respiratory diseases and suggest a potential mechanisitic pathway for viral exacerbations. Upon activation, TLR3 recruits a Toll-IL-1 receptor (TIR) -- related adaptor protein inducing interferon (TRIF), which activates both IFN-regulatory factor 3 (IRF3) and NF-kB \[[@B12]\] and \[[@B13]\]. In our model, following poly(I:C) administration to the lungs, there was an up regulation of TLR3, -2, -7, and 9 gene expression and their associated signaling molecules. Previous *in vitro*studies have demonstrated that activation of TLR3 with poly(I:C) induces up regulation of its own expression as well as the expression of other TLRs. For example, poly(I:C) up regulates mRNA for TLR2, 3 and 4 in airway smooth muscle cells \[[@B14]\] and TLR2, 3, 6 and 10 in lung epithelial cells \[[@B3]\]. *In vivo*, the up regulation of TLR mRNA expression may have occurred as a result of expression of TLRs on infiltrating cells or through up regulation on resident lung cells. Indeed, monocytes express all of the known TLRs \[[@B15]\]. In contrast, neutrophils have been shown to express all the TLRs except TLR3 \[[@B16]\]. Within the lung, all of the known TLRs have been found to be expressed by human primary bronchial epithelial \[[@B3]\] and smooth muscle cells \[[@B14]\]. The up regulation of multiple members of the TLR family, as a consequence of activation of one TLR, may indicate the creation of an environment of hyper-responsiveness to pathogen insult whereby, an exacerbation event could be triggered in the event that the lung is exposed to other toll-ligands. In support of this hypothesis, it has been shown that infection of airway epithelial cells with *Hemophilus influenza*induced the secretion of CXCL-8, up regulated TLR3 expression and increased the responsiveness to a secondary challenge of *Rhinovirus*. Interestingly, inhibition of TLR3 with small interfering RNA, inhibited the *Rhinovirus*-induced CXCL-8 production \[[@B17]\]. In addition this same group demonstrated that pretreatment with *Rhinovirus*resulted in delayed bacterial clearance when a secondary infection was induced using nontypeable *Hemophilus influenza*. Sajjan et al. showed that this may be the result of decreases in transepithelial resistance or compromised tight junctions and loss of zona occludins-1 and junctional adhesion molecule-1 \[[@B18]\]. Taken together these studies suggest that activation of TLRs, such as TLR3 can result in a perturbation of the local environment, specifically dysregulation of the airway epithelium thereby supporting an environment primed for an exacerbation. We are currently focusing efforts in our laboratory toward identifying the composition of the mononuclear cell populations in this model including the activation state of various cell types including dendritic cells. In a review by Fe *et. al.*it is summarized that TLR3 can induce a variety of cytokines in human dendritic cells including IFNβ, and CXCL10 \[[@B19]\]. *In vivo*TLR3 agonism by synthetic dsRNA also resulted in a profound up regulation of the expression and secretion of multiple pro-inflammatory cytokines, chemokines, and growth factors. *In vitro*studies have demonstrated that activation of TLR3 by dsRNA on different cell types including natural killer cells \[[@B20]\], epithelial cells \[[@B3],[@B21],[@B22]\], and smooth muscle cells \[[@B14]\] results in increased expression and/or secretion of pro-inflammatory cytokines including IL-6, CXCL-8, CCL-2, CCL-5, CXCL-10, GM-CSF, TNFα and IFNγ. A likely source of cytokines following poly(I:C) administration may be the airway epithelium since activation of BEAS-2B cells *in vitro*induced a profile of pro-inflammatory cytokines similar to that observed following *in vivo*poly(I:C) challenge. TLR3 has been identified and functionally characterized in mouse tracheal muscle \[[@B23]\] and in primary human small airway epithelial cells \[[@B21],[@B3],[@B22]\]. Previous *in vitro*studies have also demonstrated the secretion of inflammatory mediators following TLR3 activation of epithelial cells\[[@B3],[@B3],[@B21]\]. The up regulation of pro-inflammatory cytokines and chemokines provides an inflammatory milieu supporting the infiltration of inflammatory cells into the airways and lung interstitium. Accompanying the inflammation-rich pathology was the presence of bronchial epithelial cell hypertrophy. The hypertrophic cells extended into the secondary and tertiary airways. Epithelial cell hypertrophy is normally associated with increased mucus production \[[@B23]\]. However, in the current study, there was no evidence for increased mucus production by PAS staining. Given the distribution of goblet cells in normal mouse airways, which is restricted to the main bronchi and primary bronchioles, the data suggest that the hypertrophic epithelial cells are not mucus-producing goblet cells. Along with the demonstration that poly(I:C), acting as a TLR3 ligand, results in an inflammatory response *in vivo*, the study presents a novel finding that stimulation of TLR3 results in a measurable impairment of lung function both without provocation and characterized by increased AHR to methacholine. Similar changes in baseline lung function have also been described in mice exposed to Respiratory Syncytial virus (RSV) \[[@B24]\]. Recent studies have demonstrated that pre-exposure of mouse tracheas to poly(I:C) in vitro increases the expression of bradykinin B1 and B2 receptors on the smooth muscle and confers AHR to bradykinin \[[@B25]\]. Notably, inhibition of the bradykinin B1 receptor confers protection from acetylcholine-induced AHR following allergen sensitization and challenge \[[@B26]\]. In contrast, AHR to histamine following *parainfluenza-3*infection in guinea pigs was inhibited by a bradykinin B2 receptor antagonist \[[@B27]\]. Taken together these data suggest a role for bradykinin in TLR3-induced airway dysfunction. In the current study some, but not all, functional responses were protected in TLR3 KO mice following multiple administrations of poly(I:C). Specifically, they were protected from baseline lung function changes in response to poly(I:C), however protection from AHR in response to provocation with methacholine did not result in significant protection. Further, the pro-inflammatory mediators produced following poly(I:C) administration were not modulated in TLR3 KO mice. Unpublished data from our laboratory has shown that TLR3 KO mice were significantly protected from a single administration of poly(I:C) with respect to pro-inflammatory mediators in the bronchoalveolar lavage (data not shown), indicating that mediators released in response to acute activation with poly(I:C) may be more TLR3 dependent. This data suggests that another receptor for poly(I:C) may be available. Indeed, since a percentage of TLR3 KO mice succumb to poly(I:C)-induced shock, it suggests that poly(I:C) may still signal in the absence of TLR3 \[[@B1]\]. Indeed, dsRNA can also signal through dsRNA-dependent protein kinase (PKR) \[[@B28]\], RIGI \[[@B29]\] and MDA-5 \[[@B30]\]. The potential redundancy in the dsRNA downstream pathways may be an explanation for the incomplete protection observed in TLR3 KO mice. Understanding the different signaling pathways involved in recognition of dsRNA by the host has been a major area of focus by many researchers. Le Goffic *et al.*demonstrated that sensing of *influenza A virus*by TLR3 and RIG-I regulates a pro-inflammatory response. In contrast, RIG-I but not MDA-5 also mediates type I IFN-dependent antiviral signaling response\[[@B31]\]. Use of non-poly(I:C) TLR3 ligands is necessary to further define the impact of TLR3-specific signaling on pulmonary pathophysiology. Interestingly, TLR3 KO mice demonstrate protection from *influenza A*virus-induced lung function impairment accompanied by reduced inflammation and improved survival \[[@B32]\]. These data taken along with the inflammatory consequences of TLR3 activation suggest that sustained TLR3 activation may also contribute to severe exacerbations of chronic pulmonary diseases. In summary, the data presented in this study suggest that sustained TLR3 activation may play an important role in respiratory disease pathogenesis. A better understanding of the effects of TLR3 activation will provide additional insight into the mechanisms underlying virus-induced exacerbations associated with respiratory diseases. Additionally, these studies provide an opportunity to identify suitable targets for therapeutic intervention for respiratory disease exacerbations. Competing interests =================== NCS, JS, HAR, KAS, RJL, DDE, PJB, LAM, PA M, RAB, LRS, DEG, RTS, MLM, and AMD are current or former employees of Centocor Research & Development, Inc. RAF and LA declare that they have no competing interests. Authors\' contributions ======================= NCS conceived of the study and participated in its design and coordination as well as all analysis. JS, LAM, LRS, DEG, RTS, MLM, and AMD participated in the design and coordination of the studies. HAR, and KAS executed the in-life portion of the studies. RJL carried out the BEAS2B studies. DDE and PJB carried out the histopath analysis of the lungs. PAM carried out the statistical analysis of all data sets. RAB carried out the analysis of cellular infiltrates in the lung. RAF and LA made the TLR3 KO mice and gave input on the design of the studies and the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ The authors would like to thank Cory M. Hogaboam, Ph.D. Associate Professor, Immunology Program, Department of Pathology, University of Michigan Medical School, for assistance in guiding the invasive measurements of lung function.

INTRODUCTION {#s1} ============ The commonly used chemotherapeutic drugs cisplatin and carboplatin are generally considered to exert their cytotoxicity by inducing DNA damage. These drugs interact with DNA to form intra- and inter-strand crosslinks, which must be repaired for the cell to proliferate \[[@R1]\]. Hence, cells that grow more rapidly or are limited in their capacity to repair DNA should disproportionately suffer cell death, which often occurs by apoptosis. Consequently, growth factor driven tumour growth and deficits in the ability to rapidly repair DNA both enhance the ability of cisplatin to induce cell death \[[@R1]--[@R5]\]. DNA-platin adducts are also aberrantly bound by a range of nuclear proteins, and this in general enhances cell death by delaying their repair \[[@R6], [@R7]\]. Important among these nuclear proteins are members of the High Mobility A and B families (HMGA and HMGB), which display elevated affinities for the bent DNA structure of the platin adducts via their HMGA-box and HMGB-box DNA binding domains \[[@R8]--[@R10]\]. Upstream Binding Factor (UBF) is an abundant multi-HMGB-box transcription factor that defines the active state of ribosomal RNA (rRNA) gene chromatin by replacing the core histones and is essential for transcription of these genes \[[@R11]--[@R13]\]. It has long been known that UBF has a particularly high affinity for cisplatin-DNA adducts, which may act as molecular decoys to attract this factor away from the rRNA genes and in so doing suppress their transcription \[[@R14]--[@R19]\]. Since transcription of the rRNA genes is the central event in the assembly of ribosomes, the protein factories of the cell, their activity is essential for cell growth and proliferation. The ability of cisplatin adducts to act as decoys for UBF binding could, therefore, enhance the drugs cytotoxicity either by inhibiting DNA repair, by inhibiting ribosome assembly, or both. The rRNA genes are transcribed by RNA polymerase I (RPI/PolI), which is dedicated to this task. UBF is an HMGB-box DNA binding protein and one of the two essential RPI basal transcription factors \[[@R11], [@R20]--[@R23]\]. UBF is generally thought to mediate binding of the pre-initiation factor SL1/TIF1B and pre-initiation complex (PIC) assembly at the rRNA gene promoter. But UBF also forms a nucleosome-like structure that replaces histone chromatin throughout the transcribed regions of the rRNA genes and is able to regulate RPI transcription elongation in response to growth factor signalling \[[@R11], [@R24]--[@R28]\]. Ribosomal biogenesis is the process by which ribosomal RNA (rRNA) is transcribed, processed and assembled with the ribosomal proteins to create ribosomes \[[@R21], [@R29]\]. This energy consuming process is accomplished in the nucleolus and requires the action of the three RNA polymerases along with more than 200 different proteins and several hundred snoRNP complexes. Regulation of ribosome synthesis constitutes a major determinant of the increased protein synthesis needed for cell proliferation and, as such, its up-regulation occurs in many cancers \[[@R30], [@R31]\]. An increased nucleolar volume reflects this increased ribosome synthesis, and is therefore a biomarker of cancer that was recognized already 80 years ago \[[@R32]--[@R34]\]. In fact, rRNA transcription is a common and probably an essential target of many oncogenes (Myc \[[@R35], [@R36]\], SV40-T antigen \[[@R37], [@R38]\] and the Ras and mTOR signalling pathways \[[@R39]--[@R43]\]), and tumour suppressors (p53 \[[@R44]\], ARF \[[@R45]--[@R47]\], Rb \[[@R48], [@R49]\] and PTEN \[[@R50]\]). Ribosomal biogenesis is such a central process in cell growth that it is also under the direct surveillance of the p53 pathway \[[@R51]\]. Defects in rRNA gene transcription \[[@R52]\], rRNA processing \[[@R53]\] or ribosome assembly \[[@R54]\] all cause p53 stabilization and arrest of cell proliferation. These findings have led to the investigation of small molecule inhibitors of ribosomal transcription as potential chemotherapeutic agents. Inhibition of the RPI pre-initiation factor SL1/TIF1B \[[@R55]\] or induced proteasome degradation of the RPI large subunit \[[@R56]\] both lead to arrest of rRNA synthesis and mediate cell death dependent on p53 function. However, the key to successful cancer therapy remains the selective targeting of cancer cells, and since p53 is often inactivated in human cancers, therapies that depend on functional p53 have limited application. Our data now suggest that inhibition the RPI basal transcription factor UBF (Upstream Binding Factor) represents a particularly valuable p53-independent target for cancer therapy. Here we show that displacement of UBF and ablation of rRNA synthesis are very early effects of cisplatin treatment, and that in the absence of cisplatin, elimination of UBF protein is sufficient to induce fully penetrant apoptotic cell death. Using cell cultures conditional for UBF expression, we find that complete loss of ribosome biogenesis induces synchronous and fully penetrant, p53-independent cell death by apoptosis specifically in cells transformed by known oncogenes. The data argue that a major factor in the cytotoxicity of cisplatin and similar drugs is their ability to inhibit the function of UBF. This suggests that UBF itself represents a preferred target for anticancer drug development. RESULTS {#s2} ======= Previous data has clearly indicated that cisplatin treatment of human cells leads to a partial or full displacement of human UBF and inhibition of rRNA synthesis \[[@R14], [@R15], [@R17], [@R18]\]. However, to what extent this plays a role in the selective cytotoxicity of cisplatin is not known. When the Mouse Embryonic Fibroblast (MEF) derived cell line NIH3T3 was treated for 4 h with 30 μM cisplatin, a concentration calculated to be equivalent to the dose commonly used in therapy (e.g see \[[@R57], [@R58]\]), a large proportion of endogenous UBF was displaced from nucleoli and scattered throughout the nucleus at a large number of foci ([Figure S1](#SD1){ref-type="supplementary-material"}). These foci were devoid of the other nucleolar proteins fibrillarin and RPI (data not shown), which remained together in dense nuclear bodies somewhat similar to the nucleolar precursor bodies forming on conditional deletion of the *Ubf* gene \[[@R11]\]. Cisplatin displaces UBF from the mouse rRNA genes and arrests their transcription {#s2_1} --------------------------------------------------------------------------------- To better understand the effect of cisplatin, we repeated and extended these studies using the independently isolated, iMEF cell line (*Ubf^wt/wt^*/*Er-cre*^+/+^/*SvT*) previously characterized by Hamdane et al. \[[@R11]\]. Already after 4 h exposure of these cells to 30 uM cisplatin, UBF was seen to coalesce from its normal specular distribution within nucleoli into more intense foci, while fibrillarin showed some degree of coalescence but was less affected (Figure [1](#F1){ref-type="fig"}). When these cells were cultured for a further \~18 h in the absence of cisplatin, the UBF foci became more intense and UBF, but not fibrillarin, formed foci throughout the nucleus. The timing of the changes in UBF delocalization corresponded closely with changes in the interaction of UBF with the rRNA genes and with the transcription of these genes (Figure [2](#F2){ref-type="fig"}). After 4 h of cisplatin treatment a mean reduction in UBF binding of 80% was observed across the 47S precursor rRNA coding region, and this corresponded with an 80% reduction in rRNA synthesis (Figure [2B](#F2){ref-type="fig"} and [2C](#F2){ref-type="fig"}). (Due to its 5′ position in the 47S precursor, 18S rRNA synthesis was slightly less affected at 4 h than the 28S rRNA, but nevertheless was reduced by over 60% after 4 h cisplatin exposure, data not shown). 22 h after cisplatin exposure rRNA synthesis was no longer detectable. The effects of cisplatin on the activity of the rRNA genes also corresponded to an arrest of cell proliferation, no increase in the viable cells count being detected after the 4 h cisplatin treatment, and to a subsequent loss of viability (Figure [2D](#F2){ref-type="fig"}). These data suggest that the timeline of cisplatin cytotoxicity is consistent with its effects being mediated at least in part by disruption of UBF function, and the arrest of rRNA gene transcription and, hence, of ribosome biogenesis. Since cisplatin is a key chemotherapeutic agent that acts by inducing apoptotic cell death somewhat selectively in transformed cells (e.g. \[[@R3]\]), we sought to determine whether or not this activity could also be explained by the inhibition of UBF function. {#F1} ![Cisplatin coordinately displaces UBF from the rRNA genes and arrests their transcription\ **A.** Timeline of cisplatin treatment and culture of *Ubf^wt/wt^*/*Er-cre*^+/+^/*SvT* iMEFs. **B.** ChIP analyses of UBF occupancy across the rRNA gene 47S transcribed region. The positions of amplicons is indicated above the histogram showing the UBF occupancy normalized to that in the mock treated cells. **C.** Synthesis rate of rRNA determined by \[^3^H\]-uridine metabolic labelling of mock treated cells and at the indicated times post cisplatin treatment. The upper panel displays a fluorogram of \[^3^H\]-rRNA, the central panel the corresponding EtBr stained total 18S rRNA, and the lower panel quantitation of \[^3^H\] incorporation into 47S rRNA performed in triplicate. **D.** Live cell counts at indicated times following cisplatin treatment performed in triplicate.](oncotarget-06-27519-g002){#F2} UBF loss disrupts nucleolar functions in both primary and transformed MEFs {#s2_2} -------------------------------------------------------------------------- We previously generated mice conditional for the *Ubf* gene and demonstrated that loss of this gene arrested mouse development at the morula stage \[[@R11]\]. SV40Tt immortalized Mouse Embryonic Fibroblasts or iMEFs (*Ubf^fl/fl^*/*Er-cre*^+/+^/*SvT*) generated from these mice allowed us to show that UBF was essential for transcription of the rRNA genes and for the existence of a functional nucleolus \[[@R11]\]. Not surprisingly, despite their limited proliferation potential, primary MEFs derived from these mice also require UBF for rRNA synthesis and for the maintenance of nucleoli ([Figure S2](#SD1){ref-type="supplementary-material"}). Thus, UBF loss in primary MEFs recapitulated the effects observed in the transformed iMEFs. Transformed iMEFs, but not primary MEFs, undergo synchronous apoptosis following *Ubf* inactivation {#s2_3} --------------------------------------------------------------------------------------------------- Despite the apparently identical responses of the primary MEFs and the iMEFs to UBF loss, it became obvious from observing these cultures that the two cell types behaved very differently macroscopically. Inactivation of rRNA gene transcription in the *Ubf^fl/fl^*/*Er-cre*^+/+^/*SvT* iMEFs induced changes in cell morphology soon after complete UBF depletion and the shutdown of rRNA synthesis. iMEFs became highly elongated and this presaged cell death as determined by plasma membrane failure (trypan blue), mitochondrial membrane depolarization (MitoTracker) and loss of clonal viability ([Figure S3A](#SD1){ref-type="supplementary-material"} to [S3D](#SD1){ref-type="supplementary-material"}). Control *Ubf^wt/wt^*/*Er-cre*^+/+^/*SvT* iMEFs suffered none of these effects, clearly demonstrating that cell death was exclusively the result of inactivation of the *Ubf* gene. Interestingly, we detected no selective reduction of total cellular RNA in the *Ubf^fl/fl^*/*Er-cre*^+/+^/*SvT* iMEFs relative to their wild type counterparts during UBF depletion that might suggest a role of ribosome depletion in the selective induction of apoptosis (data not shown). In contrast to the behavior of the *Ubf^fl/fl^*/*Er-cre*^+/+^/*SvT* iMEFs, the primary *Ubf^fl/fl^*/*Er-cre*^+/+^ MEFs showed no evidence of major morphological changes and survived in culture for many days following complete UBF loss, maintaining plasma membrane integrity and mitochondrial function ([Figure S3A](#SD1){ref-type="supplementary-material"} to [S3C](#SD1){ref-type="supplementary-material"}). To better understand the different responses of the transformed iMEFs and primary MEFs to UBF loss, we analyzed them for typical markers of cell death. TUNEL (terminal deoxynucleotidyl transferase-mediated dUTP nick end-labeling) analysis detects the single strand DNA cleavage that is characteristic of the early stages of apoptotic cell death. *Ubf^fl/fl^*/*Er-cre*^+/+^/*SvT* iMEFs became TUNEL positive at 96 h pHT, just 24 h after complete shutdown of rRNA synthesis, while the control *Ubf^wt/wt^*/*Er-cre*^+/+^/*SvT* iMEFs remained TUNEL-negative throughout (Figure [3A](#F3){ref-type="fig"}). The TUNEL signal was fully penetrant and occurred synchronously, *Ubf^fl/fl^*/*Er-cre*^+/+^/*SvT* iMEFs being TUNEL-negative at 72 h pHT but all becoming TUNEL-positive at 96 h pHT. In contrast, the *Ubf^fl/fl^*/*Er-cre*^+/+^ primary MEFs remained TUNEL-negative at least until 144 h pHT, (Figure [3B](#F3){ref-type="fig"} and data not shown). {ref-type="supplementary-material"} and [S2C](#SD1){ref-type="supplementary-material"}.](oncotarget-06-27519-g003){#F3} Concomitant with the onset of TUNEL-positive apoptosis, the *Ubf^fl/fl^*/*Er-cre*^+/+^/*SvT* iMEFs were also found to activate Caspase 3 from 96 h pHT, as determined by the release of the 17kD peptide (p17) cleavage product (Figure [4A](#F4){ref-type="fig"}). In contrast, the control *Ubf^wt1wt^*/*Er-cre*^+/+^/*SvT* iMEFs displayed no significant cleavage of Caspase 3, consistent with the lack of a TUNEL signal. Further, Caspase 3 was not significantly activated in the primary MEFs (Figure [4B](#F4){ref-type="fig"}). Though a certain level of cleavage was detected in both *Ubf^fl/fl^* and *Ubf^wt/wt^* MEFs, this was much weaker than observed in the *Ubf^fl/fl^*/*Er-cre*^+/+^/*SvT* iMEFs as can be seen by comparison with Staurosporin-treated iMEFs. {ref-type="supplementary-material"} and [S2C](#SD1){ref-type="supplementary-material"}.](oncotarget-06-27519-g004){#F4} Interestingly, unlike the deletion of UBF, deletion of the essential RPI initiation factor TIF1A/Rrn3 did not induce apoptosis in SV40Tt transformed MEFs. 4-HT treatment of *TIF1A^fl/fl^*/*Er-cre*^+/+^/*SvT*:MEFs resulted in complete depletion of TIF1A by 48 h pHT, as observed for UBF, but did not lead to activation of Caspase 3, nor to a TUNEL signal ([Figure S4A](#SD1){ref-type="supplementary-material"} and [S4B](#SD1){ref-type="supplementary-material"}). Thus, the induction of apoptosis in the SV40Tt transformed cells was not a general property of the arrest of rRNA gene transcription, suggesting it is specific to UBF depletion. Given that the iMEFs were initially immortalized by the SV40 Tt oncogene (*Sv-T*), known to inactivate p53 \[[@R59], [@R60]\], it was not surprising to find the p53 levels in these cells were constitutively elevated and were not further induced by inactivation of the *Ubf* gene or by treatment with Staurosporin (Figure [4A](#F4){ref-type="fig"}). Thus, it was unclear whether or not p53 played a role in the apoptotic response in these cells. This question is directly addressed below using homozygous inactivation of the p53 gene. However, it should be noted that inactivation of the *Ubf* gene in the primary MEFs did not enhance the levels of p53 protein, which remained extremely low throughout (Figure [4B](#F4){ref-type="fig"}). Apoptosis is accompanied by the generation of a "nucleosomal ladder" of DNA cleavage {#s2_4} ------------------------------------------------------------------------------------ Apoptosis is often accompanied by inter-nucleosomal cleavage of genomic DNA to generate a "nucleosomal ladder" \[[@R61], [@R62]\], due to the result of the release of the nuclease EndoG from mitochondria \[[@R63], [@R64]\]. Beginning at or before 120 h pHT we observed this characteristic nucleosomal fragmentation of genomic DNA in the apoptotic *Ubf^fl/fl^*/*Er-cre*^+/+^/*SvT* but not in the control *Ubf^wt/wt^*/*Er-cre*^+/+^/*SvT* iMEFs (Figure [4C](#F4){ref-type="fig"}), nor in the corresponding primary MEFs (data not shown). Thus, three distinct markers; TUNEL signal, Caspase 3 cleavage and a nucleosomal ladder, indicated that on UBF loss MEFs underwent classic apoptotic cell death after oncogenic transformation with SV40-T, while UBF loss in untransformed MEFs induced none of these markers. UBF loss blocks proliferation and DNA replication, causing cell cycle arrest {#s2_5} ---------------------------------------------------------------------------- To better understand the mechanisms leading to apoptosis in the transformed iMEFs, we determined the effects of *Ubf* inactivation on cell cycle progression and cell division. Before tamoxifen treatment, the *Ubf^fl/fl^*/*Er-cre*^+/+^/*SvT* iMEFs displayed a large (\~50%) actively replicating S-phase population (Figure [5A](#F5){ref-type="fig"}). Their proliferation was near completely arrested by 48 pHT, corresponding with the elimination of UBF protein and with the near complete shutdown of rRNA synthesis (e.g. see [Figure S2C](#SD1){ref-type="supplementary-material"} to [S2E](#SD1){ref-type="supplementary-material"} and \[[@R11]\]). By 72 h pHT, iMEFs had also stopped active DNA replication and the G2 population abruptly increased at the expense of S-phase cells, while the fraction of G1/G0 cells remained constant (Figure [5A](#F5){ref-type="fig"} and [S5A](#SD1){ref-type="supplementary-material"}). Concomitantly, the mitotic index fell to zero as determined by the fraction of cells phosphorylated on serine 28 of histone H3 (H3-S28P) (Figure [5C](#F5){ref-type="fig"} and [5D](#F5){ref-type="fig"} and [Figure S5B](#SD1){ref-type="supplementary-material"}). Together these data suggested that many apparently G2 iMEFs were unable to complete their passage through mitosis. Parallel analysis of *Ubf^wt/wt^*/*Er-cre*^+/+^/*SvT* iMEFs post tamoxifen treatment revealed none of these effects, DNA replication and cell proliferation continuing essentially unabated (Figure [5A, 5C](#F5){ref-type="fig"} & [5D](#F5){ref-type="fig"} and [S5B](#SD1){ref-type="supplementary-material"}). {ref-type="supplementary-material"} and [S2C](#SD1){ref-type="supplementary-material"}.](oncotarget-06-27519-g005){#F5} The situation was somewhat different in the primary *Ubf^fl/fl^*/*Er-cre*^+/+^ and control *Ubf^wtlwt^*/*Er-cre*^+/+^ MEFs (Figure [5B](#F5){ref-type="fig"}). These cells proliferated more slowly than iMEFs, and only a small fraction (\~20%) was ever actively engaged in DNA synthesis. Further, regardless of UBF status these cells gradually arrested DNA replication between 24 h and 48 h pHT and displayed a corresponding increase in G2 cells, that is up to 24 h earlier than for the UBF-null iMEFs. Thus, the primary MEFs underwent a natural slowing or arrest of proliferation regardless of UBF status, while proliferation arrest in the iMEFs was a direct result of the loss of UBF protein. This suggested that the catastrophic cell death observed in the iMEF cultures was related to their inability to assume a quiescent state. In contrast, MEFs naturally arrested proliferation and became quiescent regardless of UBF status or 4-HT treatment (Figure [5B](#F5){ref-type="fig"}), and hence this may have protected them from cell death on inactivation of the *Ubf* gene. Essentially then, UBF loss specifically targeted the SV40-Tt transformed cells for apoptotic cell death, and what is more the effect was fully penetrant. This suggests that inhibition of UBF or of ribosome biogenesis might represent an ideal target for the development of cancer specific cytotoxic drugs. Apoptosis induced by UBF loss is p53 independent {#s2_6} ------------------------------------------------ P53 is often required for the induction of apoptosis, hence its inactivation in many cancers represents a serious limitation to the efficacy of chemo- and radiation therapies \[[@R65]--[@R67]\]. The SV40 Tt oncogene is known to inactivate p53 \[[@R59], [@R60]\], suggesting that apoptosis induced by UBF loss did not depend on functional p53. To directly evaluate the role of p53, we generated p53-null MEFs either wild type or conditional for UBF (*Ubf^fl/fl^*/*Er-cre*^+/+^/*p53*^−/−^) (Figures [6](#F6){ref-type="fig"} and [S6A](#SD1){ref-type="supplementary-material"}) and found that they were immortalized and hence could be passaged indefinitely. Despite this, they did not undergo apoptotic cell death on inactivation of the *Ubf* gene, and displayed neither a TUNEL signal nor Caspase 3 cleavage (Figure [6A](#F6){ref-type="fig"} and [6B](#F6){ref-type="fig"}). In contrast, after transformation with the SV40 Tt-antigens (SV40-T), the resulting p53-null (*Ubf^fl/fl^*/*Er-cre*^+/+^/*p53*^−/−^/*Sv-T*) iMEFs underwent synchronous and homogeneous TUNEL positive apoptosis two days after loss of UBF, exactly as observed for the p53 positive iMEFs (Figure [6C](#F6){ref-type="fig"}). Thus, even in the complete absence of p53 the loss of UBF was sufficient to induce apoptosis in the SV40-Tt transformed iMEFs. However, in this case no cleavage/activation of Caspase 3 was detected (Figure [6D](#F6){ref-type="fig"}). {ref-type="supplementary-material"} and [S2C](#SD1){ref-type="supplementary-material"}.](oncotarget-06-27519-g006){#F6} p53-independent apoptosis is a general response to UBF loss in oncogene stressed cells {#s2_7} -------------------------------------------------------------------------------------- It was striking that UBF loss induced fully penetrant apoptosis in SV40-Tt transformed MEFs even in the complete absence p53. To determine if this effect was specific to the SV40-Tt oncogene or occurred under other oncogenic stresses, we investigated UBF-loss in MEFs transformed by the Ras and Myc oncogenes, commonly correlated with human cancers \[[@R68]\]. *Ubf^fl/fl^*/*Er-cre*^+/+^/*p53*^−/−^ MEFs were transformed by introduction of the Ras oncogene or a combination of the Ras and Myc oncogenes and the effects of inactivation of the *Ubf* gene were followed. In each case UBF was essentially eliminated by 48 h pHT ([Figure S6B](#SD1){ref-type="supplementary-material"}) and we observed a synchronous and homogeneous onset of TUNEL-positive apoptosis 48 h later, exactly as for SV40-Tt transformation (compare Figure [7A](#F7){ref-type="fig"} with [7B](#F7){ref-type="fig"} and [7C](#F7){ref-type="fig"}). Colony forming assays also showed that in each case cell death approached 100% ([Figure S6D](#SD1){ref-type="supplementary-material"}). In the case of SV40-Tt and combined Ras/Myc transformation we also observed a "nucleosomal ladder" of apoptotic DNA cleavage starting at 96 h pHT, that is at or just after the appearance of the TUNEL signal ([Figure S6C](#SD1){ref-type="supplementary-material"}), though this cleavage was not detected in the cells transformed with Ras alone. {ref-type="supplementary-material"} and [S2C](#SD1){ref-type="supplementary-material"}.](oncotarget-06-27519-g007){#F7} Oncogenic stress may induce apoptosis by aberrantly driving cells into S-phase {#s2_8} ------------------------------------------------------------------------------ When the untransformed p53-null cells (*Ubf^fl/fl^*/*Er-cre*^+/+^/*p53*^−/−^) were analyzed by FACS, we were surprised to find that, quite unlike the SV40-Tt transformed (*Ubf^fl/fl^*/*Er-cre*^+/+^/*p53*^+/+^/*Sv-T*) iMEFs (Figure [5A](#F5){ref-type="fig"}), UBF depletion caused a significant accumulation of cells in G1 at the expense of the actively replicating S-phase cells (Figure [8A](#F8){ref-type="fig"}). The G2 cell population displayed only a small increase and this anyhow closely resembled that observed for the control *Ubf^wt/wt^*/*Er-cre*^+/+^/*p53*^−/−^ cells. In contrast, the *Sv-T*, *Ras* and *Ras/Myc* transformed *Ubf^fl/fl^*/*Er-cre*^+/+^/*p53*^−/−^ cells displayed the same G2 phase accumulation as seen for the p53-positive iMEFs (compare Figure [8B](#F8){ref-type="fig"} with [5A](#F5){ref-type="fig"}). This suggested that transformation drives cells into and through S-phase regardless of their ability to generate a full complement of ribosomes. Such a situation would be likely to lead to gross replicative errors and hence could explain the highly penetrant apoptosis occurring in both the p53-positive and p53-null transformed MEFs, but not in the untransformed p53-null MEFs. {#F8} DISCUSSION {#s3} ========== Our data suggest that the ability of cisplatin to cause the displacement of UBF from the nucleolus is a key mechanism by which this drug induces selective cell death, since the simple loss of UBF induces a rapid and highly penetrant apoptosis in oncogenically stressed cells. We have shown that conditional deletion of the *Ubf* gene induces apoptosis specifically in cells transformed by viral and cellular oncogenes. Apoptosis following UBF loss was observed not only in cells expressing SV40Tt, but also in cells expressing the oncogenes Ras and Myc. What is more, in each case apoptosis was found to be fully penetrant, all cells without exception underwent apoptotic cell death. Strikingly, the onset of apoptosis occurred synchronously in all cells two days following complete loss of UBF. Significantly, the induction of TUNEL-positive cell death was completely independent of p53, since it occurred with the same timing and penetrance even after homozygous deletion of the *p53* gene. In contrast, before oncogenic transformation primary cell cultures survived complete loss of UBF for many days after the transformed cells entered apoptosis and never underwent apoptosis. These data strongly suggest that the commonly used chemotherapeutic drug Cisplatin, and by analogy, Carboplatin exert their cytotoxicity in large part by hijacking UBF, displacing it from the nucleolus and inhibiting ribosome biogenesis. In fact, inhibition of ribosome biogenesis may be a more general property of the cytotoxic drugs used in chemotherapy than previously realized, including rapamycin analogs, 5-fluorouracyl and camptothecin \[[@R52], [@R69]\]. Azacytidine (Azacitidine, Vidaza) and deoxyazacytidine (Decitabine) are DNA methyltransferase inhibitors that have been shown to be active in treating myelodysplastic syndromes and acute myeloid leukemia (AML) \[[@R70]--[@R72]\]. The initial studies of azacytidine already showed that it strongly inhibits ribosome biogenesis, and almost certainly does so by preventing rRNA methylation \[[@R73], [@R74]\]. More recently, deoxyazacitidine was also shown to inhibit ribosome biogenesis by inhibiting rRNA processing, though the underlying mechanism of action is quite different and involves loss of rRNA gene silencing and aberrant RNA polymerase II transcription of these genes \[[@R13], [@R75]\]. Recent studies of small molecule inhibitors that target ribosome biogenesis have further shown this may be a very valid clinical approach to treating a range of cancers \[[@R55], [@R56], [@R76], [@R77]\]. However, while cell death was independent of p53 in the case of the GC-rich DNA interacting drug BMH-21 \[[@R56]\], it was found to be dependent on a functional p53 in the case of CX-5461, which is believed to target the pre-initiation complex factor SL1 \[[@R55]\]. Our data showing TIF1A/Rrn3-loss does not induce apoptosis even in the presence of p53 clearly excludes the explanation that the cytotoxicity of these drugs is simply a function of their ability to suppress rRNA synthesis. Why then inhibition of UBF can induce apoptotic cell death with such penetrance and in the complete absence of p53 is for the still a matter of conjecture. However, it is amost certainly related to the role of UBF in forming a specialized chromatin structure on the active rRNA genes \[[@R11]\]. Loss of this structure would yield the rRNA gene arrays highly susceptible to damage, and given the GC-richness of the rRNA genes the same could be argued for both cisplatin and BMH-21 drugs. The Nucleolar Organizer Regions (NORs) each encompass around 40 rRNA gene units on the short arms of the five human acrocentric chromosomes \[[@R78]\]. These loci are particularly susceptible to DNA breakage and are subject to high levels of inter- and intra-chromosomal recombination \[[@R79]--[@R81]\]. Indeed, Robertsonian translocations have long been known to predominantly involve exchanges between the short arms of human acrocentric chromatids that often create fusions with chromatids of a metacentric chromosome \[[@R82]\]. Recent data strongly suggests that these and similar chromosome translocations result from disruption of the active chromatin structure of the rRNA genes, which in turn affects chromosome pairing causing aberrant resolution of mitotic chiasmata and fusion between non-homologous chromatids \[[@R83]\]. Loss of UBF clearly disrupts the chromatin structure of the rRNA genes, leaving them at least transiently as naked DNA, and would necessarily leave these genes highly susceptible to DNA damage and breakage. Since transformed iMEFs continue replication during UBF depletion, the disruption of rRNA gene chromatin would exacerbate the effects of DNA breakage, probably inhibit homologous repair processes and hence destabilize the genome. Indeed such destabilization has recently been observed as a result of siRNA knockdown of UBF \[[@R84]\]. MATERIALS AND METHODS {#s4} ===================== Isolation and cultures of MEFs and iMEFs {#s4_1} ---------------------------------------- The generation of conditional *Ubf^fl/fl^Er-cre*^+/+^ and control mouse lines was previously described \[[@R11]\]. The p53-null allele was introduced by crossing to strain *129-Trp53^tm1Tyj^*/*J* (Jackson Laboratory Stock \# 002080). Primary mouse embryonic fibroblasts (MEFs) from E14.5 *Ubf^fl/fl^*/*Er-cre*^+/+^ and isogenetic *Ubf^wt/wt^Er-cre*^+/+^ MEFs and corresponding *p53^−/−^* MEFs were prepared as previously described \[[@R11], [@R85]\]. Cells were cultured in Dulbecco\'s modified Eagle medium (DMEM)-high glucose (Life Technologies), supplemented with 10% fetal bovine serum (Wisent) and Antibiotic/Antimycotic (Wisent). Where indicated, Cisplatin (Sandoz) was added to the cell culture medium from a 100 mM solution in DMSO to give a final concentration of 30 μM and cells incubated for 4 hr at 37°C. The culture medium was then replaced with medium without cisplatin and cells incubated for a further 16 h at 37°C, before processing for immunofluorescence as described below. MEFs were immortalized by the introduction of the SV40 Tt antigens by transfection with the pBSV0.3T/t, a modification of the pBSV-early vector \[[@R86]\] kindly provided by E. W. Khandjian. The Ras and Ras/Myc transformed MEFs were generated by transfection or co-transfection with the plasmids pWZL-Ras-hygro and pBabe-c-myc-puro (kind gifts from Gerardo Ferbeyre) into *Ubf^fl/fl^*/*Er-cre*^+/+^/*p53*^−/−^ MEFs and subsequent hygromycin or double hygromycin/puromycin selection. Inactivation of *Ubf* or *Tif1a* in cell culture, and analysis of genotype, RNA and proteins {#s4_2} -------------------------------------------------------------------------------------------- As previously described \[[@R11]\], cells were initially plated in 6 cm petri dishes (0.8 × 10^6^ cells each) and cultured for 18 hours in DMEM, high glucose, 10% fetal bovine serum. To activate ER-Cre, 4-hydroxytamoxifen (4-HT) was added to a final concentration of 50 nM, and after 4 hr incubation the medium replaced with fresh medium without 4-HT and cells harvested for analysis at various time points. In the case of *Tif1a*, cells were treated with 50 nM 4-HT, 0 h, then this treatment was repeated at 9 h, 24 h and 33 h later to ensure complete gene excision. Analyses of RNA, protein and genotype were systematically carried out on parallel cell cultures. Cells were genotyped by PCR before and after 4-HT treatment using the primers: A; 5′TGATCCCTCCCTTTCTGATG, B; 5′TGGGGATAGGCCTTAGAGAGA, C; 5′CACGGGAAAACAAGGTCACT, ([Figure S2B](#SD1){ref-type="supplementary-material"}). Metabolic labelling of RNA was carried out just before cell harvesting by addition of 10 μCi \[^3^H\]-uridine (PerkinElmer) to the culture medium and incubation for a further 3 h. RNA was extracted with Trizol (Life Technologies) according to the manufacturer\'s protocol and analyzed by gel electrophoresis, fluoroimaging (ENHance, PerkinElmer) and RNA species quantitated by scintillation counting as previously described \[[@R39], [@R40]\]. For total protein, cells were washed with cold PBS, scraped into PBS, centrifuged 30 s at 14 000 r.p.m., then resuspended in sodium dodecyl sulphate (SDS) loading buffer. After fractionation on 8%, 12% or 5--15% gradient SDS--polyacrylamide gel electrophoresis (SDS-PAGE \[[@R87]\]), cell extracts were analysed by standard Western blotting procedures. Chromatin immunoprecipitations (ChIP) {#s4_3} ------------------------------------- ChIP was performed as previously described \[[@R11], [@R88]\]. The amplicon coordinates relative to the 47S rRNA initiation site (BK000964) were as follows: 47SPr, 45133--40; ETS, 3078--3221; ITS1, 6258--6432; 28S, 10215--10411; T1--3, 13412--13607. Antibodies for western blot, immunofluorescence and ChIP {#s4_4} -------------------------------------------------------- Rabbit antibodies against UBF, RPI large subunit (A194), TTF-1 and TIF1A were generated in the laboratory. All other antibodies were obtained commercially; Anti-Caspase-3, -p53 and -H3S-28phospho (Cell Signalling), anti-Tubulin (Sigma) and anti-Fibrillarin (Covance). Immunofluorescence {#s4_5} ------------------ Cells were washed with PBS, fixed in 4% paraformaldehyde /PBS for 15 minutes and permeabilized with 0.5% Triton/PBS for 5 minutes. Incubation with primary antibody was performed for 1 h in PBS-5% BSA or 5% goat serum and cells were stained with AlexaFluor 488/568 conjugated anti-rabbit or -mouse IgG (Molecular Probes) and counterstained with DAPI. After mounting in 50% glycerol/50% 0.2 M Na-glycine, 0.3 M NaCl, 3D epifluorescent image stacks were generated on a Leica DMI6000B microscope equipped with a 63x or 100x objective and an Orca C4742--80-12AG camera (Hamamatsu). Image stacks were then deconvoluted and analyzed using Volocity software (Perkin-Elmer Improvision). Alternatively, image stacks were generated on a Leica SP5-II confocal microscope equipped with a 63x objective and running in standard scanning mode, and analyzed using Volocity software (Perkin-Elmer Improvision). FACs analysis and determination of Mitotic Index {#s4_6} ------------------------------------------------ Cells were stained for ongoing DNA synthesis using the Click-iT^®^ EdU Alexa Fluor^®^ 647 Flow Cytometry Assay Kit (Life Technologies) following the manufacturer\'s protocol and subsequently with propidium iodide (PI) immediately before analysis by the cytometry service of the CHU de Québec Research Centre using a FACSCanto II flow cytometer and FACSDiva 6.1.2 software (Becton Dickinson). Parallel cultures were stained with anti-H3S-28phospho antibody and DAPI and imaged by epifluorescence on the Leica DMI6000B microscope using 20 and 40x objectives. The Mitotic Index was calculated as the ratio of H3S-28phospho-positive to DAPI positive nuclei. Tunnel assays {#s4_7} ------------- Tunnel assays were performed with a Click-It Tunnel assay kit, Alexa 488 Imaging System, (Life Technologies). Cells were seeded in 35 mM petri dishes, fixed and processed according to the manufacturer\'s protocol and visualized by epifluorescence on the Leica DMI6000 B microscope using a 20x objective. Colony formation assays {#s4_8} ----------------------- The SV40-T, Ras and Ras/Myc transformed *Ubf^fl/fl^*/*Er-cre*^+/+^/*p53*^−/−^ and the isogenic wild-type MEF cells cultured in 100 mm petri dishes were treated with 50 nM 4-HT (Sigma) on day 0. The medium was changed after four hours to remove 4-HT, and on day 2 each culture was replated in duplicate at dilutions of 10 000, 50 000, 100 000, and 200 000 cells per 60 mm petri. On day 6 and day 12 petri dishes were fixed for 5 mins with 4% paraformaldehyde/PBS and stained with 0.05% crystal violet in distilled water (filtered) for 30 mins. Petri dishes were then washed 3 times with water and left inverted to dry before being photographed. MitoTracker assays {#s4_9} ------------------ Cells were plated in Ibidi 35 mm thin bottom petri dishes for subsequent live cell microscopy and treated for 4 h with 50 nM 4-HT (Sigma) and further cultured as standard for 96 h to induce UBF loss. Cells were then treated with 25 nM MitoTracker DeepRedTM (Life Technologies) for 20 mins at 37°C in DMEM minus serum. Petri dishes were washed once with DMEM minus serum and then incubated in FluoroBrite DMEM (Life Technologies). Finally, live image stacks were generated on the Leica SP5-II confocal microscope and analyzed using Volocity software (Perkin-Elmer Improvision). SUPPLEMENTARY DATA FIGURES {#s5} ========================== We wish to thank Dr Lucie Jeannotte for providing the p53 null allele mice and advice on their use, Dr Ross Hannan and Elaine Sanij for discussion at various stages of this work, and Dr A. Brunet of the Cytology Laboratory of the Research Centre of the Québec University Hospital Centre (RC-CHUQ) for FACS analyses. This work was funded by operating grants from the Canadian Institutes of Health Research (CIHR, MOP12205) and from the Cancer Research Society (CRS/SRC). The Research Centre of the Québec University Hospital Centre (CHU de Québec) is supported by the Fonds de recherche du Québec - Santé (FRQS). **CONFLICTS OF INTEREST** None.

González‐Varo JP, Arroyo JM, Jordano P. The timing of frugivore‐mediated seed dispersal effectiveness. Mol Ecol. 2019;28:219--231. 10.1111/mec.14850 30151871 1. INTRODUCTION {#mec14850-sec-0001} =============== Mutualistic interactions constitute an essential element of biodiversity that provides key ecological functions, from mycorrhizal‐mediated mineral nutrition to animal‐mediated pollination and seed dispersal (Jordano, [2016](#mec14850-bib-0032){ref-type="ref"}; Schupp, Jordano, & Gómez, [2017](#mec14850-bib-0054){ref-type="ref"}). A major challenge in understanding the role of mutualistic interactions in community dynamics lies in assessing not only the *immediate* outcome*,* but also the *delayed* effect that interacting species have on their partners (Schupp et al., [2017](#mec14850-bib-0054){ref-type="ref"}). The immediate outcome is the successful occurrence of interactions and can be largely assessed as a quantity component (number of events; for example, number of seeds dispersed). The delayed outcome is the "per capita" effect a species has on the demography of its interacting partner and can be assessed as a quality component (e.g., probability of recruitment of a dispersed seed). This framework allows the total effect of interactions to be estimated for both sides of the mutualism as the product of quantity and quality (quantity × quality), which results in a measure of effectiveness (Schupp et al., [2017](#mec14850-bib-0054){ref-type="ref"}). Seed dispersal mediated by frugivorous animals is a central process in the dynamics and regeneration of many vegetation types (Herrera, [2002](#mec14850-bib-0024){ref-type="ref"}; Jordano, [2013](#mec14850-bib-0031){ref-type="ref"}; Wang & Smith, [2002](#mec14850-bib-0065){ref-type="ref"}). The effectiveness framework has provided a conceptual and analytical tool for the study of seed dispersal mutualisms from the plant\'s perspective for more than two decades (Schupp, [1993](#mec14850-bib-0051){ref-type="ref"}; Schupp, Jordano, & Gómez, [2010](#mec14850-bib-0053){ref-type="ref"}). Research on seed dispersal effectiveness has largely focused both on gut passage effects and on the spatial patterns of seed deposition generated by different disperser species, with consequences for recruitment (Jordano & Schupp, [2000](#mec14850-bib-0033){ref-type="ref"}; Schupp, [1993](#mec14850-bib-0051){ref-type="ref"}; Schupp et al., [2010](#mec14850-bib-0053){ref-type="ref"}). Gut passage effects on germination can vary among groups of seed dispersers (Nogales et al., [2017](#mec14850-bib-0040){ref-type="ref"}; Traveset, [1998](#mec14850-bib-0060){ref-type="ref"}), while the fates of seeds and seedlings often differ between microhabitats and habitat types due to spatial variation in biotic and abiotic factors, such as seed predator activity, irradiance, soil humidity or intra‐ and interspecific competition (Gómez‐Aparicio, [2008](#mec14850-bib-0015){ref-type="ref"}; González‐Varo, Nora, & Aparicio, [2012](#mec14850-bib-0019){ref-type="ref"}; Rey & Alcántara, [2014](#mec14850-bib-0048){ref-type="ref"}; Schupp, [1995](#mec14850-bib-0052){ref-type="ref"}). The latter explains why space has been a major factor when considering the quality of seed dispersal services provided by different animal partners (Calviño‐Cancela & Martín‐Herrero, [2009](#mec14850-bib-0006){ref-type="ref"}; Escribano‐Ávila et al., [2014](#mec14850-bib-0009){ref-type="ref"}; Rother et al., [2016](#mec14850-bib-0050){ref-type="ref"}; Schupp et al., [2010](#mec14850-bib-0053){ref-type="ref"}). The quantity component has been assessed either by combining information on microhabitat use by different disperser species with measures of seed deposition across microhabitats (Donoso, García, Rodríguez‐Pérez, & Martínez, [2016](#mec14850-bib-0008){ref-type="ref"}; Jordano & Schupp, [2000](#mec14850-bib-0033){ref-type="ref"}) or through visual identification of dispersers from droppings with seeds (only feasible with taxonomically distant dispersers; e.g., Calviño‐Cancela & Martín‐Herrero, [2009](#mec14850-bib-0006){ref-type="ref"}; McConkey, Brockelman, & Saralamba, [2014](#mec14850-bib-0036){ref-type="ref"}). The quality component has been assessed either by field experiments of seed survival, germination and seedling establishment (Escribano‐Ávila et al., [2014](#mec14850-bib-0009){ref-type="ref"}) or by monitoring these demographic processes in naturally dispersed seeds and seedlings (Donoso et al., [2016](#mec14850-bib-0008){ref-type="ref"}). These studies have shown how effective dispersers can compensate a modest quantity component with seed deposition in high‐quality sites for recruitment (Calviño‐Cancela & Martín‐Herrero, [2009](#mec14850-bib-0006){ref-type="ref"}; Escribano‐Ávila et al., [2014](#mec14850-bib-0009){ref-type="ref"}; McConkey et al., [2014](#mec14850-bib-0036){ref-type="ref"}). Surprisingly, the temporal dimension, in terms of between‐ and within‐season variability, has remained completely overlooked in the study of seed dispersal effectiveness. The fruiting period of many plants dispersed by animals can last for many months (Hamann, [2004](#mec14850-bib-0021){ref-type="ref"}; Snow & Snow, [1988](#mec14850-bib-0055){ref-type="ref"}) and even for most of the year (Herrera, [1984](#mec14850-bib-0022){ref-type="ref"}), a common phenomenon in tropical regions (Griz & Machado, [2001](#mec14850-bib-0020){ref-type="ref"}; Peres, [1994](#mec14850-bib-0044){ref-type="ref"}). During such long periods, the biotic and abiotic factors affecting seed dispersal and seedling recruitment can change dramatically (Figure [1](#mec14850-fig-0001){ref-type="fig"}a). First, the local disperser assemblage can be temporally structured during the fruiting period because many migratory animals are frugivores, mostly birds and bats (e.g., Herrera, [1984](#mec14850-bib-0022){ref-type="ref"}; Stiles, [1980](#mec14850-bib-0056){ref-type="ref"}; Thomas, [1983](#mec14850-bib-0058){ref-type="ref"}). This involves that the contribution of a migrant species to the dispersal of a given plant species can be confined to a particular, narrow temporal window. Moreover, the populations seed predators (or parasites) can fluctuate within and between seasons (Ostfeld & Keesing, [2000](#mec14850-bib-0042){ref-type="ref"}), as well as their predation pressure on a given seed species due to changes in the abundance of alternative food resources (García, Martínez, & Obeso, [2007](#mec14850-bib-0012){ref-type="ref"}; Price & Joyner, [1997](#mec14850-bib-0045){ref-type="ref"}). In addition, climatic seasonality can determine more or less suitable periods for seedling emergence and survival, particularly in highly seasonal ecosystems (Garwood, [1983](#mec14850-bib-0014){ref-type="ref"}; Gómez‐Aparicio, [2008](#mec14850-bib-0015){ref-type="ref"}; Washitani & Masuda, [1990](#mec14850-bib-0066){ref-type="ref"}). Lastly, even the intrinsic quality of seeds might vary between early‐ and late‐ripening fruits owing to resource limitation (Vaughton & Ramsey, [1998](#mec14850-bib-0062){ref-type="ref"}) or to the activity of different pollinator species during the plant\'s flowering phenology (Ivey, Martinez, & Wyatt, [2003](#mec14850-bib-0026){ref-type="ref"}; Valverde, Gómez, & Perfectti, [2016](#mec14850-bib-0061){ref-type="ref"}). All these sources of temporal variability (Figure [1](#mec14850-fig-0001){ref-type="fig"}a) suggest that, in many plant species and across biomes, the timing of seed dispersal could be as crucial modulating seed dispersal effectiveness as the seed deposition sites. There could be also interactive "space--time" effects shifting the relative quality of microhabitats throughout the fruiting period. Tackling this issue empirically is challenging and requires answering the questions *who, where* and *when* dispersed the seeds, and *what* happened to them next. ![(a) Biotic and abiotic factors important for seed dispersal effectiveness that can vary substantially during the whole fruiting period of plant species; their position along the *x*‐axis does not indicate their timing. The *y*‐axis indicates the number of dispersed seeds; and the bell‐shaped curve is merely hypothetical and illustrative (temporal distributions could be asymmetric or multimodal). (b) Study design to address temporal variation in seed dispersal effectiveness through an assessment of quantity and quality components in different periods of the fruiting phenology and across the same target microhabitats \[Colour figure can be viewed at <http://wileyonlinelibrary.com>\]](MEC-28-219-g001){#mec14850-fig-0001} Here, we test timing effects on seed dispersal effectiveness. We used as study case a Mediterranean shrub dispersed by frugivorous birds whose fruiting period can last for up to 9 months. We approached the spatiotemporal variation in effectiveness using a factorial study design ("period × microhabitat") that allowed us to evaluate the quantity and quality components in different microhabitats and different periods of the plant\'s fruiting phenophase (Figure [1](#mec14850-fig-0001){ref-type="fig"}b). We identified the bird species responsible for seed deposition (i.e., quantity) through DNA barcoding applied on dispersed seeds---animal DNA can be successfully extracted from the surface of defecated or regurgitated seeds collected in the field (González‐Varo, Arroyo, & Jordano, [2014](#mec14850-bib-0017){ref-type="ref"}; González‐Varo, Carvalho, Arroyo, & Jordano, [2017](#mec14850-bib-0018){ref-type="ref"}). We evaluated the probability of recruitment (i.e., quality) through a series of field experiments on sequential demographic processes (seed viability, seed predation, germination and seedling survival). Then, for three groups of birds differing in their migratory behaviour, we assessed how seed dispersal effectiveness varied in space and time, and the importance of accounting for timing when estimating the overall effectiveness throughout the whole fruiting phenophase. 2. MATERIALS AND METHODS {#mec14850-sec-0002} ======================== 2.1. The plant--frugivore system {#mec14850-sec-0003} -------------------------------- The study plant species was the lentisc (*Pistacia lentiscus*, Anacardiaceae), an evergreen dioecious shrub with hemispherical shape (Supporting Information [Figure S1](#mec14850-sup-0001){ref-type="supplementary-material"}), that constitutes a dominant component of woodlands and forests in warm, lowland areas across the Mediterranean Basin (<http://www.worldwildlife.org/biomes>; Yi, Wen, Golan‐Goldhirsh, & Parfitt, [2008](#mec14850-bib-0068){ref-type="ref"}). We chose this species because its fruiting phenology can last up to 9 months, from late summer to early spring (Jordano, [1989](#mec14850-bib-0030){ref-type="ref"}), a long period during which many biotic and abiotic factors important for seed dispersal effectiveness vary considerably (Figure [1](#mec14850-fig-0001){ref-type="fig"}a, Supporting Information [Figure S1](#mec14850-sup-0001){ref-type="supplementary-material"}). Its single‐seeded fruits are spherical drupes of \~5 mm in diameter with a lipid‐rich pulp (Herrera, [1987](#mec14850-bib-0023){ref-type="ref"}). Fruits are red prior to ripening and black when ripe (Supporting Information [Figure S1](#mec14850-sup-0001){ref-type="supplementary-material"}). Its lentil‐shaped seeds are 3--5 mm in diameter and 2--3 mm in width. A significant proportion of the fruits contain unviable seeds due to abortion, parthenocarpy or parasitism by *Megastigmus pistaciae*, a chalcidoid wasp (see Supporting Information [Figure S1](#mec14850-sup-0001){ref-type="supplementary-material"}; Jordano, [1989](#mec14850-bib-0030){ref-type="ref"}; Verdú & García‐Fayos, [1998](#mec14850-bib-0064){ref-type="ref"}). Fruits can remain red throughout the fruiting season because colour is also associated with seed viability and most red fruits contain unviable seeds (Jordano, [1989](#mec14850-bib-0030){ref-type="ref"}). Lentisc fruits are consumed---and its seeds dispersed---by a diverse guild of small frugivorous birds, mainly belonging to families Sylviidae, Turdidae and Muscicapidae, which includes resident birds, sub‐Saharan migrants and European wintering migrants (Herrera, [1984](#mec14850-bib-0022){ref-type="ref"}; Jordano, [1988](#mec14850-bib-0029){ref-type="ref"}, [1989](#mec14850-bib-0030){ref-type="ref"}). The lentisc has an ephemeral seed bank because its seeds lack dormancy and germinate within the year (García‐Fayos & Verdú, [1998](#mec14850-bib-0013){ref-type="ref"}). 2.2. Study site {#mec14850-sec-0004} --------------- We conducted our study in Garrapilos, a Mediterranean lowland forest of ca. 120 ha located in southern Spain (Cádiz province; 36°39.6′N, 5°56.9′W). Its vegetation consists of large holm (*Quercus ilex* subsp. *ballota*) and cork (*Q. suber*) oaks (10--12 m height), and an understorey dominated by treelets and shrubs (2--4 m height), among which wild olive trees (*Olea europaea* var. *sylvestris*), kermes oaks (*Q. coccifera*, Fagaceae), lentiscs, evergreen buckthorns (*Rhamnus alaternus*, Rhamnaceae) and hawthorns (*Crataegus monogyna*, Rosaceae) are the dominant species (Supporting Information [Figure S1](#mec14850-sup-0001){ref-type="supplementary-material"}). A lower layer of scrubs (\<1 m height) is dominated by rockroses (*Cistus salviifolius*, Cistaceae). The mean lentisc cover was 30%, and the mean cover of the main vegetation elements was as follows: oak trees 31%, shrubs 49%, scrubs 15% and uncovered soil (both with and without oak canopy above) 36% (cover data from 20 × 30‐m line transects); only uncovered soil, shrubs and scrubs account for 100% because the tree cover can overlap with these elements. Our sampling area covered ca. 20 ha within this forest. 2.3. Sampling design {#mec14850-sec-0005} -------------------- We studied different demographic processes associated with the quantity and quality components of seed dispersal effectiveness in three microhabitat types replicated in three periods of the fruiting phenology (early, mid and late), following a factorial study design (Figure [1](#mec14850-fig-0001){ref-type="fig"}b). We divided the 9‐month fruiting period (August--April) previously observed in the study site into three 3‐month periods classified as *early* (August--October), *mid* (November--January) and *late* (February--April) of the lentisc fruiting phenology. These same‐sized temporal frames allow the magnitude of seed dispersal to be compared between periods. We expected this period length (3 months) to properly capture local turnover in disperser species (see Section [2.5](#mec14850-sec-0007){ref-type="sec"} and Supporting Information [Table S1](#mec14850-sup-0001){ref-type="supplementary-material"}) and to include contrasting climatic conditions that are important for seedling recruitment (the mid‐period is generally colder and wetter than the early and late periods; Supporting Information [Figure S2](#mec14850-sup-0001){ref-type="supplementary-material"}). We evaluated the quantity of seed deposition as the contribution of different disperser species to seed rain across "microhabitat--period" combinations (Table [1](#mec14850-tbl-0001){ref-type="table"}). For the quality sub components, we assessed variation in seed viability only between periods, whereas we assessed post dispersal processes (survival to seed predation, germination and seedling survival) across "microhabitat--period" combinations (Table [1](#mec14850-tbl-0001){ref-type="table"}). We focused on three microhabitat types (Figure [1](#mec14850-fig-0001){ref-type="fig"}b): (a) on uncovered soil under the canopy of oak trees (*trees*, hereafter); (b) under treelets/shrubs bearing fleshy fruits (*fruit‐bearing shrubs*, hereafter); and (c) under shrubs not bearing fleshy fruits (*non‐fb shrubs*, hereafter); these microhabitats accounted for 53% cover in the study site (8%, 17% and 28%, respectively). We chose these microhabitats because birds typically use trees and shrubs as perches, dropping most seeds beneath them (Izhaki, Walton, & Safriel, [1991](#mec14850-bib-0027){ref-type="ref"}; Jordano & Schupp, [2000](#mec14850-bib-0033){ref-type="ref"}; Rey & Alcántara, [2014](#mec14850-bib-0048){ref-type="ref"}); in fact, we have found that lentisc seed rain densities on open ground are negligible (González‐Varo et al., [2014](#mec14850-bib-0017){ref-type="ref"}) and very low beneath *Cistus* scrubs (J.P. González‐Varo unpublished data). Besides, germination and establishment of lentisc seedlings are favoured beneath trees and shrubs due to favourable microclimatic conditions produced under their canopy (Verdú & García‐Fayos, [1996](#mec14850-bib-0063){ref-type="ref"}). We differentiated between types of shrubs because bird‐generated seed rain is generally higher beneath fruit‐bearing plants (Herrera, Jordano, López‐Soria, & Amat, [1994](#mec14850-bib-0025){ref-type="ref"}; Montesinos, Verdú, & García‐Fayos, [2007](#mec14850-bib-0037){ref-type="ref"}) and also because post dispersal processes such as seed predation can be both conspecific and heterospecific density‐dependent (García et al., [2007](#mec14850-bib-0012){ref-type="ref"}; Kwit, Levey, & Greenberg, [2004](#mec14850-bib-0034){ref-type="ref"}). The observational and experimental procedures to assess different demographic process are specified below. ###### Demographic processes assessed in this study belonging to the quantity or quality components of seed dispersal effectiveness (SDE). We assessed variation in these processes between disperser species (D), periods (P) within the fruiting phenology of the plant and microhabitats (M) of seed arrival and seedling recruitment SDE component Demographic process Metric Factors --------------- ---------------------------------------------------------- -------------------------------------- ------------- Quantity Seed deposition Seed rain density (seeds/m^2^) D*,* P*,* M Quality Seed viability Proportion of viable seeds P Quality Escape to seed predation Proportion of seeds that survive P*,* M Quality Germination Proportion of seeds that germinate P*,* M Quality Seedling survival[a](#mec14850-note-0001){ref-type="fn"} Proportion of seedlings that survive P*,* M Seedling survival until early autumn after the first summer. John Wiley & Sons, Ltd 2.4. Bird‐mediated seed dispersal {#mec14850-sec-0006} --------------------------------- We sampled lentisc seeds dispersed by birds in the study site during the whole 2014--2015 fruiting season, from summer 2014 to spring 2015 (August--April; 9 months in total). We used seed traps placed beneath the three target microhabitats (*trees*,*fruit‐bearing shrubs* and *non‐fb shrubs*) to quantify the magnitude of seed deposition. Seed traps consisted of plastic trays (40 cm × 55 cm, 8 cm height) with small holes (1 mm diameter) to allow the drainage of rainwater and covered with wire mesh (1 cm × 1 cm) to prevent post dispersal seed predation by vertebrates (Supporting Information [Figure S1](#mec14850-sup-0001){ref-type="supplementary-material"}). We monitored a total of 37 seed traps placed beneath different oak trees (*n *=* *12), treelets/shrubs bearing fleshy fruits (*n *=* *13; 5 wild olive trees, 4 female lentiscs and 4 hawthorns) and shrubs not bearing fleshy fruits (*n *=* *12; 4 kermes oaks, 4 male lentiscs and 4 male evergreen buckthorns); distance between seed traps ranged from 5 to 530 m. We conducted sampling surveys fortnightly where we recorded the number of bird‐dispersed lentisc seeds per seed trap and sampled individual seeds or droppings for DNA barcoding analysis (see Section [2.5](#mec14850-sec-0007){ref-type="sec"}). We did so putting each sample with a minimum of handling into a 1.5‐ml sterile tube with the aid of the tube cap. Tubes were labelled and stored in a freezer at −20°C until DNA extraction (González‐Varo et al., [2014](#mec14850-bib-0017){ref-type="ref"}). In each sampling survey, we generally sampled either all or most seeds from those seed traps receiving few seeds (1--4), whereas we generally sampled a subsample from seed traps receiving many seeds (\>5). Overall, we sampled 44% of all seeds found in the seed traps (457 out of 1,030 seeds). 2.5. Seed disperser identification through DNA barcoding {#mec14850-sec-0007} -------------------------------------------------------- We used DNA barcoding to identify the bird species that dispersed the seeds sampled (*n* = 457 seeds in 443 samples; 13 samples contained 2--3 seeds in the same bird dropping). DNA of animal origin can be extracted from the surface of defecated or regurgitated seeds, allowing the identification of the frugivore species responsible of each dispersal event (González‐Varo et al., [2014](#mec14850-bib-0017){ref-type="ref"}, [2017](#mec14850-bib-0018){ref-type="ref"}). Briefly, disperser species identification was based on a 464‐bp mitochondrial DNA region (COI: cytochrome *c* oxidase subunit I). For DNA extraction, we used a GuSCN/silica protocol, incubating each seed directly in extraction buffer (added to the 1.5‐ml tube where the seed was sampled in the field) (see details in González‐Varo et al., [2014](#mec14850-bib-0017){ref-type="ref"}). For PCR amplification, we used the primers COI‐fsdF (5′--GCATGAGCCGGAATAGTRGG--3′) and COI‐fsdR (5′--TGTGAKAGGGCAGGTGGTTT--3′) following the PCR protocol described by González‐Varo et al. ([2014](#mec14850-bib-0017){ref-type="ref"}). For a subset of sampled seeds (ca. 10%) that failed to amplify using COI‐fsd primer pair, we used an additional protocol using other primer sets to gain in amplification success for smaller DNA fragments. Details are provided in González‐Varo et al. ([2017](#mec14850-bib-0018){ref-type="ref"}). Briefly, this protocol consisted of nested PCRs, using a new primer set designed for shorter sequences (COI‐fsd‐degR: 5′--GTTGTTTATTCGGGGGAATG--3′, to be combined with COI‐fsdF; COI‐fsd‐degF: 5′--GGAGCCCCAGACATAGCAT--3′, to be combined with COI‐fsdR) (González‐Varo et al., [2017](#mec14850-bib-0018){ref-type="ref"}) on the amplicon AWCintF2--AWCintR4 (avian DNA barcodes; Lijtmaer, Kerr, Stoeckle, & Tubaro, [2012](#mec14850-bib-0035){ref-type="ref"}) as template (following Alcaide et al., [2009](#mec14850-bib-0002){ref-type="ref"}). We only sequenced one strand (forward primer) of the amplified COI fragments because in most cases the electrophoretic patterns were clear and resulting sequences (length: mean = 365 bp; median = 397 bp; range = 104--417 bp) allowed successful discrimination between species. Sequences (i.e., barcodes) were aligned and edited using [sequencher]{.smallcaps} 4.9, and then identified using the "[barcode of life data]{.smallcaps}" identification system ([bold]{.smallcaps}: <http://www.boldsystems.org>; Ratnasingham & Hebert, [2007](#mec14850-bib-0047){ref-type="ref"}). [bold]{.smallcaps} accepts sequences from the 5′ region of the COI gene and returns species‐level identification and assigns a percentage of similarity to matched sequences. We classified the DNA‐identified bird species as *residents*,*sub‐Saharan migrants* and *European migrants* in order to analyse whether the quantity and quality components of seed dispersal effectiveness are dependent on birds' migratory strategy. Sub‐Saharan migrants include species that either breed in the study area or use it as a stopover during their autumn migration to Africa. European migrants include species that overwinter in the study area. Classification was based on the online Encyclopaedia of the Birds of Spain (<http://www.seo.org/listado-aves>) and the species' occurrences in the study site, which were assessed through monthly bird censuses (Supporting Information [Table S1](#mec14850-sup-0001){ref-type="supplementary-material"}). 2.6. Pre dispersal loses: Seed viability {#mec14850-sec-0008} ---------------------------------------- We evaluated differences between periods in the viability (%) of bird‐dispersed seeds sampled in the field through the "flotation/sink" method: Only seeds that sink have a well‐developed embryo inside (validated by Albaladejo, González‐Martínez, Heuertz, Vendramin, & Aparicio, [2009](#mec14850-bib-0001){ref-type="ref"}). We conducted this test when the extraction buffer for DNA barcoding analysis was added to the tube containing the seed. We expected temporal differences in seed viability because (a) the ratio between red and black fruits varies through the fruiting season; (b) fruits can remain red in colour because colour is also associated with seed viability and most red fruits contain unviable seeds; and (c) although birds have a strong preference for black fruits, they also consume red ones (Herrera, [1984](#mec14850-bib-0022){ref-type="ref"}; Jordano, [1989](#mec14850-bib-0030){ref-type="ref"}). We also evaluated the viability of seeds (%) inside black fruits in order to obtain viable seeds for the sowing experiments (see Section [2.6](#mec14850-sec-0008){ref-type="sec"}). We collected black fruits from nine mother plants in the early period (October), 10 in the mid‐period (December) and 11 in the late period (February). We collected fruits from the same mother plants, whenever possible, to minimize maternal effects. However, some plants did not have black fruits in the late period, so we had to sample fruits of neighbour plants (shared mother plants between periods: early--mid: 9; early--late: 4; mid--late: 5). We collected a total of 2,288 black fruits (555--925 per period), for which we tested the viability of depulped seeds through the "flotation/sink" method. We opened a subset of seeds that floated (*n *=* *147) and corroborated that they were either empty (69%; due to abortion or parthenocarpy) or contained a wasp larva (31%) (Supporting Information [Figure S1](#mec14850-sup-0001){ref-type="supplementary-material"}). 2.7. Post dispersal fates: Seed predation, germination and seedling establishment {#mec14850-sec-0009} --------------------------------------------------------------------------------- We assessed three post dispersal processes (seed predation, germination and seedling establishment) through experiments conducted in the nine "microhabitat--period" combinations of our study design (Figure [1](#mec14850-fig-0001){ref-type="fig"}, Table [1](#mec14850-tbl-0001){ref-type="table"}). The experiments were set up on October 17, 2014, December 23, 2014, and February 27, 2015, for the periods *early*,*mid* and *late*, respectively (see Supporting Information [Figure S2](#mec14850-sup-0001){ref-type="supplementary-material"}). The experiments were conducted in an area of \~0.5 m^2^ under the canopy of individual trees or shrubs, which represented the microhabitat replicates. We assessed post dispersal seed predation by rodents in each study period by placing experimental units (seed depots, hereafter) across the three target microhabitats. Seed depots consisted of 10 lentisc seeds glued firmly to one side of a 10 cm × 10 cm of green plastic mesh (1‐mm pore size), which was nailed to the ground (see Rey et al., [2002](#mec14850-bib-0049){ref-type="ref"}). In each period, we placed 7--10 seed depots per microhabitat (i.e., 23--26 per period; *n* ~total~ = 75 depots with 750 seeds), which were monitored only after 2 weeks due to very high predation rates (see [Section 3](#mec14850-sec-0011){ref-type="sec"}). We considered a seed to have been preyed upon if it disappeared from the square or if it was still on the square but was gnawed and empty (Supporting Information [Figure S1](#mec14850-sup-0001){ref-type="supplementary-material"}). Seedling emergence and survival were assessed through a sowing experiment, which is known to be a robust tool for disentangling processes affecting seedling establishment (González‐Varo et al., [2012](#mec14850-bib-0019){ref-type="ref"}). In each period, we sowed seeds in the three target microhabitats to assess seed germination and seedling survival. All lentisc seeds used in this experiment were viable seeds collected from 9 to 11 mother plants per period (see details in Section [2.6](#mec14850-sec-0008){ref-type="sec"}). Seeds were sown in 7--10 replicated stations per microhabitat (i.e., 22--26 per period; *n* ~total~ = 71 stations with 639 seeds sowed). In each station, we removed any naturally dispersed seeds and then sowed nine seeds uniformly distributed in a matrix of 3 rows × 3 columns, separated by 5 cm, at a depth of 0.5--1 cm. We added litter to match the natural conditions as close as possible and protected the sowing stations with a wire mesh cage (with a grid of 15 × 15 cm area and 15 cm height; 1 cm × 1 cm) to prevent predation by rodents (e.g., González‐Varo et al., [2012](#mec14850-bib-0019){ref-type="ref"}). Additionally, we added thin wire mesh (1‐mm pore size) on the top of the cage to prevent the deposition of lentisc seeds into the sowing stations (Supporting Information [Figure S1](#mec14850-sup-0001){ref-type="supplementary-material"}). Seed germination and seedling survival were monitored fortnightly during the first \~4 months after sowing and monthly thereafter until October 28, 2015. We chose early autumn, after the first summer faced by the seedlings, as the end of the experiment because seedling survival to the summer drought is a crucial process affecting recruitment dynamics across Mediterranean plant species (Gómez‐Aparicio, [2008](#mec14850-bib-0015){ref-type="ref"}). 2.8. Data analyses {#mec14850-sec-0010} ------------------ All analyses were performed using the [r]{.smallcaps} computing environment [v]{.smallcaps}. 3.3.3 (R Core Team, [2015](#mec14850-bib-0046){ref-type="ref"}). We used different types of models (according to the data structure and the nature of the response variable) to test for significant effects of period, microhabitat and their interaction (P × M) on the demographic processes studied. The interaction term allowed us to test whether the effects of the period were consistent across microhabitats. We used a linear model (LM) to test for spatiotemporal differences in seed rain density (seeds/m^2^), which was ln (*x *+* *1) transformed to meet the normality and homoscedasticity assumptions. We used binomial distributions and logit link functions to analyse seed viability, seed predation, seed germination and seedling survival, all of which were Bernoulli‐distributed response variables (1 =  success, 0 = failure). For seed viability, we used a generalized linear model (GLM) to test for the effect of period. For seed predation, seed germination and seedling survival, we used generalized linear mixed models (GLMM) where period and microhabitat were included as fixed effects and the identity of the experimental stations (i.e., seed depots or sowing stations) was included as a random effect (Bolker et al., [2009](#mec14850-bib-0005){ref-type="ref"}). GLMMs were fitted using the package [lme4]{.smallcaps} (v. 1.1--12) (Bates, Maechler, Bolker, & Walker, [2014](#mec14850-bib-0004){ref-type="ref"}), and the significance of fixed effects (*p*‐values of Wald χ^2^ tests) was computed using the "Anova" function of the package [car]{.smallcaps} (v. 2.1--6) (Fox & Weisberg, [2011](#mec14850-bib-0010){ref-type="ref"}). For the analyses of seed dispersal effectiveness, we grouped the bird species contributing to seed rain (DNA‐identified) by their migratory strategy (i.e., residents, sub‐Saharan migrants and European migrants; see Section [2.5](#mec14850-sec-0007){ref-type="sec"}). We did so because: these groups are expected to contribute to seed dispersal in different periods of the fruiting phenology (Figure [1](#mec14850-fig-0001){ref-type="fig"}a); the sub‐Saharan migrants included many species but with small contributions to the lentisc seed rain (see [Section 3](#mec14850-sec-0011){ref-type="sec"}); all bird species within and across groups were relatively similar in terms of body size (i.e., small passerines of 12--70 g); and for the sake of simplicity because 3 bird groups × 3 periods × 3 microhabitats already account for 27 potential combinations of effectiveness. We calculated the quantity and quality components of seed dispersal effectiveness for different bird species groups (*i*) contributing to seed dispersal across the study periods (*j*) and microhabitats (*k*). We calculated the quantity component as follows: $$QT_{\mathit{ijk}} = d_{\mathit{jk}} \times f_{\mathit{ijk}}$$ where *d* ~*jk*~ is the magnitude of seed deposition (i.e., seed rain density) in period *j* and microhabitat *k*, and *f* ~*ijk*~ is the relative contribution (frequency) of bird species group *i* to period *j* and microhabitat *k*. In other words, the quantity of seed deposition contributed by each bird species group in each "microhabitat--period" combination. We calculated the quality component as follows: $$QL_{\mathit{jk}} = v_{j} \times p_{\mathit{jk}} \times g_{\mathit{jk}} \times s_{\mathit{jk}}$$where *v* ~*j*~ is the probability of viability among bird‐dispersed seeds in period *j*, whereas *p* ~*jk*~, *g* ~*jk*~ and *s* ~*jk*~ are, respectively, the probabilities of escaping to post dispersal seed predation, germinating and surviving as seedling for seeds dispersed in period *j* and microhabitat *k*. In other words, QL is the cumulative probability of recruitment of dispersed seeds in each "microhabitat--period" combination (see Table [1](#mec14850-tbl-0001){ref-type="table"}). We obtained zero probabilities in most *p* ~*jk*~ and in two *s* ~*jk*~, which likely reflected sample size limitations to accurately measure these demographic processes, rather than they were fully collapsed. For operational purposes, we replaced these zeros with low values in order to avoid zeros in the computed [ql]{.smallcaps} ~*jk*~ (e.g., González‐Varo et al., [2012](#mec14850-bib-0019){ref-type="ref"}; Rey & Alcántara, [2014](#mec14850-bib-0048){ref-type="ref"}). First, we assigned a constant probability of *p* ~*jk*~ = 0.01 (1%) because predation rates showed no variability and were almost total across periods and microhabitats. Second, we conservatively replaced the two zero values obtained for *s* ~*jk*~ with the minimum non zero value we obtained for the probability of seedling survival (*s* = 0.09) across "microhabitat--period" combinations. We calculated the seed dispersal effectiveness for each bird species group contributing to seed dispersal across periods and microhabitats as follows: $$\text{SDE}_{\mathit{ijk}} = QT_{\mathit{ijk}} \times QL_{\mathit{jk}}$$ We also calculated the overall effectiveness across periods (SDE~*ik*~ = [qt]{.smallcaps} ~*ik*~ × [ql]{.smallcaps} ~*k*~), where [qt]{.smallcaps} ~*ik*~ is the sum of [qt]{.smallcaps} ~*ijk*~ across periods for each "bird species group--microhabitat" combination, and [ql]{.smallcaps} ~*k*~ is the weighted mean of [ql]{.smallcaps} ~*jk*~ across periods (weighted by [qt]{.smallcaps} ~*ijk*~); the reason for weighting is that the cumulative probabilities of recruitment represented by [ql]{.smallcaps} ~*jk*~ were associated with different fractions of seeds arriving to each microhabitat. We used the package [effect.lndscp]{.smallcaps} (v. 0.2.8) (by P. Jordano; see Schupp et al., [2017](#mec14850-bib-0054){ref-type="ref"}) to represent "quantity × quality" effectiveness landscapes. 3. RESULTS {#mec14850-sec-0011} ========== 3.1. Seed rain density and frugivore contributions {#mec14850-sec-0012} -------------------------------------------------- We found that seed rain density mediated by birds varied significantly between periods (Table [2](#mec14850-tbl-0002){ref-type="table"}). Not surprisingly, the greatest seed densities were found in the mid‐period, with values that were \~3 times higher than in the early and late periods, when densities were very similar (Figure [2](#mec14850-fig-0002){ref-type="fig"}a). Seed rain density also varied significantly between microhabitats (Table [2](#mec14850-tbl-0002){ref-type="table"}), with the highest values found beneath both types of shrubs (with and without fruits) and the lowest beneath trees (\~2--3 times lower; Figure [2](#mec14850-fig-0002){ref-type="fig"}a). The non significant interaction term of the LM indicated that the differences observed between periods were consistent across microhabitats (Table [2](#mec14850-tbl-0002){ref-type="table"}). ###### Significance of the fixed factors of our sampling design ("period," "microhabitat type" and their interaction) on models analysing the demographic processes studied (*df*: degrees of freedom) Response Model Tests Period Microhabitat type P × M ------------------------------------------------------- ------- ----------- -------- ---------------------------- ------- --------------------------- ----- ------- Seed rain LM *F* 18.5 **1.4** × **10** ^**−7**^ 19.3 **8.0** × **10** ^**−8**^ 1.7 0.146 Seed viability[a](#mec14850-note-0004){ref-type="fn"} GLM Wald χ^2^ 43.5 **3.5** × **10** ^**−10**^ -- -- -- -- Seed predation[b](#mec14850-note-0004){ref-type="fn"} GLMM Wald χ^2^ 0.0 1.000 \<0.1 0.999 0.0 0.999 Germination GLMM Wald χ^2^ 20.8 **3.0** × **10** ^**−5**^ 2.2 0.327 4.9 0.300 Seedling survival GLMM Wald χ^2^ 2.6 0.457 7.3 *0.063* 0.8 0.936 *p*‐Values \< 0.05 are shown in bold; a *p*‐value \< 0.10 is shown in italics. LM: linear model; GLM: generalized linear model (binomial); GLMM: generalized linear mixed models (binomial with seed depot or sowing station as random factor). ^a^Differences in seed viability were only assessed between periods. ^b^Seed predation was 100% in 73 of the 75 seed depots placed in the field (see Figure [3](#mec14850-fig-0003){ref-type="fig"}b). John Wiley & Sons, Ltd ![(a) Frugivore‐mediated seed rain density (mean ± 95% CI) of *Pistacia lentiscus* in three microhabitat types for each of the three study periods of the 2014--2015 fruiting season. (b) Relative contribution (%) to seed rain densities by the 11 bird species identified through DNA barcoding applied to defecated/regurgitated seeds (*n* ~total~ = 435 seeds; mean = 48, range = 26--74 per "microhabitat--period" combination). Full bird species names: *Sylvia melanocephala* (Sm), *Luscinia megarhynchos* (Lm), *Ficedula hypoleuca* (Fh), *Muscicapa striata* (Ms), *Phoenicurus phoenicurus* (Pp), *Sylvia borin* (Sb), *Sylvia communis* (Sc), *Sylvia hortensis* (Sh), *Erithacus rubecula* (Er), *Sylvia atricapilla* (Sa) and *Turdus philomelos* (Tp). According to their migratory strategy, these species included resident birds (1 species), sub‐Saharan migrants (7 species) and European migrants (3 species) \[Colour figure can be viewed at <http://wileyonlinelibrary.com>\]](MEC-28-219-g002){#mec14850-fig-0002} We successfully identified through DNA barcoding a total of 11 bird species from 422 samples (435 seeds) of the total 443 samples analysed (457 seeds); that is, the disperser was successfully identified in 95.3% of samples (PCR amplification failed in the remaining 4.7%). These 11 species included one resident bird, seven sub‐Saharan migrants and three European migrants (Figure [2](#mec14850-fig-0002){ref-type="fig"}b). Remarkably, DNA barcoding tools allowed us to identify seed dispersal by two sub‐Saharan migrants (*Sylvia communis* and *Sylvia hortensis*) that were not recorded in the bird censuses (see Supporting Information [Table S1](#mec14850-sup-0001){ref-type="supplementary-material"}). The resident bird and the European migrants contributed to seed rain in all "microhabitat--period" combinations, whereas sub‐Saharan migrants only contributed in four combinations, and mostly in the early period (Figure [2](#mec14850-fig-0002){ref-type="fig"}b). Indeed, seed dispersal in the early period was mediated by a diverse avian assemblage (11 species), and seed rain contributions were evenly distributed between residents, and sub‐Saharan and European migrants (Figure [2](#mec14850-fig-0002){ref-type="fig"}b). In contrast, seed dispersal in periods mid and late was mediated by a simpler avian assemblage (4 species), and seed rain contributions were dominated by European migrants (83--96%; Figure [2](#mec14850-fig-0002){ref-type="fig"}b). The early period also showed the highest differences between microhabitats in seed rain contributions: Most seeds deposited beneath trees were dispersed by sub‐Saharan migrants (57%), whereas most deposited beneath non‐fruit‐bearing shrubs were dispersed by European migrants (58%). 3.2. Pre dispersal loses: Seed viability {#mec14850-sec-0013} ---------------------------------------- We found a striking decrease in the viability of dispersed seeds through the fruiting phenology (Figure [3](#mec14850-fig-0003){ref-type="fig"}a). Mean viability dropped from 63% in the early period to 35% in the mid‐period and then to 5% in the late period (Figure [3](#mec14850-fig-0003){ref-type="fig"}a). Accordingly, period has highly significant effects in the GLM (Table [2](#mec14850-tbl-0002){ref-type="table"}). We found a parallel decrease in seed viability for ripe fruits collected from lentisc plants (see Supporting Information [Figure S3](#mec14850-sup-0001){ref-type="supplementary-material"}), yet viability rates were higher (82%, 63% and 17% in periods early, mid and late, respectively). {#mec14850-fig-0003} 3.3. Post dispersal fates: Seed predation, germination and seedling establishment {#mec14850-sec-0014} --------------------------------------------------------------------------------- We found huge post dispersal seed predation rates, which were above 95% across all "microhabitat--period" combinations and complete (100%) in most combinations (Figure [3](#mec14850-fig-0003){ref-type="fig"}b); only 5 of the 750 seeds survived the experiment after 2 weeks. Some sowing stations were lost due to damage by wild boars (*Sus scrofa*): Nine stations (13%) were lost from the germination data and 18 stations (25%) from the survival data (see details in Supporting Information [Table S2](#mec14850-sup-0001){ref-type="supplementary-material"}). We recorded a total of 168 emerged seedlings in the 62 sowing stations (556 seeds) comprising the germination data. Seed germination varied significantly between periods but not between microhabitats, and the non significant interaction term of the GLMM indicated that the differences observed between periods were also consistent across microhabitats (Table [2](#mec14850-tbl-0002){ref-type="table"}). Germination rates were higher in periods early and late than in the mid‐period (Figure [3](#mec14850-fig-0003){ref-type="fig"}c). Such higher germination rates were associated with germination speeds, as these were significantly faster in periods early and late (mean = 8.1 and 6.9 weeks after sowing, respectively) than in the mid‐period (mean = 11.3 weeks; see details in Supporting Information [Figure S4](#mec14850-sup-0001){ref-type="supplementary-material"}). A total of 39 seedlings survived until the end of the experiment in the 53 sowing stations (138 seedlings) comprising the survival data. Seedling survival showed more erratic patterns, both across periods and microhabitats, although the highest survival was always found beneath fruit‐bearing shrubs (Figure [3](#mec14850-fig-0003){ref-type="fig"}d). Period and microhabitat had non significant effects on the GLMM (*p *\>* *0.4), but the interaction term was marginally significant (*p *=* *0.063; Table [2](#mec14850-tbl-0002){ref-type="table"}). 3.4. Seed dispersal effectiveness {#mec14850-sec-0015} --------------------------------- Seed dispersal effectiveness provided by different bird groups (residents, and sub‐Saharan and European migrants) changed remarkably in time due to temporal changes in the quantity and quality components (Figure [4](#mec14850-fig-0004){ref-type="fig"}a). The early period was a "low quantity -- high quality" period and included the highest quality values obtained across periods and microhabitats (Figure [4](#mec14850-fig-0004){ref-type="fig"}a). The disproportionate high quality found beneath fruit‐bearing shrubs resulted from the highest seed viability in the early period along with the highest germination and high seedling survival recorded beneath fruit‐bearing shrubs (Figure [3](#mec14850-fig-0003){ref-type="fig"}). The three bird groups contributed to seed deposition in this "microhabitat--period" combination (sub‐Saharan migrants \< residents \< European migrants). The mid‐period was a "high quantity -- intermediate quality" period and included the highest quantity values obtained across periods and microhabitats, mostly contributed by European migrants (Figure [4](#mec14850-fig-0004){ref-type="fig"}a). The intermediate quality found in the mid‐period across microhabitats resulted from intermediate levels of seed viability, low germination rates and high rates of seedling survival beneath shrubs (Figure [3](#mec14850-fig-0003){ref-type="fig"}). Finally, the late period was a "low quantity -- low quality" period (Figure [4](#mec14850-fig-0004){ref-type="fig"}a), the low quality mostly resulting from the minimal levels of seed viability in this period (Figure [3](#mec14850-fig-0003){ref-type="fig"}). These quantity--quality differences resulted in similar values of total seed dispersal effectiveness (sum across microhabitats and dispersers) in the periods early (SDE = 0.053) and mid (SDE = 0.057), which were one order of magnitude higher than in the late period (SDE = 0.005). ![Seed dispersal effectiveness (SDE = Quantity × Quality) landscapes of *Pistacia lentiscus*' seed dispersers grouped by their migratory strategy (residents, and sub‐Saharan and European migrants) and considering the three studied microhabitats (see legend). Isoclines represent all combinations of quantity and quality components with the same SDE. (a) SDE landscapes per period. (b) Overall SDE landscape \[Colour figure can be viewed at <http://wileyonlinelibrary.com>\]](MEC-28-219-g004){#mec14850-fig-0004} These period‐specific patterns shaped the overall seed dispersal effectiveness across periods (Figure [4](#mec14850-fig-0004){ref-type="fig"}b). The overall effectiveness was highest for European migrants dispersing seeds beneath fruit‐bearing and non‐fruit‐bearing shrubs (SDE = 0.049 and 0.027, respectively) followed by resident birds and sub‐Saharan migrants dispersing seeds beneath fruit‐bearing shrubs (SDE = 0.017 and 0.009, respectively) (Figure [4](#mec14850-fig-0004){ref-type="fig"}b). Accounting for temporal variation allowed us to unveil a non‐negligible contribution of resident birds and sub‐Saharan migrants to the overall SDE beneath fruit‐bearing shrubs (23.0% and 12.3%, respectively), despite their reduced contribution to the overall quantity component (12.6% and 4.2%, respectively) (Figure [4](#mec14850-fig-0004){ref-type="fig"}b). 4. DISCUSSION {#mec14850-sec-0016} ============= The effectiveness of seed dispersal mutualisms has been widely explored in previous studies aiming to assess not only the immediate, but also the delayed effects frugivorous animals have on the plant populations they disperse (Schupp, [1993](#mec14850-bib-0051){ref-type="ref"}; Schupp et al., [2010](#mec14850-bib-0053){ref-type="ref"}). These studies have usually focused on the spatial patterns of seed deposition and seedling recruitment (Calviño‐Cancela & Martín‐Herrero, [2009](#mec14850-bib-0006){ref-type="ref"}; Escribano‐Ávila et al., [2014](#mec14850-bib-0009){ref-type="ref"}; Jordano & Schupp, [2000](#mec14850-bib-0033){ref-type="ref"}; McConkey et al., [2014](#mec14850-bib-0036){ref-type="ref"}; Rother, Pizo, & Jordano, [2016](#mec14850-bib-0050){ref-type="ref"}), whereas the temporal patterns of seed dispersal effectiveness during the fruiting phenophase have remained completely overlooked. Here, we fill this knowledge gap by demonstrating shifts in the identity and contribution of seed dispersers, the magnitude of seed rain (quantity component) and multiple demographic processes (quality sub components) necessary for seedling recruitment (quality component). We show how small contributions to seed rain by migratory species can result in a relevant effectiveness if they disperse seeds during high‐quality periods for recruitment. These types of temporal shifts in effectiveness are to be expected in dynamic and seasonal environments (Carnicer, Jordano, & Melian, [2009](#mec14850-bib-0007){ref-type="ref"}), where plant--frugivore interactions are pivoting around frugivore assemblages with a marked component of migratory and transient species. 4.1. The timing of the quantity component {#mec14850-sec-0017} ----------------------------------------- The lentisc is a keystone species of Mediterranean woodlands, occupying central positions in plant--frugivore interaction networks (Olesen et al., [2011](#mec14850-bib-0041){ref-type="ref"}), and with a key role as a food item for both generalized frugivores (Jordano, [1988](#mec14850-bib-0029){ref-type="ref"}) and insectivores (Jordano, [1987](#mec14850-bib-0028){ref-type="ref"}). The contributions of different bird species to the lentisc seed rain were marked by the extreme seasonality and temporal dynamics of the local avifauna. The disperser assemblage actually includes bird species with three distinct migratory strategies (Moreau, [1952](#mec14850-bib-0038){ref-type="ref"}; Wernham et al., [2002](#mec14850-bib-0067){ref-type="ref"}) that overlay temporally along the lentisc fruiting phenophase, namely residents, sub‐Saharan long‐distance migrants and overwintering species from Northern Europe. Most sub‐Saharan migrants are transient in the study area between summer and early autumn: Only one of the seven species actually breeds in the study site (*Luscinia megarhynchos*; Supporting Information [Table S1](#mec14850-sup-0001){ref-type="supplementary-material"}), and the other six species only use the lowland forests and woodlands of South Spain as stopover sites for fuelling during their autumn migration (Herrera, [1984](#mec14850-bib-0022){ref-type="ref"}). This transience explains why the seed rain contribution of sub‐Saharan migrants, mostly contributed by *Ficedula hypoleuca*, was confined to the early period (Figure [2](#mec14850-fig-0002){ref-type="fig"}b). In contrast, both the resident species (*Sylvia melanocephala*) and the European migrants (*Erithacus rubecula*,*Sylvia atricapilla* and *Turdus philomelos*) occur in the study area during all or most of the lentisc fruiting phenophase (Supporting Information [Table S1](#mec14850-sup-0001){ref-type="supplementary-material"}). The dispersal peak observed in the mid‐period coincides not only with the ripening peak of the lentisc (Jordano, [1989](#mec14850-bib-0030){ref-type="ref"}), but also the massive arrival to the study region of the hyper abundant European migrants, especially *S*. *atricapilla* and *E*. *rubecula* (González‐Varo, [2010](#mec14850-bib-0016){ref-type="ref"}; Tellería, Ramírez, & Pérez‐Tris, [2008](#mec14850-bib-0057){ref-type="ref"}). European migrants generally stay in their Mediterranean wintering grounds from October to March (e.g., González‐Varo, [2010](#mec14850-bib-0016){ref-type="ref"}), which explains the similar seed rain contributions observed in the periods mid (November--January) and late (February--April), despite the lower seed rain densities in the latter. 4.2. The timing of the quality component {#mec14850-sec-0018} ---------------------------------------- We analysed four quality sub components of effectiveness and two of them, namely seed viability and germination, varied markedly between periods. The parallel decrease in viability of seeds from fruits and dispersed seeds indicates that such decrease was not caused by the birds, but took place at a pre dispersal stage (Supporting Information [Figure S3](#mec14850-sup-0001){ref-type="supplementary-material"}). As pointed, fruits can remain red in colour because colour is also associated with seed viability and most red fruits contain unviable seeds. This explains why lentisc plants typically bear more red than black fruits at the end of the fruiting season (Jordano, [1989](#mec14850-bib-0030){ref-type="ref"}). However, our results show that black fruits are only a reliable signal of seed viability at early and mid‐periods of the fruiting phenophase, because we only tested viability in black fruits and most seeds from the late period were unviable. This suggests that some fruits with unviable seeds might ripen very slowly up to, eventually, acquiring the black colour at the end of the season. As the fruit supply is depleted by the frugivores, the incidence of empty seeds becomes predominant, and such incidence was higher in bird‐dispersed seeds than in seeds sampled from black fruits (Supporting Information [Figure S3](#mec14850-sup-0001){ref-type="supplementary-material"}). The latter suggests that birds could also feed on red fruits despite their preference for the black ones (Jordano, [1989](#mec14850-bib-0030){ref-type="ref"}). Unviable seeds include abortion, parthenocarpy or parasitism by the chalcidoid wasp *M. pistaciae* (Jordano, [1989](#mec14850-bib-0030){ref-type="ref"}; see also Traveset, [1993](#mec14850-bib-0059){ref-type="ref"}) and, thus, the proximate causes underlying viability loss are biotic. We also expected to find temporal differences in post dispersal seed survival, a demographic process governed by the local abundance and foraging preferences of seed predators (García et al., [2007](#mec14850-bib-0012){ref-type="ref"}; Ostfeld, Manson, & Canham, [1997](#mec14850-bib-0043){ref-type="ref"}), that is, by biotic factors. Unfortunately, our sample sizes did not allow us to detect variation in post dispersal seed predation neither between periods nor between microhabitats. This was unforeseen because similar sample sizes successfully characterized differences in seed predation among populations and between microhabitats in other Mediterranean shrub species (González‐Varo et al., [2012](#mec14850-bib-0019){ref-type="ref"}). Seed germination was the other quality sub component that varied between periods and that variation was higher than that observed between microhabitats. Germination was both lower and slower in the mid‐period, that is, in the seed dispersal peak. Apparently, the main factors underlying temporal differences in germination were abiotic, and there are reasons to think that temperature played a crucial role. The speed and success of germination in lentisc seeds are positively associated with soil humidity (Verdú & García‐Fayos, [1996](#mec14850-bib-0063){ref-type="ref"}), but this does not explain the lower and slower germination rates observed in the mid‐period because soil humidity during the first month after sowing was very similar between periods (see details in Supporting Information [Figure S5](#mec14850-sup-0001){ref-type="supplementary-material"}). In contrast, seeds sowed in the mid‐period faced lower temperatures than those sowed in the periods early and late (average air temperature can be 5--8°C lower; see Supporting Information [Figure S2](#mec14850-sup-0001){ref-type="supplementary-material"}). We also expected to find temporal differences in seedling survival between periods because this process can be a demographic bottleneck in strongly seasonal Mediterranean ecosystems due to summer drought (Gómez‐Aparicio, [2008](#mec14850-bib-0015){ref-type="ref"}). Yet, we think our limited sample sizes also prevented clearer patterns; we sowed more than 639 seeds, but survival was only assessed in 138 seedlings across nine "microhabitat--period" combinations. Interestingly, the greatest effect in the GLMM analysing seedling survival was accounted by the interaction term, slightly supporting the idea that the quality of microhabitats can vary between periods (see Figure [3](#mec14850-fig-0003){ref-type="fig"}d). 4.3. The timing of seed dispersal effectiveness {#mec14850-sec-0019} ----------------------------------------------- The overlay of the temporally dynamic frugivore assemblage, and the temporally variable seed viability and germination caused heterogeneous effectiveness of lentisc seed dispersers. Intensity of seed rain (maximum in the mid‐period) is temporally decoupled from good conditions for germination (early and late in the season) and seed viability (maximum in the early season). The overall seed dispersal effectiveness was dominated by hyper abundant European migrants, not only owing to their huge quantity contribution throughout the lentisc fruiting phenophase, but also because their quantity contribution was higher than that of residents and sub‐Saharan migrants in the "microhabitat--period" combination with the highest quality for seedling recruitment (i.e., *fruit‐bearing shrubs* -- *early*; Figure [4](#mec14850-fig-0004){ref-type="fig"}). However, most seed dispersal by sub‐Saharan migrants was uniquely concentrated in that "microhabitat--period" combination, making their mutualistic services to be the ones with the highest overall quality. This evidences that the timing of dispersal can also compensate for quantity inequalities in seasonally dynamic disperser assemblages within seasonal ecosystems. In particular, we would overlook the relevance of sub‐Saharan migrants if we simply consider the overall frugivore contribution to seed rain over the whole fruiting season, or if we assess the quality sub components in the peak or at the end of the fruiting phenophase (e.g., Escribano‐Ávila et al., [2014](#mec14850-bib-0009){ref-type="ref"}; García, [2001](#mec14850-bib-0011){ref-type="ref"}; González‐Varo et al., [2012](#mec14850-bib-0019){ref-type="ref"}). Interestingly, the effects of timing on the quantity and quality components of effectiveness are expected to vary at wider spatiotemporal scales, between years and among populations. Simply considering the disperser assemblage studied, one would expect that fruit crops should be depleted earlier in years of low fruit production or advanced fruiting phenology. In fact, the lentisc ripening peaks can differ in nearly 1 month between consecutive years (Jordano, [1989](#mec14850-bib-0030){ref-type="ref"}), and the local abundance of lentisc fruits in the study site was nearly 10 times greater in the study season than in the previous season (i.e., 2013--2014: J.P. González‐Varo unpublished data). Similarly, fruit crops should be depleted earlier in less dense lentisc populations (see González‐Varo, [2010](#mec14850-bib-0016){ref-type="ref"}). Under such scenarios, the relative contributions of Sub‐Saharan migrants could be greater than the ones reported here. 5. CONCLUDING REMARKS {#mec14850-sec-0020} ===================== Plant--animal mutualisms are intrinsically dynamic forms of ecological interactions. We show here that the timing of plant--frugivore interactions matters for the quantity and quality components of seed dispersal effectiveness. Plants offer a resource provisioning for mutualistic animals, and the seasonal dynamics of animal assemblages along the flowering or fruiting phenophases typically result in a high temporal turnover of interactions (Morente‐López, Lara‐Romero, Ornosa, & Iriondo, [2018](#mec14850-bib-0039){ref-type="ref"}). Timing effects on effectiveness are therefore expected to happen in other types of mutualisms like, for instance, pollination. In fact, during the flowering period of a plant, flowers can be exposed to different pollinator faunas differing in the quantity of pollen they can transport (Ivey et al., [2003](#mec14850-bib-0026){ref-type="ref"}; Valverde et al., [2016](#mec14850-bib-0061){ref-type="ref"}), while pollen germinability (a quality sub component) may depend on local climatic conditions (Aronne, Buonanno, & De Micco, [2015](#mec14850-bib-0003){ref-type="ref"}). We think the "quantity--quality" compensatory effects uncovered here for transient sub‐Saharan migrants are likely to occur in highly seasonal environments, where biotic and abiotic conditions change considerably during the whole flowering or fruiting phenophases. We hope our study will foster future research on timing effects on effectiveness of mutualistic interactions and their relevance for ecological functionality and community dynamics. AUTHOR CONTRIBUTIONS {#mec14850-sec-0023} ==================== J.P.G.‐V. conceived the study, planned the sampling design and collected the data in the field. J.M.A. performed laboratory work. J.P.G.‐V. conducted the statistical analyses and wrote the first manuscript draft. P.J. discussed the idea, the sampling design and the results, and contributed during manuscript writing. All authors approved the final manuscript. Supporting information ====================== ######   ###### Click here for additional data file. We thank the "Servicio de Cría Caballar de las Fuerzas Armadas" for permission to work at the study site. We thank Bea Rumeu, Carolina Carvalho, Néstor Pérez‐Méndez, Cande Rodríguez, Pablo González‐Moreno and David Varo for their help and pleasant company during fieldwork. Benno Simmons kindly checked the English grammar and style. Logistical support was provided by the Molecular Ecology Laboratory, Estación Biológica de Doñana (LEM‐EBD‐CSIC), a facility certified to ISO9001:2015 and ISO14001:2015 quality and environmental management protocols. This study was funded by grants (to P.J.) of the Spanish MINECO (CGL2017‐82847‐P) and Junta de Andalucía Excellence Projects (RNM‐5731), and supported by a Severo Ochoa Award for Centres of Excellence in R+D+I (SEV‐2012‐0262). J.P.G.‐V. was funded by a postdoctoral fellowship from the Severo Ochoa Program (SEV‐2012‐0262) and, while writing this paper, by an Individual Fellowship from the Marie Sklodowska‐Curie Actions (H2020‐MSCA‐IF‐2014‐656572: MobileLinks). DATA ACCESSIBILITY {#mec14850-sec-0022} ================== Data associated with this article (a: seed rain density; b: sample‐level information with seed dispersers identified through DNA barcoding; c: seed viability in ripe fruits; d: seed predation experiment; e: sowing experiment; and f: seed dispersal effectiveness) are deposited in Dryad (<https://doi.org/10.5061/dryad.3c4n8nc>).

This is the story of the work of women\'s religious orders in setting up a system of health care in New York City in the mid-nineteenth century, and running it successfully for over a hundred years. It is not for the uninitiated in the history of the city, or even those coming for the first time to the worlds of health and women\'s history. Starting in the 1840s, by the beginning of the twentieth century, women\'s Catholic religious orders ran fourteen of New York\'s non-public hospitals, seven general care institutions, and specialized services for infants and children, women, tuberculosis patients, the aged and the dying. Bed capacity accounted for one quarter of the total supply in the city by 1904. The first Roman Catholic hospital in New York was founded in 1849 (sixteen years after the first such hospital in the United States), in part in response to increased immigration of Roman Catholics, and a perceived prejudice against them, and visiting priests, in the established hospitals of Bellevue and New York Hospital. Unlike the majority of specialist hospitals in Britain, St Vincent\'s (and its thirteen successors in the city) did not spring from the vision of medical men. The Roman Catholic Hospitals of New York City were the products of the vocation of nursing sisterhoods to care for the sick of this rapidly-expanding metropolis. As such, their history forms part of the growing body of work on women\'s pivotal role in initiating and developing health care in the United States. Within ten years of the first hospital\'s foundation, the patient population of New York was "overwhelmingly foreign-born". By 1866, 50 per cent of hospital admissions gave Ireland as their birthplace, and were presumed to be Roman Catholic. It is not clear from this work what percentage of the inhabitants of New York (old and new) were members of the Church, so no conclusion can be drawn about the health profile of the notoriously poor Irish of the growing city, or of that of the German and Italian immigrants who formed the patient population of several of the new hospitals. The timing of the hospital initiatives was no accident. Roman Catholic nursing sisterhoods had begun to be accepted by the establishment during the Civil War, when the Sisters of Mercy had nursed the wounded of both sides, in spite of opposition from the Church hierarchy and the formidable Dorothea Dix, superintendent of women nurses in the Union Army. It would have been intriguing to discover the *antebellum* attitude of New Yorkers to the sisters, but context for this (and much more) is missing from this slender volume. Bernadette McCauley refutes the assertion by the contemporary Catholic press that the sisters were resuming a European pre-Reformation tradition of women religious caring for the sick, but were rather in the seventeenth-century model of "active communities". She points out that most of the orders which established hospitals in the city were relatively young, and that the Sisters of Charity (the order that established St Vincent\'s) had been founded in the United States in the early nineteenth century. Who were these women? With a few exceptions, the reader cannot say. We are told early on (and it is reiterated several times) that the first administrator of St Vincent\'s, Ellen Hughes, was the sister of the Roman Catholic Archbishop of New York, but she is one of the few identifiable women in the hospital movement. This may be the natural result of studying groups of women whose life choice was a binding commitment to remove exterior traces of individual personality through their titles, behaviour and dress, but it does not help in understanding the specific impetus to begin---and maintain for over a century---such a significant part of health care in one of the largest, and most culturally diverse, cities of the New World. There are some half-hearted attempts to assess their ethnic, class and educational backgrounds, but with little statistical evidence presented these do not enlighten. The sisters were clearly women of great resourcefulness, as well as piety. All but one of the orders who embarked on the mission were immigrants themselves, and received little support from their mother houses. Once they had decided to open their own hospitals, they raised the seed money by the more traditional means of establishing fee-paying schools. New buildings were impossible at first, so they converted old buildings in the geographical area in which they felt they were most needed. With little or no municipal financial support, they generated funds from within the constituencies they served. The sisters of St Dominic, which ran St Catherine\'s Hospital in Brooklyn, was an enclosed order. In order to undertake their mission, they extended the boundaries of the cloister to include the hospital. The Sisters of Charity were forbidden from treating boys, and therefore separated from the mother house in Maryland in order to respond to the Archbishop\'s plea to take over the running of the Roman Catholic Orphan Asylum. These were ingenious solutions to potential barriers to their mission. Contemporary accounts praised the sisters for their selfless devotion, and this quality, allied to their vows of poverty ("we will live with the poor and like the poor"), was the principal selling point for the hospitals when they were founded, and for much of their existence under the sisters\' direction. The daily discipline of convent life was considered by some nurse leaders to mitigate against their being truly devoted nurses, but it was recognized that they offered excellent, reliable, service at minimum cost, and with none of the disciplinary problems that lay nurses could bring. The sisters were barred from studying medicine until the 1930s, and posed no threat to the male medical establishment. They asked little of the archdiocese, and claimed no miracle cures, the treatments on offer being thoroughly orthodox. They responded to developments, setting up nurse training schools in the early twentieth century, and erecting purpose-built hospitals for the demands of scientific medicine. The author is more comfortable with the financial and administrative history of her selected institutions, although, without supplementary information, it is hard to digest the long list of donors and significant individual figures in the various hospitals. Tables or graphs would have made the financial details easier to comprehend, and a table giving the names, founding dates and religious affiliation of each institution would have helped in distinguishing those under consideration. The title of the book, the first part of which is a quote from the *Catholic World* in 1868, implies that its focus is the Roman Catholic sisters who nursed New York Roman Catholics. This may have been the intention of the author, but the target is missed. The women themselves are absent, and so are their patients. One might assume that they nursed only Roman Catholics, but this seems not to have been the case. There is a one-page overview of patient diseases, but too much is either left unsaid, or merely hinted at. How did the hospitals get their patients? Most of the patients paid something towards their care, but there is scant consideration of the economics of sickness, or the class structure of patient admissions. There is a throwaway comment on page 42 that hospital patients were rather like paupers, in that accepting institutional care was a shameful admission of failure to provide in times of sickness, but that there was "prestige" attached to being nursed by the sisters. This begs many questions, none of which are answered. Hospital rules (long the bugbear of patients and their families in the nineteenth and early twentieth century) are said to have been more acceptable in their establishments, as they were neither more nor less than those by which the nuns lived, but the evidence is missing. One hospital was close to the docks and therefore was effectively an accident and emergency facility, but we do not know the outcomes of treatment, nor the relationship between the institutions and the employers and unions. At one point, we are told that St Vincent\'s hospital had an enormous number of patients suffering from alcoholism, but the fact is left hanging, and one longs to know more. What is one to make of the following, "The patient regulations at Seton Hospital, a tuberculosis hospital run by the Sisters of Charity where the patients were almost entirely charity cases, illustrate how the sisters attempted to maintain what they considered propriety, and demonstrate that class distinctions among patients and staff were not absent from Catholic institutions" (pp. 46--7). There is no account of the rules, no consideration of what was and was not propriety, and nothing on the class structure of the hospital, let alone the society it served. Several important points are highlighted in the work. The first is that the sisters did not view hospital treatment as an end in itself, but as just one part of a mosaic of care for the bodies and souls of the disadvantaged in this city of immigrants. Death was part of this picture, and was not viewed as failure, but as the path to a higher life. In a world where fund-raisers competed on the basis of the statistics of success, this attitude must have been either refreshing, or contrary. The author does succeed in upsetting preconceived notions of what being a religious sister was in New York in this period. She presents an account of innovation, adaptability, patience, skill in care-giving and financial administration---allied to a life choice that rejected materialism and self-advancement. As she concludes rather inelegantly, by the late twentieth century, "New York\'s hospital sisters had accomplished quite a bit". We are left with the impression that this little book (just ninety-six pages when the long introduction, acknowledgements, footnotes and excellent bibliography are removed) is part of a much longer study. While there are flashes of great insight, and it is clearly the result of much diligent research in an impressive array of sources, it is also evident that the author has done a hatchet job on her original manuscript. It is a little like sitting down to a meal, and being served with just a morsel from each course. It is to be hoped that her next volume will provide the banquet for which this book is merely a taster.

INTRODUCTION {#sec1-1} ============ Prostate cancer (PCa) is the second most common cancer diagnosed in men and the fifth leading cause of male cancer deaths worldwide.[@ref1] Despite the availability of various treatment strategies, including surgery, radiotherapy, chemotherapy, and endocrine and targeted therapies, prognosis remains very poor especially for patients with castration-resistant PCa (CRPC).[@ref2][@ref3] Current diagnostic and prognostic indicators, such as prostate-specific antigen (PSA) and Gleason score, have highly variable, which can lead to failed diagnosis and prognosis.[@ref4][@ref5] Additionally, the pathogenesis of PCa development and metastasis is not completely understood. Hence, there is an urgent need for better understanding of the molecular mechanisms underlying PCa carcinogenesis and progression to facilitate identification of novel molecular targets for diagnosis and therapy. Phosphoglycerate mutase 1 (PGAM1) is an important enzyme in glycolysis, which catalyzes the conversion of 3-phosphoglycerate (3-PG) to 2-phosphoglycerate (2-PG).[@ref6] In a study that used chemistry-based proteome reactivity profiling to identify drug targets in breast cancer, PGAM1 was identified as a potential metabolic enzyme related to breast carcinogenesis.[@ref7] Several subsequent studies have demonstrated that PGAM1 is usually upregulated in a range of human cancers, such as hepatocellular carcinoma, lung cancer, breast cancer, colorectal cancer, and renal clear cell carcinoma, and that its enzymatic activity was increased in cancerous tissues compared with adjacent normal tissues.[@ref8][@ref9][@ref10][@ref11] Moreover, knocking down PGAM1 attenuated lung and breast cancer cell growth in nude mouse xenograft models.[@ref10][@ref12] However, the relationship between PGAM1 and PCa has not been extensively investigated. In this study, we investigated changes in PGAM1 expression in PCa tissues compared with normal prostate tissues, and examined the relationship of PGAM1 with clinicopathological variables and its cellular function in relation to tumorigenesis, to evaluate its potential value as a biomarker and for use in targeted therapy for PCa. MATERIALS AND METHODS {#sec1-2} ===================== {#sec2-1} ### Tissue microarray and cell culture {#sec3-1} A human prostate tissue microarray (TMA; PR1921) was purchased from US Biomax, Inc. (Rockville, MD, USA). The TMA contained both normal prostate tissues and PCa tissues along with each patient\'s age, clinical stage, Gleason score, and metastasis status. Four human PCa cell lines, PC-3, 22Rv1, DU145, and LNCap (ATCC, Manassas, VA, USA), and a human prostate epithelial cell line, RWPE1 (ATCC), were used in this study. The cell lines were identified by short tandem repeat genotype analysis and cultured in RPMI-1640 (Gibco, Grand Island, NY, USA) medium with 10% bovine serum at 37°C in an atmosphere containing 5% CO~2~. ### Immunohistochemistry {#sec3-2} Immunohistochemistry was performed on the TMA according to the manufacturer\'s recommended protocols.[@ref13][@ref14] Briefly, the TMA slides were deparaffinized and rehydrated, and then endogenous tissue peroxides were quenched by incubation with 0.3% H~2~O~2~ for 30 min. For antigen retrieval, the slides were boiled in sodium citrate buffer (10 mmol l^−1^, pH 6.0) in a pressure cooker for 7 min. Subsequently, nonspecific binding was blocked with 5% normal goat serum, and then the slides were incubated with primary rabbit anti-human PGAM1 polyclonal antibody (1:200, Abcam Inc., Cambridge, MA, USA) overnight at 4°C, then with anti-rabbit secondary antibody (Zhongshan Biotech, Zhongshan, China). Diaminobenzidine was visualized as a chromogen substrate. The slides were counterstained with hematoxylin, then dehydrated and mounted with glass coverslips according to standard laboratory protocol. The positive staining intensity of PGAM1 was scored into four categories: 0, negative; 1, weakly positive; 2, intermediately positive; and 3, strongly positive. The percentage of PGAM1 positive cells was scored as four categories: 0, no staining; 1: \<25% cells; 2: 25%--75% cells; and 3: \>75% cells. The total protein expression score (ranging from 0 to 9) of a sample was obtained by the multiplication of the intensity and percentage scores, as previously described.[@ref15] The staining pattern of TMA was scored based on the total protein expression scores as follows: total protein expression score 0, −; 1--3, +; 4--6, ++; and 6--9, +++. We subsequently divided our cases into two groups using total protein expression scores: cases with total expression score 0--3 (−/+) were assigned to the low expression group, and cases with total expression score 4--9 (++/+++) were assigned to the high expression group. ### Western blotting {#sec3-3} Total protein was extracted from cell lysates using radioimmunoprecipitation assay buffer (Beyotime Biotechnology, Shanghai, China). Equal amounts of total protein samples were separated by SDS-polyacrylamide gel electrophoresis and electrotransferred from the gel to polyvinylidene fluoride membranes (Millipore Corporation, Billerica, MA, USA). The membranes were blocked with 5% fat-free milk or bovine serum albumin, and then immunoblotted using the following primary antibodies: rabbit anti-PGAM1 (1:1000; Abcam), rabbit polyclonal anti-cleaved caspase-3 (1:500, \#9664; Cell Signaling Technology, Danvers, MA, USA); and rabbit polyclonal anti-Bcl-2, anti-Bax, anti-matrix metallopeptidase (MMP)-2, and anti-MMP-9 (all 1:500; all Immunoway, Plano, TX, USA). Anti-β-actin staining (1:1000; Bioworld Technology, Louis Park, MN, USA) was used as an internal control. Finally, the membranes were incubated with the appropriate secondary antibodies (1:5000; Boster Ltd., Wuhan, China). Signals were visualized using an enhanced chemiluminescence detection system (Pierce Biotechnology, Rockford, IL, USA) in accordance with the manufacturer\'s instructions. ### Short interfering (si) RNA transfection {#sec3-4} Two siRNAs were designed for PGAM1 knockdown. The sequences were as follows: 5′-GUCCUGUCCAAGUGUAUCUTT-3′ and 5′-AGAUACA CUUGGACAGGACTT-3′. The sequences of the negative control (NC) siRNA were as follows: 5′-UUCUUCGAACGUGUCACGUTT-3′ and 5′-ACGUGACACGUUCGGAGAATT-3′. The 22Rv1 and PC-3 cells were seeded (3 × 10^5^ cells per well) in six-well plates (Corning Costar, Corning, NY, USA). When the cells reached 70% confluence, they were transfected with siRNA using Lipofectamine 3000 reagent (Life Technologies, Carlsbad, CA, USA) according to the manufacturer\'s protocol. ### Cell proliferation, migration, and invasion assays {#sec3-5} The proliferation of transfected cells was evaluated by a CCK-8 assay (Kit Dojindo, Kumamoto, Japan) according to the manufacturer\'s protocol. Briefly, cells were seeded (3 × 10^3^ cells per well) in 96-well plates and cultured for 24, 48, 72, or 96 h. CCK-8 reagent (10 μl) was added to each well and the cells were incubated for 2 h, and then the absorbance at 450 nm was measured with a SpectraMax M5 microplate reader (Molecular Devices, Sunnyvale, CA, USA). The migration and invasion abilities of siRNA-transfected cells were evaluated using a Transwell assay (Corning Costar, Corning, NY, USA). Briefly, 3 × 10^4^ cells resuspended in 2000 μl of serum-free medium were added to the upper chamber of a Transwell system with an 8-μm pore membrane. The chamber was uncoated (for the migration assay) or coated with Matrigel (BD Biocoat, Bedford, Mass, USA; for the invasion assay). The lower chamber contained 300 μl medium containing 10% fetal bovine serum. Cells were allowed to migrate for 24 h or invade for 48 h, and then the cells that had not penetrated the membrane were removed with a cotton swab. The cells on the lower surface of the membrane were fixed, stained, and counted under a light microscope in five randomly selected fields. ### Flow cytometry analysis {#sec3-6} Twenty-four hours after transfection, the cells were collected and washed twice with cold phosphate-buffered saline. Cell apoptosis was evaluated using an annexin V FITC apoptosis detection kit I (BD biosciences, Franklin Lakes, NJ, USA) according to the manufacturer\'s instructions. Apoptotic cells were detected by flow cytometry using a BD FACSVerse system. ### Tumor xenograft model in nude mice and shRNA treatment {#sec3-7} Lentivirus-mediated PGAM1 knockdown in PC-3 cells was achieved using a lentivirus kit according to the manufacturer\'s instruction (GeneCopoeia, Carlsbad, CA, USA). Briefly, the cells were infected with a lentivirus bearing short hairpin (sh) RNA targeting PGAM1[@ref12] (5′-CCGGCAAGAACTTGA AGCCTATCAACTCGAGTT GATAGCTTCAAGTTCTTGTTTTTTG-3′) and a recombinant hnRNP-L lentivirus. The NC groups were infected with the empty lentiviral vector. The infection efficiency was validated by Western blotting analyses. Female athymic mice (BALB/c-nu/nu; 4--5 weeks old) were purchased from the Animal Center of Southern Medical University and were housed in specific-pathogen-free conditions and bred in accordance with the institutional guidelines. To evaluate PCa tumor growth *in vivo*, 5 × 10^6^ PC-3 cells stably expressing PGAM1 shRNA via lentiviral infection or NC cells were injected subcutaneously and bilaterally into the flanks of athymic mice (6 mice per group). Tumor dimensions were measured on two perpendicular axes and tumor volume was calculated with the formula: volume = (length × width^2^)/2. The mice were euthanized by CO~2~ inhalation after 25 days and the tumors were removed and weighed, then the primary tumors were fixed, paraffin-embedded, and sectioned. The sections were stained with hematoxylin and eosin and observed under a microscope. All of the procedures were approved by the Institutional Animal Care and Use Committee of Southern Medical University. ### Statistical analysis {#sec3-8} Statistical analyses were performed using SPSS 20.0 software (SPSS Inc., Chicago, IL, USA). Data are expressed as mean ± s.e.m. The Student\'s *t*-test was used to analyze continuous data, the Chi-square test was used for categorical data, and factorial analysis of variance was used to analyze differences between groups. *P* \< 0.05 was considered statistically significant. RESULTS {#sec1-3} ======= {#sec2-2} ### PGAM1 expression levels in PCa tissues and cells {#sec3-9} To detect PGAM1 expression in human PCa, immunohistochemical staining was performed on a total of 96 prostate tissue specimens processed in a TMA, including 16 and 80 PCa tissues. Based on PGAM1 staining levels (**Figure [1a](#F1){ref-type="fig"}**-**[1d](#F1){ref-type="fig"}**), all prostate tissues were divided into two groups: a low expression group (− and +) and a high expression group (++/+++). Among the PCa tissues, 56 cases (70%) were classified as having high expression of PGAM, compared with only 4 cases (25%) among the nonneoplastic (normal or adjacent) tissues. Chi-square analysis revealed that PGAM1 expression levels were higher in PCa tissues than in nonneoplastic (normal or adjacent) tissues (*P* = 0.001, **[Table 1](#T1){ref-type="table"}**). As shown in **[Figure 1e](#F1){ref-type="fig"}**, PGAM1 localized to the cytoplasm and nucleus in prostate cells. To validate the TMA data, Western blotting was carried out in four PCa cell lines (PC-3, 22Rv1, DU145, and LNCap) and a normal prostate epithelial cell (RWPE1). The results showed that PGAM1 protein expression levels were markedly higher in four PCa cell lines than in the RWPE1 cell line (**[Figure 1f](#F1){ref-type="fig"}**). {#F1} ###### Correlation between phosphoglycerate mutase 1 expression and clinicopathological variables of prostate cancer  We then analyzed the relationship between PGAM1 expression levels and clinicopathological variables. Chi-square analysis showed that PGAM1 expression was not associated with patient age, clinical stage, or lymph node or distant metastasis status, but was statistically associated with Gleason score (*P* = 0.01) and T-stage (*P* = 0.009) (**[Table 1](#T1){ref-type="table"}**). ### Inhibition of cell proliferation by PGAM1 knockdown {#sec3-10} To determine the biological function of PGAM1 in PCa, siRNA targeting PGMA1 (si-PGAM1) was transfected into PC-3 and 22Rv1 cells to inhibit endogenous PGAM1 expression. Western blotting confirmed that the PGAM1 protein level markedly decreased in PC-3 and 22Rv1 cells transfected with si-PGAM1 compared with NC cells (**[Figure 2a](#F2){ref-type="fig"}**). To determine the influence of PGAM1 knockdown on the proliferation ability of PCa cells *in vitro*, a CCK-8 assay was performed. Factorial analysis of variance demonstrated that PGAM1 knockdown inhibited proliferation of PC-3 and 22Rv1 cells at 72 h and 96 h, respectively, compared with NC cancer cells (*P* \< 0.001, **[Figure 2b](#F2){ref-type="fig"}**). {#F2} ### PGAM1 knockdown enhanced apoptosis in PCa cells {#sec3-11} Next, we used flow cytometry to determine whether the PGAM1 knockdown-induced inhibition of cell proliferation resulted from apoptosis. Twenty-four hours after transfection of PC-3 and 22Rv1 cells with siRNA, the number of both early and later apoptotic cells was markedly increased among cells transfected with si-PGAM1 compared with NC cells (**[Figure 3a](#F3){ref-type="fig"}**). Student\'s *t*-test analysis revealed that the mean total number of apoptotic cells increased from 6.68 ± 0.64 to 20.50 ± 0.94 and from 4.77 ± 0.58 to 16.93 ± 1.55 in response to PGAM1 knockdown in PC-3 cells (*P* \< 0.001) and 22Rv1 cells (*P* \< 0.01), respectively (**[Figure 3b](#F3){ref-type="fig"}**). {#F3} To further analyze the potential mechanism underlying this promotion apoptosis by PGAM1 knockdown, we examined the expression of apoptosis-related proteins (Bcl-2, Bax, and cleaved caspase-3) by Western blotting. As shown in **[Figure 3c](#F3){ref-type="fig"}**, the protein expression levels of Bax and cleaved caspase-3 significantly increased in response to PGAM1 knockdown compared with NC and blank PC-3 and 22Rv1 cells. Conversely, the expression of Bcl-2 decreased in si-PGAM1-transfected PCa cells. ### PGAM1 knockdown inhibited PCa cells migration and invasion {#sec3-12} We used Transwell chamber and Matrigel assays to investigate the effect of PGAM1 on PCa cell migration and invasion. The migration and invasion abilities of si-PGAM1-transfected PC-3 and 22Rv1 cells were prominently decreased compared with those of NC cells (**[Figure 4a](#F4){ref-type="fig"}**). This was confirmed by Student\'s *t*-test analysis (*P* \< 0.01, **[Figure 4b](#F4){ref-type="fig"}**). {#F4} Because MMP-2 and MMP-9 are crucial for tumor cell migration and invasion, we examined their expression in PCa cells with PGAM1 knocked down. As expected, PGAM1 silencing downregulated MMP-2 and MMP-9 protein levels in PCa cells (**[Figure 4c](#F4){ref-type="fig"}**). ### Knockdown of PGAM1 inhibited xenograft tumor growth in vivo {#sec3-13} To evaluate the impact of PGAM1 knockdown on tumor growth *in vivo*, we established a subcutaneous xenograft tumor model in athymic nude mice by injecting PC-3 cells infected with shRNA targeting PGAM1 or NC PC-3 cells. Results from Western blotting showed an obvious reduction of PGAM1 expression in PC-3 cells transfected with sh-PGAM1 compared with NC cells (**[Figure 5a](#F5){ref-type="fig"}**). As expected, the cells with PGAM1 knocked down formed slower-growing xenografts compared with NC cells (*P* \< 0.01, **Figure [5b](#F5){ref-type="fig"}**-**[5d](#F5){ref-type="fig"}**). Hematoxylin and eosin staining revealed the histopathological features of the tumor tissues in xenograft tumors (**[Figure 5e](#F5){ref-type="fig"}**). {#F5} DISCUSSION {#sec1-4} ========== Tumor cells prefer the glycolysis pathway to make the oxidative phosphorylation more efficient, even under the nonhypoxia conditions. This characteristic of tumor cells is called "Warburg effect" or aerobic glycolysis, and it helps tumor cells produce more energy than normal cells.[@ref16] PGAM1 is a key enzyme that catalyzes the conversion of 3-PG and 2-PG in the glycolysis pathway. Hitosugi *et al*.[@ref12] found that PGAM1 knockdown elevated 3-PG levels, whereas it reduced 2-PG levels. By regulating the intracellular 3-PG and 2-PG levels in the glycolysis pathway, PGAM1 plays a specific role in coordinating biosynthesis and glycolysis to promote the cancer cell growth. Many studies have found that PGAM1 is overexpressed in diverse cancers, likely due to loss of TP53,[@ref17][@ref18] and is important for tumorigenesis, invasion, and metastasis.[@ref19] Narayanan *et al*.[@ref20] used real-time polymerase chain reaction to demonstrate the expression levels of PGAM1mRNA were higher in LNCaP cell than that in the normal cell. Zhang *et al*.[@ref21] used modified serum-guided immunoblotting, two-dimensional gel electrophoresis, and MALDI-TOF mass spectrography in a differential proteomic study and found that PGAM1 expression levels were higher in PCa tissues than that in benign prostatic hyperplasia tissues. These previous studies strongly suggest that PGAM1 may be associated with PCa. However, neither study investigated the relationship between PGAM1 and PCa in depth. In the present study, we demonstrated upregulated expression of PGAM1 in PCa. These results are consistent with previous reports about PGAM1 protein expression in PCa.[@ref20][@ref21] Furthermore, our results revealed that PGAM1 expression was not associated with patients\' age, clinical stage, or metastasis status, but that patients with higher Gleason scores and T stages exhibited increased PGAM1 expression, suggesting that PGAM1 might contribute to progression and aggressiveness in PCa to some extent. We also investigated the function of PGAM1 in PCa. To date, relatively few studies have examined the biological function of PGAM1. Ren *et al*.[@ref11] demonstrated that knocking down PGAM1 expression with shRNA targeting PGAM1 induced liver cancer cell growth arrest and apoptosis *in vitro* and *in vivo*. Hitosugi *et al*.[@ref12] found that targeting PGAM1 using shRNA or a small molecule inhibitor resulted in notably decreased glycolysis and biosyntheses, accompanied by inhibited leukemia cell proliferation. They further reported that Y26 phosphorylation could enhance PGAM1 activation by stabilizing the active conformation of PGAM1, which promoted cancer cell proliferation and tumor growth.[@ref22] Sanzey *et al*.[@ref23] found that silencing of PGAM1 increased cell death in U87 cells and increased survival in mice with glioblastoma xenografts. A recent study reported that knockdown of PGAM1 by siRNA notably inhibited glioma cell proliferation, migration, and invasion and promoted cancer cell apoptosis.[@ref13] Here, we showed that silencing PGAM1 by transfecting cells with siRNA markedly inhibited cell proliferation. Additionally, siRNA knockdown of PGAM1 notably enhanced cell apoptosis in PC-3 and 22Rv1 cells by downregulating Bcl-2 expression, upregulating Bax expression, and activating the caspased-3 signal. Moreover, knockdown of PGAM1 expression inhibited PCa cell migration and invasion. These results suggest that PGAM1 plays an important role in the progression of PCa by regulating MMP-2 and MMP-9. Some cell lines we used for the study are androgen independent. Thus, our finding may provide new information for further researches of treating CRPC. Our findings strongly indicate that PGAM1 plays an important role in PCa development and progression. Further molecular and functional studies of PGAM1 in PCa, for example to determine the molecular mechanisms underlying the effects observed here, should be conducted. CONCLUSION {#sec1-5} ========== Our data suggest that PGAM1 was upregulated in PCa cells and tissues. Additionally, PGAM1 knockdown efficiently inhibited PCa cell proliferation, migration, and invasion and enhanced cancer cell apotosis *in vitro*. Moreover, PGAM1 knockdown suppressed xenograft tumor growth *in vivo*. These results indicate that PGAM1 may play an important role in the progression and aggressiveness of PCa, and that it might be a valuable marker of poor prognosis and a potential therapeutic target for PCa. AUTHOR CONTRIBUTIONS {#sec1-6} ==================== YAW, BWZ, and SCZ conceived this study, conducted the searching, and drafted the manuscript. DJL participated in statistical analysis and help draft the manuscript. FPS and XLS contributed to the design of this study. BH and CW participated in experimental work. SCZ checked the design of this study, participated in coordination, and provided proposals for the manuscript. All authors read and approved the final manuscript. COMPETING INTERESTS {#sec1-7} =================== All authors declared no competing interests. This study was supported by three Science and Technology planning Projects of Guangdong Province (No. 2013B051000050, No. 2014A020212538, and No. 2016A020215175), the Natural Science Foundation of Guangdong Province (No. 2016A030313583), the Medical Scientific Research Foundation of Guangdong Province (No. A2016555), the Outstanding Youths Development Scheme of Nanfang Hospital, Southern Medical University (No. 2015J005), and the Science and Technology planning Project of Guangzhou (No. 201704020070). [^1]: These authors contributed equally to this work.

INTRODUCTION {#s1} ============ Atherosclerosis is a leading cause of death worldwide ([@BIO041913C4]). The atherosclerotic lesion is characterized by the development of fibroinflammatory lipid plaques in the inner lining of arteries ([@BIO041913C33]). This involves the directed migration of vascular smooth muscle cells (VSMCs) from the medial to the intimal layer of the vessel wall, where they undergo proliferation and synthesize extracellular matrix to form a fibrous cap that covers the atherosclerotic plaque ([@BIO041913C20]). VSMCs are also the predominant cell type found in the lesions during restenosis after angioplasty, stenting and in vein graft disease, where they migrate and proliferate to form the thickened neointima ([@BIO041913C8]). It is well established that soluble growth factors regulate VSMC proliferation and migration ([@BIO041913C22]). Moreover, within the vessel wall, VSMCs make adhesive contacts with neighbouring cells as well as the extracellular matrix. The dynamic remodelling of cell--cell and cell--matrix contacts can regulate cellular and molecular processes in VSMCs to coordinate cell behaviours such as proliferation and migration ([@BIO041913C11]; [@BIO041913C18]; [@BIO041913C22]). Balloon denudation of the endothelium is a model of mechanical injury to the artery that triggers the directional migration of VSMCs to form a thickened neointimal layer ([@BIO041913C8]). Following mechanical injury of the carotid artery in the rat, our lab showed elevated expression of both discoidin domain receptor 1 (DDR1) ([@BIO041913C13]) and N-cadherin in VSMCs ([@BIO041913C19]). DDR1 is a collagen-binding receptor tyrosine kinase, and its deletion reduces VSMC migration towards collagen *in vitro* and attenuates neointimal thickening and atherosclerotic plaque formation *in vivo* ([@BIO041913C10]; [@BIO041913C13]). Recent research has shown that DDR1 can stabilize cadherin-containing contacts, but many studies have focused on the effects of DDR1 in stabilizing E-cadherin contacts in epithelial cells ([@BIO041913C7]; [@BIO041913C9]; [@BIO041913C34]). Furthermore, these effects were found to be context-dependent. In normal epithelial cells, DDR1 forms a complex with E-cadherin, stabilizing cell--cell adhesions ([@BIO041913C9]; [@BIO041913C34]). By contrast, in cancer, DDR1 is upregulated and promotes epithelial-mesenchymal transition (EMT) by increasing the expression of N-cadherin, promoting cell migration and invasion ([@BIO041913C2]; [@BIO041913C16]; [@BIO041913C25]; [@BIO041913C30]). Clearly, the effects of DDR1 on cadherin-based contacts cannot be extrapolated between different cell types and conditions. To the best of our knowledge, there has been no research studying the effects of DDR1 on N-cadherin cell--cell contacts in VSMCs. VSMCs express several types of cadherin molecules, including N-cadherin, T-cadherin, R-cadherin, FAT1-cadherin and OB-cadherin ([@BIO041913C26]; [@BIO041913C33]). OB-cadherin promotes cell--cell adhesion and collectivization of VSMCs ([@BIO041913C3]). T-cadherin ([@BIO041913C17]) stimulates proliferation and induces migration of VSMCs, potentially contributing to intimal hyperplasia in atherosclerotic lesions and vessel stenosis. FAT1- ([@BIO041913C14]) and R-cadherin ([@BIO041913C31]) may have an antiproliferative function through the sequestration of β-catenin, preventing its translocation to the nucleus to activate cyclin D1. FAT1-cadherin increases cell--cell adhesive force and reduces migration and invasion in epithelial cells ([@BIO041913C15]). Previous research from our lab showed that N-cadherin was the most abundant cell--cell adhesion molecule expressed by VSMCs, and that it played an important role in regulating directional migration ([@BIO041913C28]). Specifically, in mechanical wounding experiments performed *in vitro*, the disruption of N-cadherin contacts at the wound edge led to a polarized posterior-lateral distribution of N-cadherin at the cell periphery, and this was required for the directional migration of VSMCs into the wound ([@BIO041913C28]). This migratory response was inhibited by treatment with either N-cadherin blocking antibody or peptide, which bind to N-cadherin molecules on the cell surface to inhibit its function ([@BIO041913C28]). Consistent with our findings, George and colleagues reported that inhibiting N-cadherin using HAV peptides attenuated VSMC migration and neointimal thickening in a human saphenous vein organ culture model ([@BIO041913C23]). N-cadherin belongs to the family of classical cadherins, which are transmembrane cell--cell adhesion molecules that undergo calcium-dependent homophilic binding via the extracellular domain ([@BIO041913C1]). The cytoplasmic domain of N-cadherin can bind to catenin molecules which interact with the actin cytoskeleton to mediate strengthening and stabilization of the junctional complex ([@BIO041913C12]). In addition, N-cadherin contacts are stabilized by association with lipid rafts, which are cholesterol/sphingolipid-enriched microdomains in the cell membrane. Lipid rafts facilitate the functional grouping of proteins during a wide range of processes including cell migration, intracellular trafficking, and signal transduction ([@BIO041913C5]; [@BIO041913C27]; [@BIO041913C29]). Little is known about the lipid raft localization of the N-cadherin adhesion complex or the DDR1 in VSMCs. The role of DDR1 in stabilizing N-cadherin contacts has not yet been investigated. Therefore, in this study, using VSMCs from a DDR1-knockout mouse model, we describe a novel mechanism by which DDR1 regulates N-cadherin cell--cell contacts in VSMCs. RESULTS {#s2} ======= N-cadherin cell--cell contacts were disrupted in DDR1−/− VSMCs {#s2a} -------------------------------------------------------------- Primary DDR1+/+ and DDR1−/− VSMCs were grown to confluence, fixed and stained for N-cadherin to visualize cell--cell contacts. In DDR1+/+ VSMCs, N-cadherin staining was observed in a zipper-like conformation around the cell periphery ([Fig. 1](#BIO041913F1){ref-type="fig"}A). By contrast, in DDR1−/− VSMCs zipper-like structures were reduced and most N-cadherin staining was diffuse in the cytoplasm ([Fig. 1](#BIO041913F1){ref-type="fig"}B). Quantification of N-cadherin junctional staining demonstrated significantly higher (60% greater) N-cadherin staining intensity at cell--cell contacts of DDR1+/+ VSMCs compared to DDR1−/− VSMCs ([Fig. 1](#BIO041913F1){ref-type="fig"}E). While more intense F-actin staining and a greater number of stress fibres was observed in DDR1+/+ VSMCs ([Fig. 1](#BIO041913F1){ref-type="fig"}C), an intact actin network was still observed in DDR1−/− VSMCs ([Fig. 1](#BIO041913F1){ref-type="fig"}D), showing that overall cell integrity was not compromised by DDR1 deletion. Furthermore, there were no differences in cell viability or density between DDR1+/+ and DDR1−/− VSMCs. Immunoblotting whole cell lysates of DDR1+/+ and DDR1−/− VSMCs revealed similar levels of total N-cadherin protein in both genotypes ([Fig. 2](#BIO041913F2){ref-type="fig"}A). To determine whether the level of N-cadherin protein in the membrane was reduced in DDR1−/− SMCs, the membrane protein fraction was isolated by ultracentrifugation and probed with an antibody against N-cadherin ([Fig. 2](#BIO041913F2){ref-type="fig"}B Total, [2](#BIO041913F2){ref-type="fig"}C Membrane). However, there were no significant differences in the total ([Fig. 2](#BIO041913F2){ref-type="fig"}D), or in the membrane protein levels ([Fig. 2](#BIO041913F2){ref-type="fig"}E) of N-cadherin between genotypes. Taken together, this data suggests that the major consequence of DDR1 deficiency is a reduction in the organization of N-cadherin into stable, morphologically distinct cell--cell junctions. Fig. 1.**N-cadherin cell--cell contacts were disrupted in DDR1−/− VSMCs.** (A--D) Immunostaining of N-cadherin (green; 610921 BD Biosciences) and F-actin (red; Alexa 568 phalloidin) in DDR1+/+ and DDR1−/− VSMCs. Nuclei are counterstained with Hoechst (blue). (E) Relative fold change in intensity of junctional N-cadherin staining in DDR1+/+ and DDR1−/− VSMCs. Experiment was repeated three times. Data are plotted as mean±s.e.m. Student's *t*-test was performed. \* indicates statistical significance of *P*\<0.05.Fig. 2.**There was no difference in total cell lysate or whole membrane N-cadherin protein levels between DDR1+/+ and DDR1−/− VSMCs.** (A) Total cell lysate of DDR1+/+ and DDR1−/− VSMCs, immunoblotted for N-cadherin, DDR1 and β-actin. Results were quantified as the level of N-cadherin normalized to β-actin, and expressed relative to DDR1+/+ VSMCs, and Student's *t*-test was performed (*n*=3). Protein levels of N-cadherin and DDR1 in total cell lysate B, and whole membrane fractions C, of DDR1+/+ and DDR1−/− VSMCs. GAPDH and Flotillin-2 were used as controls. Blots were spliced to compose the images, but samples were immunoblotted and developed simultaneously for each blot. Blots in panels B and C were processed separately. Relative fold change in total cell lysate (D), and whole membrane (E), N-cadherin protein levels in DDR1+/+ and DDR1−/− VSMCs normalized to flotillin-2 and expressed relative to DDR1+/+ VSMCs (*n*=3). Data are plotted as mean±s.e.m. Student's *t*-test was performed. N-cadherin co-immunoprecipitated with DDR1 and catenins {#s2b} ------------------------------------------------------- Next we investigated whether DDR1 could associate with N-cadherin in VSMCs. DDR1 and N-cadherin were pulled down together using reciprocal co-immunoprecipitation ([Fig. 3](#BIO041913F3){ref-type="fig"}A,B). Catenins associate with the cytoplasmic tail of N-cadherin, facilitate interaction with the cytoskeleton, and strengthen the junctional complex ([@BIO041913C1]). We performed co-immunoprecipitations to determine whether there were differences in N-cadherin association with various catenins comparing DDR1+/+ and DDR1−/− VSMCs. N-cadherin co-immunoprecipitated with α-, β-, γ-, and p120-catenins, and there were no differences in association comparing the two genotypes of VSMCs ([Fig. 3](#BIO041913F3){ref-type="fig"}A). Quantification of the western blots showed that the association of N-cadherin with catenins did not differ significantly between DDR1+/+ and DDR1−/− VSMCs ([Fig. 3](#BIO041913F3){ref-type="fig"}C). Fig. 3.**N-cadherin co-immunoprecipitated with DDR1, α-, β-, γ-, and p120-catenin.** (A) Immunoprecipitation of N-cadherin or total cell lysate of DDR1+/+ and DDR1−/− VSMCs blotted for N-cadherin, DDR1, α-catenin, β-catenin, γ-catenin and p120-catenin. Blots were spliced to compose the image, but all samples were immunoblotted and developed simultaneously. (B) Immunoprecipitation of DDR1 or total cell lysate of DDR1+/+ VSMCs blotted for DDR1 and N-cadherin. Experiments were repeated three times. (C) Relative fold change of DDR1, α-catenin, β-catenin, γ-catenin, and p120-catenin levels (*n*=3) pulled down with N-cadherin in DDR1+/+ and DDR1−/− VSMCs was normalized to N-cadherin and expressed relative to DDR1+/+ VSMCs. Data are plotted as mean±s.e.m. Student's *t*-test was performed. \* indicates statistical significance of *P*\<0.05. Cytoplasmic N-cadherin in DDR1−/− VSMCs was disrupted by Triton treatment {#s2c} ------------------------------------------------------------------------- To further probe the localization of N-cadherins in DDR1+/+ and DDR1−/− VSMCs, we performed Triton extraction to determine the cellular compartments in which the N-cadherin proteins reside. This takes advantage of the fact that proteins in the cytoplasm, during synthesis, and recycling are Triton soluble, while stable membrane proteins are Triton insoluble. In DDR1+/+ VSMCs, Triton treatment did not disrupt N-cadherin cell--cell contacts, which remained in a zipper-like configuration at the cell periphery ([Fig. 4](#BIO041913F4){ref-type="fig"}A). However, in DDR1−/− VSMCs, cytoplasmic N-cadherin staining was markedly reduced after Triton treatment, with more disruption occurring with increased concentration or time of Triton exposure ([Fig. 4](#BIO041913F4){ref-type="fig"}A). This suggests more N-cadherin was present in the Triton soluble pool in DDR1−/− VSMCs. To measure the amount of N-cadherin present in Triton soluble versus insoluble fractions, DDR1+/+ and DDR1−/− VSMCs were lysed in 1% Triton, and N-cadherin protein levels were assessed by western blotting ([Fig. 4](#BIO041913F4){ref-type="fig"} B,C). DDR1−/− VSMCs showed significantly higher levels of N-cadherin in the Triton soluble fraction compared to DDR1+/+ cells ([Fig. 4](#BIO041913F4){ref-type="fig"}D). By contrast, the DDR1−/− cells had significantly less N-cadherin in the Triton insoluble fraction compared to DDR1 +/+ cells ([Fig. 4](#BIO041913F4){ref-type="fig"}E). Fig. 4.**N-cadherin in DDR1−/− VSMCs was disrupted by Triton treatment.** (A) Immunofluorescence of DDR1+/+ and DDR1−/− VSMCs treated with or without 0.25% or 1% Triton for 1 or 5 min on ice, fixed and stained for N-cadherin (red; 610921 BD Biosciences). Nuclei are counterstained with Hoechst (blue). Protein levels of N-cadherin, DDR1, and β-actin in triton-soluble (B), and triton-insoluble (C), fractions of cell lysate from DDR1+/+ and DDR1−/− VSMCs. Blots were spliced to compose the image, but all samples were immunoblotted and developed simultaneously. Blots in panels B and C were processed separately. Relative fold changes of triton-soluble (*n*=8) (D), and triton-insoluble (*n*=4) (E). N-cadherin protein in DDR1+/+ and DDR1−/− VSMCs normalized to β-actin, and expressed relative to DDR1+/+ VSMCs. Data are plotted as mean±s.e.m. Student's *t*-test was performed. \* indicates statistical significance of *P*\<0.05. Levels of N-cadherin and catenins were reduced in lipid rafts from DDR1−/− VSMCs {#s2d} -------------------------------------------------------------------------------- Lipid rafts are sphingolipid- and cholesterol-rich microdomains in the cell membrane that can cluster and organize proteins into signalling complexes, and comprise a cell membrane structure that is resistant to Triton solubilisation ([@BIO041913C21]). To determine whether N-cadherin cell--cell junctions associate with lipid rafts in wild-type VSMCs, we first labelled lipid rafts with fluorescent cholera toxin B that binds to ganglioside GM1, and then fixed and stained the cells for N-cadherin. N-cadherin and GM1 were co-localized at regions of cell--cell contact, and quantification revealed that approximately 2% of N-cadherin signal co-localized with GM1 ([Fig. 5](#BIO041913F5){ref-type="fig"}A). To compare the levels of N-cadherin protein in lipid rafts between DDR1+/+ and DDR1−/− VSMCs, cells were lysed in 1% Triton, lipid raft proteins were separated by ultracentrifugation on a sucrose density gradient, then fractions were subjected to western blotting. Staining with antibodies for the lipid raft markers flotillin-2 and caveolin-1 showed that lipid raft proteins were concentrated in fractions 3--5 ([Fig. 5](#BIO041913F5){ref-type="fig"}B). GAPDH, a cytoplasmic protein loading control, was absent from these fractions and present in fractions 6--9 ([Fig. 5](#BIO041913F5){ref-type="fig"}B). N-cadherin, DDR1, α-, β-, and p120-catenin were present in the lipid raft fractions from DDR1+/+ VSMCs, but markedly reduced in the lipid raft fractions of DDR1−/− VSMCs ([Fig. 5](#BIO041913F5){ref-type="fig"}B). Because proteins that associate with membrane lipid rafts constitute only a small fraction of the total proteins inside the cell, the majority of N-cadherin, DDR1 and catenin proteins were found in non-lipid raft fractions (fractions 7--9). Fig. 5.**Levels of N-cadherin and catenins were reduced in lipid rafts from DDR1−/− VSMCs.** (A) Immunofluorescence of DDR1+/+ VSMCs stained for N-cadherin (red; 610921 BD Biosciences) and GM1 (green; Alexa 488 cholera toxin B, C34775 Life Technologies). Nuclei are counterstained with Hoechst (blue). (B) Protein levels of N-cadherin, DDR1, α-catenin, β-catenin, p120-catenin, flotillin-2, caveolin-1, GAPDH, and β-actin in total cell lysate (T), and fractions (1--9) after separation of cell lysate from DDR1+/+ and DDR1−/− VSMCs along a 40%/30%/5% sucrose gradient. (C) Protein levels of N-cadherin, DDR1, α-catenin, β-catenin, p120-catenin, flotillin-2, caveolin-1, GAPDH and β-actin in total cell lysate and fractions 3--5. Blots were spliced to compose the image, but all samples were immunoblotted and developed simultaneously. (D) Relative fold changes of N-cadherin, α-catenin, β-catenin, p120-catenin and β-actin (*n*=3) in lipid raft fractions of DDR1+/+ and DDR1−/− VSMCs normalized to caveolin-1, expressed relative to DDR1+/+ VSMCs. Data are plotted as mean±s.e.m. Student's *t*-test was performed. \* indicates statistical significance of *P*\<0.05. To facilitate better comparison of protein levels in the lipid raft fractions of DDR1+/+ and DDR1−/− VSMCs, the lipid raft fractions from several dishes were pooled and run side by side on an SDS-PAGE gel and analysed by western blotting ([Fig. 5](#BIO041913F5){ref-type="fig"}C). There were significant reductions in the levels of N-cadherin, α-, β, and p120-catenins in the lipid rafts from DDR1−/− VSMCs compared to DDR1+/+ VSMCs ([Fig. 5](#BIO041913F5){ref-type="fig"}D). β-actin levels were also reduced in the lipid raft fractions of DDR1−/− compared to DDR1+/+ VSMCs ([Fig. 5](#BIO041913F5){ref-type="fig"}C), likely due to reduced anchoring of N-cadherin to the membrane and cytoskeleton in the DDR1−/− VSMCs. There were no differences in levels of these proteins in total cell lysates between DDR1+/+ and DDR1−/− VSMCs ([Fig. 5](#BIO041913F5){ref-type="fig"}C). Disruption of lipid rafts resulted in the reduction of N-cadherin, catenins and DDR1 {#s2e} ------------------------------------------------------------------------------------ Cholesterol oxidase and methyl-β-cyclodextrin (MβCD) disrupt lipid rafts by altering their cholesterol content, either through catalysing cholesterol catabolism or directly removing cholesterol molecules. Cells were treated with cholesterol oxidase or MβCD to disrupt lipid rafts, then stained for N-cadherin to observe the effects on cell--cell contacts. In untreated cells, N-cadherin contacts were arranged in a zipper conformation around the periphery of DDR1+/+ VSMCs, while N-cadherin staining was more diffuse in the cytoplasm of DDR1−/− VSMCs ([Fig. 6](#BIO041913F6){ref-type="fig"}A). Treatment with cholesterol oxidase disrupted peripheral, junctional N-cadherin staining in DDR1+/+ VSMCs, with dose and time dependent increases in disruption ([Fig. 6](#BIO041913F6){ref-type="fig"}A). By contrast, cholesterol oxidase treatment did not affect the diffuse cytoplasmic N-cadherin staining in DDR1−/− VSMCs ([Fig. 6](#BIO041913F6){ref-type="fig"}A). MβCD treatment had similar results, disrupting N-cadherin at cell--cell contacts of DDR1+/+ VSMCs with little effect on cytoplasmic N-cadherin staining in DDR1−/− VSMCs ([Fig. 6](#BIO041913F6){ref-type="fig"}B). The disruptive effect of MβCD on the cell membrane was rescued by treating the cells with exogenous cholesterol, which restored N-cadherin at cell--cell contacts in DDR1+/+ VSMCs ([Fig. 6](#BIO041913F6){ref-type="fig"}B). Quantification of N-cadherin junctional staining in these experiments revealed a decrease from 0.94±0.03 in untreated cells to 0.44±0.04 in MβCD treated cells. Subsequent rescue of the MβCD treated cells with exogenous cholesterol resulted in the restoration of junctional N-cadherin staining to control levels 0.96±0.08. These results were confirmed by analysing proteins in lipid rafts from MβCD treated cells. MβCD treatment led to the displacement of N-cadherin, the catenins and DDR1 from lipid raft fractions of DDR1+/+ VSMCs ([Fig. 7](#BIO041913F7){ref-type="fig"}A). Cholesterol treatment restored the presence of N-cadherin, DDR1, and associated catenins in lipid raft fractions of DDR1+/+ VSMCs ([Fig. 7](#BIO041913F7){ref-type="fig"}A). In DDR1−/− VSMCs MβCD treatment did not affect N-cadherin or catenin levels in lipid rafts ([Fig. 7](#BIO041913F7){ref-type="fig"}B). However, cholesterol treatment actually increased the amount of N-cadherin and its associated catenins in the lipid raft fractions of DDR1−/− VSMCs ([Fig. 7](#BIO041913F7){ref-type="fig"}B), but this was not accompanied by restoration of N-cadherin at morphologic cell--cell contacts ([Fig. 6](#BIO041913F6){ref-type="fig"}B). Fig. 6.**Disruption of lipid rafts resulted in the reduction of N-cadherin contacts between cells.** (A) Immunofluorescence of DDR1+/+ and DDR1−/− VSMCs treated with or without 1 or 2 units of cholesterol oxidase for 1 or 2 h, fixed and stained for N-cadherin (red; 610921 BD Biosciences). Nuclei are counterstained with Hoechst (blue). (B) Immunofluorescence of DDR1+/+ and DDR1−/− VSMCs treated with or without 1 mM of MβCD for 30 min, followed by 1 mM of MβCD-cholesterol for 30 min or 1 h, fixed and stained for N-cadherin (green; 610921 BD Biosciences).Fig. 7.**Disruption of rafts resulted in reduction of N-cadherin, catenins and DDR1 in the lipid raft fraction.** Protein levels of N-cadherin, DDR1, α-catenin, β-catenin, p120-catenin, flotillin-2, caveolin-1, GAPDH and β-actin in total cell lysate (T) and fractions (1--9) after separation of cell lysate along a 40%/30%/5% sucrose gradient from DDR1+/+ (A), and DDR1−/− (B), VSMCs that were untreated or treated with 1 mM MβCD, followed by 1 mM of MβCD-cholesterol for 30 min or 1 h. Experiments were repeated three times. N-cadherin junction staining was reduced after DDR1 knockdown, and rescued after DDR1b overexpression {#s2f} ----------------------------------------------------------------------------------------------------- To ensure the disruption in N-cadherin cell--cell contacts observed in DDR1−/− VSMCs was specifically due to DDR1 deletion instead of possible confounding effects in the global DDR1 knockout model, DDR1+/+ and DDR1−/− VSMCs were treated with DDR1 siRNA and stained for N-cadherin. DDR1 levels were decreased ∼70% in cells treated with DDR1siRNA ([Fig. 8](#BIO041913F8){ref-type="fig"}C). In untreated cells and cells treated with RNAiMAX (vehicle) or scrambled siRNA, N-cadherin contacts were arranged in a zipper conformation around the periphery of DDR1+/+ VSMCs, while N-cadherin staining was more diffuse in the cytoplasm of DDR1−/− VSMCs ([Fig. 8](#BIO041913F8){ref-type="fig"}A). DDR1 silencing reduced the junctional staining of N-cadherin and its zipper-like conformation in DDR1+/+ VSMCs, but had very little effect on cytoplasmic N-cadherin staining in DDR1−/− VSMCs ([Fig. 8](#BIO041913F8){ref-type="fig"}A). Quantification revealed that junctional staining for N-cadherin was reduced by 43% after DDR1 siRNA treatment in DDR1+/+ VSMCs, but the low level of junctional staining was not significantly affected by DDR1 siRNA treatment in DDR1−/− VSMCs ([Fig. 8](#BIO041913F8){ref-type="fig"}D). Fig. 8.**N-cadherin junction staining was reduced after DDR1 knockdown, and rescued after DDR1b overexpression.** (A) Immunofluorescence of DDR1+/+ and DDR1−/− untreated VSMCs, cells transfected with RNAiMAX only, or scrambled siRNA or DDR1 siRNA, then fixed and stained for N-cadherin (red; 610921 BD Biosciences). Nuclei are counterstained with Hoechst (blue). (B) DDR1+/+ and DDR1−/− VSMCs were transfected with DDR1b plasmid DNA, then fixed and stained for DDR1 (green; D1G6 Cell Signaling Technology) and N-cadherin (red; 610921 BD Biosciences). Nuclei are counterstained with Hoechst (blue). Green staining identifies transfection and expression of the DR1b plasmid. Arrows point to junctional N-cadherin staining in VSMCs expressing exogenous DDR1b. (C) DDR1 protein levels in cells transfected with control or DDR1 siRNA. Experiment was repeated three times. Data are plotted as mean±s.e.m. Student's *t*-test was performed. (D) Relative fold change in the intensity of junctional N-cadherin staining in DDR1+/+ and DDR1−/− VSMCs after siRNA treatment. Designation with the same letter represents values not significantly different. Experiment was repeated three times. Data are plotted as mean±s.e.m. One-way ANOVA followed by Holm-Sidak test was performed. (E) Relative fold change in the intensity of junctional N-cadherin staining in DDR1+/+ and DDR1−/− VSMCs after DDR1b transfection. Designation with the same letter represents values not significantly different. Experiment was repeated three times. Data are plotted as mean±s.e.m. One-way ANOVA followed by Holm-Sidak test was performed. To further confirm the direct effect of DDR1 on N-cadherin cell--cell contact formation in VSMCs, DDR1+/+ and DDR1−/− VSMCs were transfected to overexpress full length DDR1b plasmid. Cells were immunostained with an antibody against DDR1 to detect successful transfection of the DDR1b plasmid (green, intense stain due to overexpression of high levels of DDR1) followed by immunostaining for N-cadherin (red). [Fig. 8](#BIO041913F8){ref-type="fig"}B, lower panels shows several DDR1−/− cells overexpressing DDR1b (green), and the establishment of N-cadherin stained cell--cell junctions (red) ([Fig. 8](#BIO041913F8){ref-type="fig"}B, arrows). Quantification of the immunofluorescent staining revealed that expression of DDR1b increased junctional N-cadherin by 73% in DDR1−/− VSMCs ([Fig. 8](#BIO041913F8){ref-type="fig"}E). By contrast, DDR1b transfection did not further increase N-cadherin junctional staining in DDR1+/+ VSMCs ([Fig. 8](#BIO041913F8){ref-type="fig"}E). The cytoskeleton was disrupted in DDR1−/− VSMCs {#s2g} ----------------------------------------------- Cell contacts are stabilized by anchoring to the cytoskeleton, but our previous studies have revealed actin and microtubule disruption in DDR1−/− VSMCs. We stained DDR1+/+ and DDR1−/− VSMCs for F-actin and α-tubulin to visualize the cytoskeleton. Immunostaining cells with DDR1 antibody revealed some non-specific staining in the DDR1−/− cells due to antibody cross-reacting with other proteins ([Fig. 9](#BIO041913F9){ref-type="fig"}), as has been noted previously on western blots. DDR1−/− VSMCs showed weak F-actin staining, with a dramatic decrease in the number of stress fibres, a reduction in cortical actin, and increased membrane ruffles that overlapped at contact sites between neighbouring cells ([Fig. 9](#BIO041913F9){ref-type="fig"}, arrows and inset panel). Microtubules were less abundant and less aligned in DDR1−/− VSMCs ([Fig. 9](#BIO041913F9){ref-type="fig"}). Cells were treated with cytochalasin D to inhibit actin polymerization and determine whether this affected N-cadherin cell--cell contacts. DDR1+/+ VSMCs showed disruption of N-cadherin cell--cell contacts after cytochalasin D treatment compared to untreated cells (data not shown). By contrast, while cytochalasin D had a disruptive effect on the actin cytoskeleton of DDR1−/− VSMCs, it did not lead to disruption of the cytoplasmic N-cadherin staining in DDR1−/− VSMCs compared to untreated cells (data not shown). Fig. 9.**The cytoskeleton was disrupted in DDR1−/− VSMCs.** Immunofluorescence of DDR1+/+ and DDR1−/− VSMCs fixed and stained for DDR1 (green; D1G6 Cell Signaling Technology), F-actin (red; Alexa 568 phalloidin), α-tubulin (red; ab52866 Abcam). Arrows identify areas of membrane ruffling and overlap in DDR1−/− VSMCs. Dotted lines enclose an area showing a membrane ruffle which is shown at higher magnification below. DISCUSSION {#s3} ========== This study was conducted to further our understanding of the molecules and structures in VSMCs that are important for proper N-cadherin junction establishment, and to determine whether DDR1 associates with and promotes the expression or function of N-cadherin. Our results show that DDR1 regulates N-cadherin junction stability and localization to lipid rafts in VSMCs. Knockout or knockdown of DDR1 in VSMCs resulted in disruption of N-cadherin contacts. Analysis of lipid raft fractions revealed decreases in the amounts of N-cadherin and associated catenins in DDR1−/− compared to DDR1+/+ VSMCs. Importantly, transfection of DDR1−/− cells with full-length DDR1b rescued the KO phenotype and promoted the formation of N-cadherin junctions. N-cadherin is the most abundant cell--cell adhesion molecule expressed by VSMCs, and it plays an important role in regulating cellular functions including migration, proliferation, and survival ([@BIO041913C23]; [@BIO041913C28]; [@BIO041913C32]). Research from our lab has shown that wound edge VSMCs after mechanical wounding *in vitro* displayed a polarized posterior-lateral distribution of N-cadherin cell--cell contacts, which was required for front polarization of the microtubule organizing centre, anterior positioning of hyper-stabilized microtubules to facilitate membrane transport, activation of Cdc42 at the leading edge, inhibition of GSK3β at the posterior-lateral edge, and directional migration into the wound ([@BIO041913C28]). The effects of N-cadherin on Rho GTPases were also found in C2C12 myoblasts where the establishment of N-cadherin contacts inhibited Cdc42 and Rac1 activity as well as filopodia and lamellipodia formation ([@BIO041913C6]). In VSMCs, downregulation and disruption of N-cadherin cell--cell contacts were associated with increased proliferation caused by the translocation of β-catenin into the nucleus to activate transcription ([@BIO041913C32]). Furthermore, inhibiting N-cadherin function and abolishing N-cadherin expression increased apoptosis in VSMCs and greatly impacted cell survival ([@BIO041913C23]). Overall, these findings suggest that the ability to establish proper N-cadherin cell--cell contacts is crucial to VSMC function. While interactions between DDR1 and N-cadherin have not been previously investigated in VSMCs, both molecules were found in separate studies to be upregulated in the neointima after mechanical injury of the carotid arteries coincident with the time course of active proliferation and migration of these cells ([@BIO041913C13]; [@BIO041913C19]). Upon deletion of DDR1 in mice, VSMC migration after denuding injury was reduced, mice developed smaller atherosclerotic plaques and DDR1−/− VSMCs exhibited reduced migration *in vitro* ([@BIO041913C10]; [@BIO041913C13]). VSMC migration and neointimal formation were also impaired after the functional inhibition of N-cadherin ([@BIO041913C23]; [@BIO041913C28]). Our current results show for the first time that DDR1 and N-cadherin interact in VSMCs, and suggest that DDR1 influences the localization and stability of cell adhesion junctions, identifying a pathway whereby matrix and cell adhesions coordinate to regulate cell migration and proliferation. Since DDR1 has been shown to associate with E-cadherin in epithelial cells to enhance the stabilization of E-cadherin-based junctions ([@BIO041913C9]; [@BIO041913C34]), we investigated whether DDR1 could associate with N-cadherin in VSMCs. DDR1 and N-cadherin co-immunoprecipitated with each other, suggesting that the two molecules exist in physical proximity to each other in VSMCs, likely in the lipid rafts. In epithelial cells DDR1 has been shown to upregulate N-cadherin mRNA expression and protein synthesis during the process of EMT ([@BIO041913C25]). However, in VSMCs, total protein levels of N-cadherin were similar between DDR1+/+ and DDR1−/− VSMCs, arguing against a role for DDR1 in regulating N-cadherin expression and synthesis. This suggests that DDR1-mediated regulation of N-cadherin junctions in VSMCs occurs through a different mechanism. Since there was disruption of N-cadherin contacts on the cell membrane in DDR1−/− VSMCs, we measured total membrane levels of N-cadherin protein in DDR1+/+ and DDR1−/− VSMCs. However, we did not detect a difference between the two genotypes. Another factor that contributes to N-cadherin contact formation and stabilization is its association with catenin molecules. However, when we immunoprecipitated N-cadherin from DDR1+/+ and DDR1−/− VSMCs, we detected similar levels of α-, β-, γ-, and p120-catenin associated with N-cadherin, suggesting that the disruption of N-cadherin contacts in DDR1−/− VSMCs is unlikely due to altered binding of catenin molecules during junctional complex formation. Lipid rafts comprise specialized microdomains within the cell membrane and serve as an important nexus for cell signalling; cell--cell and cell--matrix adhesion molecules are known to concentrate in rafts. Previous reports in C2C12 myoblasts showed that concentration of N-cadherin in lipid rafts decreased lateral diffusion and mobility within the plasma membrane, and this stabilizing effect occurred without changes in the total level of N-cadherin in the membrane ([@BIO041913C5]). To determine how DDR1 might influence N-cadherin-mediated adhesion in VSMCs, we studied the localization of N-cadherin junctional complexes to lipid rafts in DDR1+/+ and DDR1−/− cells. Performing Triton extraction, co-immunostaining of N-cadherin and GM1, fractionation over a sucrose density gradient, and cholesterol depletion and rescue, we showed that the N-cadherin and catenin junctional complex proteins were associated with lipid rafts, and that this association was reduced in DDR1−/− VSMCs. Moreover, we present convincing evidence that DDR1 also localizes to lipid rafts. Therefore, we propose that the close association between DDR1 and N-cadherin in lipid rafts of VSMCs helps cluster N-cadherin junctional complex into lipid rafts, immobilizing and stabilizing contacts, and in turn stabilizing DDR1. Another important factor in contact stabilization is the association with the actin cytoskeleton, which can help anchor and cluster junctional complexes. In DDR1−/− VSMCs actin staining intensity and stress fibres were reduced, and microtubule organization was altered. Both cytoskeletal structures play important roles in adherens junction formation. Transport along microtubules is important for trafficking of N-cadherin to the plasma membrane where it can form cell--cell contacts ([@BIO041913C24]). N-cadherin staining in the cell cytoplasm was increased in DDR1−/− cells, so it is possible that N-cadherin trafficking was altered. It is also possible that recycling of N-cadherin in endocytotic vesicles was impaired. Another explanation is that in the absence of DDR1, N-cadherin in the plasma membrane cannot localize to lipid rafts and/or anchor to the actin cytoskeleton, therefore adherens junctions do not mature and stabilize. This was further supported by our data demonstrating the disruptive effect of cytochalasin D on N-cadherin contacts in DDR1+/+ VSMCs. In conclusion, we have shown that DDR1 facilitates the stabilization of N-cadherin cell--cell contacts in VSMCs. Specifically, we have shown that DDR1 associates with N-cadherin, and the deletion of DDR1 disrupted N-cadherin cell--cell contacts in VSMCs through a reduction in lipid raft association as well as effects on the cytoskeleton. The molecular mechanisms underlying N-cadherin-mediated adhesion in VSMCs are important in regulating the migratory and proliferative behaviour of VSMCs, which are essential processes contributing to atherosclerotic plaque formation and restenosis after angioplasty, stenting or in vein grafts. MATERIALS AND METHODS {#s4} ===================== All reagents were obtained from Sigma Chemical unless stated otherwise. Cell culture {#s4a} ------------ Animal experiments were performed in accordance with the guidelines of the Canada Council on Animal Care, with the approval of the University of Toronto Faculty of Medicine Animal Care Committee. Primary mouse VSMCs from the carotid arteries of male and female DDR1+/+;C57Bl6 and DDR1−/−;C57Bl6 mice were used between passages 5 and 9. VSMCs were grown in Dulbecco\'s Modified Eagles Medium (DMEM) supplemented with 10% foetal bovine serum, and 2% penicillin, streptomycin and amphotericin (GIBCO BRL, Life Technologies Inc., Rockville, MD, USA) in humidified 95% air and 5% carbon dioxide in an incubator at 37°C. Membrane fractionation {#s4b} ---------------------- DDR1+/+ and DDR1−/− VSMCs were plated on 150-mm plastic tissue culture plates at 25,000 cells/cm^2^ and grown to confluence for 4--5 days then serum starved overnight. Cells were then rinsed with ice-cold PBS, lysed by scraping on ice into 600 μl sucrose buffer (250 mM sucrose, 20 mM HEPES, 10 mM KCl, 2 mM MgCl~2~, 1 mM EDTA, 1 mM EGTA), supplemented with 1 mM DTT, 100 µM vanadate, 100 µM phenylmethylsulfonyl fluoride (PMSF), protease inhibitor cocktail (Catalogue No. 11836170001, Roche Applied Science, Germany), and homogenized by passing the cell lysate through a 30G needle ten times using a 1 ml syringe. A portion was saved as the whole cell lysate. Cells were left on ice for 20 min to lyse and then centrifuged at 800 ***g*** for 5 min at 4°C to pellet nuclei. The supernatant was centrifuged again at 10,000 ***g*** for 5 min to pellet mitochondria. Finally, the supernatant was ultracentrifuged at 100,000 ***g*** for 45 min to pellet membrane. The resulting supernatant was saved as the cytoplasmic portion, and the pellet was resuspended in TBS+0.1% SDS and saved as the membrane portion. All lysate portions were dissolved in Laemmli buffer with 5% β-mercaptoethanol, boiled for 5 min, and analysed by western blotting. Co-immunoprecipitation {#s4c} ---------------------- DDR1+/+ and DDR1−/− VSMCs were plated on 150-mm plastic tissue culture plates at 25,000 cells/cm^2^ for 4--5 days to near confluence then serum starved overnight. Cells were lysed with 1× cell lysis buffer (\#9803, Cell Signaling Technology, Inc., MA, USA, 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM Na~2~EDTA, 1 mM EGTA, 1% Triton, 2.5 mM sodium pyrophosphate, 1 mM beta-glycerophosphate, 1 mM Na~3~VO~4~, 1 µg/ml leupeptin) supplemented with 1 mM PMSF, 20 mM NaF, and protease inhibitor cocktail, sonicated or passed through a 30 G needle five times, rotated at 4°C for 15 min to lyse, and centrifuged at 14,000 ***g*** for 15 min at 4°C. Supernatant was incubated with anti-DDR1 antibody (1:100, D1G6, \#5583, Cell Signaling Technology Inc.) or anti-N-cadherin antibody (1:100, ab18203, Abcam) overnight at 4°C with constant rotation. Afterwards, Protein A agarose beads (\#9863, Cell Signaling Technology Inc.) were added to the lysate to incubate for 3 h at 4°C with constant rotation. Afterwards beads were washed three times with lysis buffer. Beads and total cell lysates were denatured in Laemmli buffer with 5% β-mercaptoethanol, boiled for 5 min, and subjected to western blotting. Triton extraction {#s4d} ----------------- DDR1+/+ and DDR1−/− VSMCs were plated on 18 mm round glass coverslips at 20,000--30,000 cells/cm^2^ and grown to near confluence, serum starved overnight, treated with 1% Triton in ice-cold 10 mM PIPES, pH 6.8, 50 mM NaCl, 3 mM MgCl~2~, 300 mM sucrose, and 1 mM PMSF for 5 min, then fixed in 4% paraformaldehyde and stained for N-cadherin. DDR1+/+ and DDR1−/− VSMCs were grown on 100-mm plastic tissue culture plates to near confluence and lysed for 15 min at 4°C with constant rotation in 1× cell lysis buffer (\#9803, Cell Signaling Technology Inc.), supplemented with 1 mM PMSF, 20 mM NaF, and protease inhibitor cocktail (Roche). Cell lysate was sonicated and centrifuged at 13,200 ***g*** for 15 min at 4˚C. Supernatant and pellet was saved as triton-soluble and triton-insoluble portion, respectively. Lysates were dissolved in Laemmli buffer with 5% β-mercaptoethanol, boiled for 5 min, and subjected to western blotting. GM1 staining {#s4e} ------------ DDR1+/+ VSMCs were plated on 22 mm×22 mm glass coverslips at 20,000--30,000 cells/cm^2^ and grown to near confluence and incubated with 15 μg/ml of Alexa488-conjugated cholera toxin subunit B (C34775, Life Technologies Inc.), which binds to GM1 enriched in lipid rafts, for 15 min on ice. Afterwards, cells were fixed in 4% paraformaldehyde and stained for N-cadherin. Lipid raft isolation {#s4f} -------------------- DDR1+/+ and DDR1−/− VSMCs were plated on 150-mm plastic tissue culture plates at 25,000 cells/cm^2^ and grown to near confluence and serum starved overnight, rinsed with ice-cold PBS, and lysed in 25 mM Tris, pH 7.5, 150 mM NaCl, and 5 mM EDTA containing 1% Triton, 200 μM NaF, 100 μM PMSF, 100 μM sodium orthovanadate, and protease inhibitor tablet. Cell lysate was rotated at 4°C for 30--60 min, then passed ten times through a 30 G needle. Equal concentrations and volumes of cell lysate were mixed with 80% sucrose to obtain 40% final sucrose concentration, overlaid with 30% sucrose followed by 5% sucrose for a final volume of 4.5 ml, and centrifuged at 200,000 ***g*** at 4°C for 16--18 h. From the top, 500 μl fractions were collected, dissolved in 2× Laemmli buffer with 5% β-mercaptoethanol, boiled for 5 min, and subjected to western blotting. Cholesterol oxidase treatment {#s4g} ----------------------------- DDR1+/+ and DDR1−/− VSMCs were plated on 18 mm round glass coverslips at 20,000--30,000 cells/cm^2^ and grown on to near confluence and serum starved overnight, treated with 1 or 2 units of cholesterol oxidase (C8649, Sigma-Aldrich) for 1 or 2 h at 37°C. Afterwards, cells were fixed in 4% paraformaldehyde and stained for N-cadherin. Methyl-β-cyclodextrin treatment and cholesterol rescue {#s4h} ------------------------------------------------------ DDR1+/+ and DDR1−/− VSMCs were plated at 20,000--30,000 cells/cm^2^ on 18 mm round glass coverslips and grown to near confluence and serum starved overnight, treated with 1 mM of methyl-β-cyclodextrin (MβCD, C4555, Sigma-Aldrich) for 30 min at 37°C. Afterwards, cells were treated with or without 1 mM of MβCD-cholesterol complex (C4951, Sigma-Aldrich) for 30 min or 1 h at 37°C. Cells were then fixed in 4% paraformaldehyde and stained for N-cadherin. DDR1 siRNA transfection {#s4i} ----------------------- DDR1+/+ and DDR1−/− VSMCs were plated at 30,000 cells/cm^2^ and grown to 90% confluence and transfected with or without RNAiMAX only (Thermo Fisher Scientific), scrambled siRNA (AM4635, Thermo Fisher Scientific), or DDR1-specific siRNA (AM16704, ID 159939, Ambion, Thermo Fisher Scientific) using Lipofectamine RNAiMAX Transfection Reagent (Cat No. 13778075, Thermo Fisher Scientific) according to the manufacturer\'s instructions. Briefly, lipofectamine RNAiMAX was mixed with siRNA, allowed to incubate at room temperature for 5 min, and added to the cells for a final siRNA concentration of 10 nM. Cells were cultured for 24 h, fixed in 4% paraformaldehyde, and stained for N-cadherin. Parallel cultures were lysed and prepared for western blots to measure the efficiency of DDR1 knockdown. DDR1b transfection {#s4j} ------------------ Plasmid containing full-length DDR1b isoform (a gift from the late Dr Wolfgang Vogel) was transformed into competent DH5-α *E. coli* according to the manufacturer\'s instructions (C2987H; New England Biolabs). Plasmid purification was performed using Maxi Prep DNA isolation kit according to the manufacturer\'s instructions (K210026; Invitrogen). Plasmids containing DDR1b or empty vector were transfected into DDR1+/+ and DDR1−/− VSMCs using Lipofectamine-3000 according to the manufacturer\'s instructions (L3000015; Thermo Fisher Scientific). Briefly, cells were plated at 30,000 cells/cm^2^ in six-well plates and grown to 90% confluence. Lipofectamine 3000 (3.75 μl/well) and plasmid DNA (2.5 μg/well) were mixed, allowed to incubate at room temperature for 5 min, and added directly to cells. Cells were cultured for 24 h, fixed in 4% paraformaldehyde, and stained for DDR1 and N-cadherin. Cytochalasin D treatment {#s4k} ------------------------ DDR1+/+ and DDR1−/− VSMCs were grown on 18 mm round glass coverslips to near confluence and serum starved overnight, treated with 1 μg/ml of cytochalasin D (C2618, Sigma-Aldrich) for 1 h at 37°C. Afterwards, cells were fixed in 4% paraformaldehyde and stained for N-cadherin. Immunofluorescent staining and confocal microscopy {#s4l} -------------------------------------------------- Cells were fixed with 4% paraformaldehyde for 5 min or 100% ice-cold methanol for 10 min, rinsed in PBS, incubated with 0.2% Triton X-100 in PBS for 5 min, and washed in PBS three times at 5 min intervals. The coverslips were incubated with mouse monoclonal anti-N-cadherin (1:250, 610921, BD Biosciences), rabbit anti-DDR1 (1:700, D1G6, Cell Signaling Technology Inc.), or rabbit anti-α-tubulin (1:250, ab52866, Abcam), primary antibody for 1 h at room temperature. The coverslips were washed three times with PBS at 5-min intervals. Secondary antibodies donkey anti-mouse Alexa 488, goat-anti-rabbit Alexa 488, donkey-anti-mouse Alexa 568, goat-anti-rabbit Alexa 568 (1:200, Invitrogen), Alexa 568-conjugated phalloidin (1:100, Invitrogen), and Hoechst 33342 (1:5000, Lonza) were applied to washed coverslips and incubated for 20 min. The coverslips were then washed again three times with PBS at 5 min intervals followed by dipping in distilled water. Prolong Gold (Invitrogen) was used for mounting of coverslips to glass slides for microscopic observation. Coverslips were examined using the 60× objective of a laser scanning confocal microscope (FluoView 1000 Laser Scanning Confocal Microscope, Olympus IX81 Inverted Microscope, Olympus Co. Canada). Serial optical sections were taken at 0.5-μm thickness to include the entire thickness of the cells. Z-stacks were analysed and displayed for each experiment. Western blot {#s4m} ------------ Proteins were separated on SDS-PAGE gels at 70--90 V for 1.5--2 h at room temperature in running buffer (2.5 mM Tris, 25 mM glycine, 0.1% SDS), and transferred overnight at 35 V at 4°C in transfer buffer (2.5 mM Tris 19.2 mM glycine) to polyvinylidene fluoride membranes. Immunoblotting was performed with rabbit anti-DDR1 antibody (D1G6, 1:1000, Cell Signaling Technology), rabbit anti-N-cadherin antibody (ab76057, 1:5000, Abcam), rabbit anti-β-catenin antibody (8480, 1:1000, Cell Signaling Technology), rabbit anti-γ-catenin antibody (2309, 1:1000, Cell Signaling Technology), rabbit anti-p120-catenin antibody (ab92514, 1:1000, Abcam), rabbit anti-flotillin-2 antibody (3436, 1:1000, Cell Signaling Technology), rabbit anti-caveolin-1 antibody (D46G3, \#3267, 1:1000, Cell Signaling Technology), rabbit anti-β-actin antibody (ab8224, 1:1000, Abcam), followed by HRP-conjugated anti-mouse or anti-rabbit secondary antibody (1:10,000, Jackson ImmunoResearch Laboratories, Inc. USA), or HRP-conjugated anti-GAPDH antibody (ab9482, 1:5000, Abcam). Blots were stripped and reprobed for different proteins. Immunoreactive bands were visualized using Westar SuperNova Chemiluminescence Detection Kit (Cyanagen, Bologna, Italy) or Western Lightning Plus-ECL (PerkinElmer Inc.) and the image was captured using MicroChemi bio-imaging system (Froggabio, Toronto, Canada) or ChemiDoc Touch Imaging System (Bio-Rad Laboratories Inc.). Western blots were quantified by measuring the intensity of bands (ImageJ, NIH), and normalizing to β-actin ([Figs 2](#BIO041913F2){ref-type="fig"}A, [4](#BIO041913F4){ref-type="fig"}D,E, [8](#BIO041913F8){ref-type="fig"}C), flotillin-2 ([Fig. 2](#BIO041913F2){ref-type="fig"}D,E), N-cadherin ([Fig. 3](#BIO041913F3){ref-type="fig"}C), or caveolin-1 ([Fig. 5](#BIO041913F5){ref-type="fig"}D). To enable statistical analysis, for each experiment one control sample was arbitrarily chosen and set to a value of one. The value of every other control and experimental sample was then calculated relative to that sample, and the mean of the control and experimental groups and standard error on the means were determine and plotted. This enabled us to assess the variance in the control and experimental sample groups, and perform statistical analysis on the data. Image processing and analysis {#s4n} ----------------------------- Quantitative analysis of the junctional intensity of N-cadherin was performed in ImageJ software (NIH) with the line scan functions by drawing straight lines orthogonal to, and centred on, randomly selected homotypic junctions. Optical Z-stacks (0.5-μm intervals) were acquired to correct for cell heights and to focus on all junctions analysed. The maximum and minimum pixel intensities along the selected lines were recorded, and the minimum pixel intensities were subtracted from the maximum pixel intensities as background fluorescence. A total of 20--30 junctions were analysed for each individual experiment, and at least three independent experiments were performed. Statistical analysis {#s4o} -------------------- Statistical analysis was performed using GraphPad Prism. A value of *P*\<0.05 was considered significant. Student\'s *t*-test was used for pair-wise comparisons. One-way ANOVA followed by Holm-Sidak test was used for comparison between multiple groups. The authors thank Dr Guangpei Hou for assistance with cell harvest and maintenance. David Ngai and Marsel Lino provided DDR1b plasmid and advice on protocols for the plasmid and siRNA transfection experiments. **Competing interests** The authors declare no competing or financial interests. **Author contributions** Conceptualization: S.X., M.P.B.; Methodology: S.X., M.P.B.; Formal analysis: S.X., S.B.; Investigation: S.X., S.B.; Resources: M.P.B.; Data curation: S.X.; Writing - original draft: S.X.; Writing - review & editing: M.P.B.; Visualization: M.P.B.; Supervision: M.P.B.; Project administration: M.P.B.; Funding acquisition: M.P.B. **Funding** This work was supported by the Heart and Stroke Foundation of Canada \[G-15-0009032, G-18-0022067 to M.P.B.\] and the Canadian Institutes of Health Research \[MOP133592 to M.P.B.\]. S.X. was supported by the Ontario Graduate Scholarship, Peterborough K.M. Hunter Graduate Studentship, Heart & Stroke Lewar Centre of Excellence Studentship and Queen Elizabeth II Graduate Scholarship in Science and Technology.

Introduction ============ The ability for cells to organize their interior is ubiquitous across all domains of life. In bacteria, the ParA/MinD family of ATPases has been primarily studied for their ability to segregate genetic cargos, such as chromosomes and plasmids (reviews in [@msz308-B10]; [@msz308-B8]). Less studied are ParA/MinD family members implicated in the positioning of diverse protein complexes, including those involved in secretion ([@msz308-B99]; [@msz308-B71]), chemotaxis ([@msz308-B93]; [@msz308-B76]; [@msz308-B4]), conjugation ([@msz308-B6]), cell division ([@msz308-B74]; [@msz308-B57]), and cell motility ([@msz308-B103]; [@msz308-B47]), as well as protein-based bacterial microcompartments (BMCs), such as the carboxysome ([@msz308-B81]; [@msz308-B56]). For partitioning plasmids, the ParABS system is the best characterized to date. Mechanistically, ParA proteins dimerize and nonspecifically bind DNA in the presence of ATP ([@msz308-B50]; [@msz308-B37]; [@msz308-B19]; [@msz308-B97]). Subsequently, ParB, which binds around a centromere-like site on the plasmid, *parS*, displaces ParA from the nucleoid (potentially through ATP hydrolysis stimulation) ([@msz308-B22]; [@msz308-B29]; [@msz308-B23]; [@msz308-B12]; [@msz308-B14]). ParA then recycles its nucleotide and rebinds the nucleoid at a random location. The local formation of ParA depletion zones around individual ParB-bound plasmids results in a global break in ParA symmetry along the nucleoid that ParB-bound plasmids utilize to migrate in a directed and persistent manner toward increased concentrations of ParA on the nucleoid ([@msz308-B1]; [@msz308-B36]; [@msz308-B40]; [@msz308-B98]; [@msz308-B39]); a recursive mechanism that ensures equidistant plasmid positioning and faithful plasmid inheritance following cell division. In our recent study, we identified a new self-organizing ParA-type ATPase system, McdAB, which is responsible for equidistantly positioning the carbon-fixing organelles of cyanobacteria, carboxysomes ([@msz308-B56]). Carboxysomes are essential for photoautotrophic growth of cyanobacteria. Since O~2~ competes with CO~2~ as a substrate for ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO), the encapsulation of RuBisCO and carbonic anhydrase within a selectively permeable protein shell (the carboxysome) is necessary for generating the high CO~2~ environment needed to drive internal RuBisCO reactions toward the Calvin--Benson--Bassham cycle (CO~2~ substrate) and away from the wasteful process of photorespiration (O~2~ substrate) ([fig. 1*A* and *B*](#msz308-F1){ref-type="fig"}) (reviewed in [@msz308-B104]). Comprised thousands of proteins ([@msz308-B91]), carboxysomes were once thought to be completely paracrystalline ([@msz308-B43]). However, recent reports have challenged this idea by showing that carboxysome formation involves the process of liquid--liquid phase separation (LLPS) ([@msz308-B70]; [@msz308-B101]); a process that describes the ability of proteins to spontaneously demix into dilute and dense phases that resemble liquid droplets (reviewed in [@msz308-B3]). Through these mechanisms, carboxysomes contribute to \>25% of global carbon-fixation through atmospheric CO~2~ assimilation ([@msz308-B73]). Using the model rod-shaped cyanobacterium *Synechococcus elongatus* PCC 7942, we identified a small novel protein, McdB, responsible for emergent oscillatory patterning of ATP-bound McdA on the nucleoid ([@msz308-B56]). Although McdB had no identifiable sequence similarities with any known ParB-family members, we found that McdB localized to carboxysomes through multiple shell protein interactions and removed McdA from the nucleoid in their vicinity ([@msz308-B56]); observations that are analogous to ParA patterning following ParB binding to the *parS* site of plasmids. Thus, like plasmids, we showed that carboxysomes utilize a Brownian-ratchet-based mechanism whereby McdB-bound carboxysome motion occurs in a directed and persistent manner toward increased concentrations of McdA on the nucleoid ([@msz308-B98]; [@msz308-B39]; [@msz308-B56]) ([fig. 1*C*](#msz308-F1){ref-type="fig"}). {#msz308-F1} Several important questions remained following our study. First, it was unclear whether the McdAB system was widespread among cyanobacteria, especially given that cyanobacteria are an incredibly diverse and widely distributed phylum of bacteria that display complex morphologies, including: 1) unicellular, 2) baeocystous, 3) filamentous, 4) heterocystous, and 5) ramified ([@msz308-B86]). Cyanobacteria are also taxonomically classified as α or β depending on the form of RuBisCO they encapsulate; β-cyanobacteria encapsulate form 1B RuBisCO in β-carboxysomes and α-cyanobacteria encapsulate form 1 A RuBisCO in α-carboxysomes ([@msz308-B73]). These two types of carboxysomes are structurally distinct, and α-carboxysomes are thought to have been horizontally transferred to cyanobacteria from chemoautotrophic Proteobacteria ([@msz308-B73]). Thus, it is not obvious whether α-cyanobacteria possess a McdAB system and whether it would share similarities to that of the β-cyanobacterial McdAB system. Second, we noted in our previous study that McdA surprisingly lacked the signature lysine residue in the Walker A ATP-binding motif---a critical residue that defines the ParA family of ATPases. Although BLASTp results for McdA identified additional McdA-like sequences in other cyanobacteria that also lacked this lysine residue, the results were few and many were plasmid encoded. Therefore, it was not clear whether the McdAB system was unique to *S. elongatus* or if more than one type of McdAB system evolved in cyanobacteria. Lastly, it was not obvious why reliable results from BLASTp for *S. elongatus* McdB could not be obtained. Interestingly, however, our neighborhood analysis around the carboxysome operon in the distantly related cyanobacterium *Gloeobacter kilaueensis* JS1 identified a ParA-type ATPase that possessed the signature lysine residue absent in *S. elongatus* McdA and a small downstream coding sequence. The protein product of this small gene loaded onto *S. elongatus* carboxysomes, which was surprising because it had no sequence homology to *S. elongatus* McdB ([@msz308-B56]). The findings suggested a more rigorous gene neighborhood analysis was necessary to identify other McdAB systems. Here, we performed a neighborhood analysis for McdAB-like sequences encoded near carboxysome components in 537 cyanobacterial genomes (205 α-cyanobacterial and 332 β-cyanobacterial). Our analysis revealed that the McdAB system is widespread among β-cyanobacteria and is surprisingly absent in α-cyanobacteria. Across these β-cyanobacteria, McdAB were found near carboxysome components in ∼31% of genomes and near the minor shell components CcmK3 and CcmK4 in ∼25% of genomes; suggesting a strong functional association. Our analysis also shows that there are two types of McdAB systems, which we term types 1 and 2. Type 1 systems, like that of *S. elongatus*, consist of a McdA without the signature lysine residue and a McdB with a predicted C-terminal coiled coil. Type 2 systems alternatively, which were found to be the most abundant (\>98% of genomes), consist of a McdA ATPase with the signature lysine residue present, and a McdB with a predicted central coiled coil. The low representation of the *S. elongatus* Type 1 McdAB system suggested a possible unique origin. In support of this hypothesis, our cyanobacterial phylogeny strongly suggests that *S. elongatus* is immediately adjacent to α-cyanobacteria, which lack the McdAB system, and shares a common ancestor with cyanobacteria where a McdAB system could not be identified. Lastly, when comparing identified McdB proteins, little to no sequence homology was observed. However, all shared the known hallmarks of proteins capable of LLPS: 1) intrinsic disorder, 2) biased amino acid compositions, 3) low hydrophobicity, and 4) extreme multivalency. We purified representatives of both McdB types and found that all formed phase-separated droplets. Moreover, we found that the LLPS activity of *S. elongatus* McdB was highly influenced by pH; forming droplets in a pH range that correlates to the proposed acidic carboxysome environment and not forming droplets in a pH range that correlates to the basic cyanobacterial cytosol. To our knowledge, this is the first experimental demonstration of LLPS behavior in a ParA-type ATPase partner protein and is an interesting finding given that carboxysome formation has recently been shown to involve LLPS ([@msz308-B70]; [@msz308-B101]). Collectively, these results have broad implications for understanding carboxysome formation, homeostasis, positioning, and function. Results ======= Finding Homologs of *S. elongatus* McdAB ---------------------------------------- To explore whether the McdAB system is widespread among cyanobacteria, we began our analysis by performing BLASTp searches for *S. elongatus* McdA (Synpcc7942_1833) and McdB (Synpcc7942_1834). We previously reported that the deviant Walker A box of *S. elongatu*s McdA lacks the signature lysine residue that defines the ParA family of ATPases (**K**GGXXKS/T) ([@msz308-B56]). The serine substitution at this position in McdA (**S**GGQGKT) may underlie the unusually high ATPase activity of McdA, which displays a maximum specific activity that is roughly two orders of magnitude greater than that of other well-studied ParA systems ([@msz308-B2]; [@msz308-B97]; [@msz308-B56]). Our BLASTp search results for *S. elongatus* McdA returned only a few McdA-like sequences where the signature lysine residue was replaced with a serine ([supplementary fig. S1*A*](https://academic.oup.com/mbe/article-lookup/doi/10.1093/molbev/msz308#supplementary-data), [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Four of these hits were nearly identical to *S. elongatus* McdA (*Synechococcus elongatus* PCC 6301, *Synechococcus elongatus* PCC 11801, *Synechococcus elongatus* UTEX 3055, and *Synechococcus* sp. UTEX 2973). Recent crystallization of a plasmid-encoded McdA-like protein from the cyanobacterium *Cyanothece* sp. PCC 7424 (PCC7424_5529) showed that a lysine residue in the middle of the protein (K151) functioned analogously to the signature lysine residue of classical ParA-type ATPases found within the Walker A box; K151 was found to contact the oxygen between the β and γ phosphates of ATP and promote formation of a sandwich dimer ([@msz308-B83]). Consistent with this finding, our McdA-like BLASTp hits also possessed the K151 residue ([supplementary fig. S1*A*](https://academic.oup.com/mbe/article-lookup/doi/10.1093/molbev/msz308#supplementary-data), [Supplementary Material](#sup1){ref-type="supplementary-material"} online). BLASTp results for *S. elongatus* McdB were extremely poor. However, when performing a gene neighborhood analysis of the newly identified McdA-like sequences above, a short open reading frame was identified immediately downstream. Although these sequences shared high similarity among themselves, they largely differed from *S. elongatus* McdB outside the first ∼25 amino acids ([supplementary fig. S1*B*](https://academic.oup.com/mbe/article-lookup/doi/10.1093/molbev/msz308#supplementary-data), [Supplementary Material](#sup1){ref-type="supplementary-material"} online). ParB proteins possess a small-charged region at their N-terminus responsible for interacting with its cognate ParA protein and stimulating its ATPase activity ([@msz308-B72]; [@msz308-B75]; [@msz308-B9]; [@msz308-B2]). Consistent with this observation, the recent crystallization and analysis of the *Cyanothece* sp. PCC 7424 McdB-like protein (PCC7424_5530) revealed that the N-terminus (AA 1--150) mediated interaction with the *Cyanothece* sp. PCC 7424 McdA-like protein ([@msz308-B83]). Given the similarity between *S. elongatus* McdA and the McdA-like sequences, we identified by BLASTp, it is not surprising that the N-terminal region of *S. elongatus* McdB and the McdB-like sequences identified by BLASTp were highly conserved in this region ([supplementary fig. S1*B*](https://academic.oup.com/mbe/article-lookup/doi/10.1093/molbev/msz308#supplementary-data), [Supplementary Material](#sup1){ref-type="supplementary-material"} online). However, outside this N-terminal region, it was not obvious why *S. elongatus* McdB greatly differed from our newly identified McdB-like sequences. The structure of the McdB-like protein from *Cyanothece* sp. PCC 7424 possessed two small central helices followed by a large C-terminal coiled-coil region ([@msz308-B83]). Although *S. elongatus* McdB is predicted to also possess a C-terminal coiled coil ([@msz308-B56]), *S. elongatus* McdB possesses a large glutamine-rich region and several extensions and gaps relative to the McdB-like BLASTp hits ([supplementary fig. S1*B*](https://academic.oup.com/mbe/article-lookup/doi/10.1093/molbev/msz308#supplementary-data), [Supplementary Material](#sup1){ref-type="supplementary-material"} online); potentially due to recognition of different cargos (i.e., carboxysomes vs. plasmids). Indeed, many of these McdAB-like proteins were plasmid encoded and were also the sole ParA-type system on these plasmids. Collectively, given that: 1) several of our McdAB hits were plasmid encoded, suggesting that these proteins might function in plasmid partitioning instead of carboxysome positioning, 2) only a few homologs were identified among the hundreds of available cyanobacterial genomes, 3) the McdB-like protein in *G. kilaueensis* JS1 that we previously showed loaded onto *S. elongatus* carboxysomes ([@msz308-B56]) followed a ParA-type ATPase that possessed the signature lysine residue and lacked K151, and 4) *G. kilaueensis* JS1 McdAB were encoded within the carboxysome operon, we reasoned that these proteins identified by BLASTp above, as well as the crystallized *Cyanothece* sp. PCC 7424 McdAB-like proteins ([@msz308-B83]), were not true McdAB carboxysome positioning systems. Therefore, an alternative more rigorous analysis was needed to identify McdAB homologs across cyanobacteria. McdAB Cooccur with Carboxysome Components in Many Cyanobacteria --------------------------------------------------------------- The increased availability of genomic data, in combination with targeted bioinformatic analyses, has resulted in a wealth of data suggesting that the vast majority of BMC-related genes tend to form operons with their respective encapsulated enzymes. For example, the most recent bioinformatic efforts to identify novel BMCs among bacteria resulted in the identification of 23 different types of BMCs across 23 different bacterial phyla ([@msz308-B7]). This study showed that neighborhood analyses are a powerful and reliable tool for the identification of new factors involved with BMC function. Therefore, as BLASTp was an unreliable method to identify new candidate McdA and McdB proteins, we next performed neighborhood analysis for McdAB-like sequences that clustered near carboxysome-related components across 537 cyanobacterial genomes (205 α-cyanobacterial and 332 β-cyanobacterial). We defined candidate McdA sequences as having a deviant Walker A motif with global homology to ParA-type ATPases and candidate McdB sequences as the protein product of the open reading frame immediately following these McdA sequences. We note that additional carboxysome-related genes are often located at distant loci from the main *ccm* (Carbon Concentrating Mechanism) operon ([fig. 1*D*](#msz308-F1){ref-type="fig"}) ([@msz308-B7]; [@msz308-B88]). Using these criteria, our analysis identified 75 examples of McdAB clustering near carboxysome components (genomic distances of 2209 ± 1679 bp or 2 ± 1.6 genes up- or downstream of the carboxysome component). Of these, McdAB-like sequences clustered near the *ccm* operon in 15 species and near the distant loci of the minor shell proteins CcmK3 and CcmK4 in 41 species ([fig. 1*E*](#msz308-F1){ref-type="fig"}). Moreover, in 16 species, only a McdB-like sequence clustered with CcmK3 and CcmK4, and in three species, a McdB-like sequence was found near either the RuBisCO transcription factor (RbcR) or carbonic anhydrase (CcaA) ([fig. 1*E*](#msz308-F1){ref-type="fig"}). In both these instances, a McdA-like sequence was not identified, suggesting McdB proteins in general may have additional functions independent of its role in positioning carboxysomes via interaction with McdA. To identify additional McdA and McdB proteins among cyanobacteria that did not cluster near carboxysome components, we needed to establish a new criterion to aid our search. To accomplish this, we generated multiple sequence alignments for the McdA and McdB proteins that clustered near carboxysome components to identify highly conserved regions and/or features among these proteins ([supplementary fig. S2*A*](https://academic.oup.com/mbe/article-lookup/doi/10.1093/molbev/msz308#supplementary-data) and *B*, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Among the McdA proteins, the first notable features we found were that they could range in size from 210 to 228 amino acids and that each of these proteins possessed the signature lysine residue within the Walker A motif ([fig. 1*F*](#msz308-F1){ref-type="fig"} and [supplementary fig. S2*A*](https://academic.oup.com/mbe/article-lookup/doi/10.1093/molbev/msz308#supplementary-data), [Supplementary Material](#sup1){ref-type="supplementary-material"} online), unlike that of *S. elongatus* McdA which instead possesses a serine residue ([@msz308-B56]). Secondly, two additional threonine residues followed the last threonine residue of the Walker A motif ([supplementary fig. S2*A*](https://academic.oup.com/mbe/article-lookup/doi/10.1093/molbev/msz308#supplementary-data), [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Lastly, two tryptophan residues adjacent to the Walker A' motif (1---gray cylinder) and three additional regions toward the last half of these proteins (2---yellow cylinder, 3---brown cylinder, and 4---green cylinder) were present among all McdA proteins encoded near carboxysome components, but not among classical ParA proteins or *S. elongatus* McdA ([fig. 1*F*](#msz308-F1){ref-type="fig"} and [supplementary fig. S2*A*](https://academic.oup.com/mbe/article-lookup/doi/10.1093/molbev/msz308#supplementary-data), [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Unlike these McdA proteins, conservation among McdB proteins near carboxysome components was extremely low; partly due to the surprising observation that McdB proteins ranged in size from 150 to 326 amino acids. However, several shared features existed that allowed us to establish criteria to define an McdB protein. First, we found that all McdB proteins were largely polar and biased in amino acid composition ([fig. 1*G*](#msz308-F1){ref-type="fig"}). Second, Predictor of Natural Disordered Regions (PONDR) predicted McdB proteins to be highly disordered (average disorder = 68%) ([fig. 1*H*](#msz308-F1){ref-type="fig"}) ([@msz308-B79]), which explains our poor BLASTp results here and prior ([@msz308-B56]). Lastly, all identified McdB proteins possessed a highly charged N-terminal region, a predicted coiled coil, a glutamine-rich region centrally located within the predicted coiled coils, and a tryptophan residue within the last four residues of each sequence ([fig. 1*I*](#msz308-F1){ref-type="fig"} and [supplementary fig. S2*B*](https://academic.oup.com/mbe/article-lookup/doi/10.1093/molbev/msz308#supplementary-data), [Supplementary Material](#sup1){ref-type="supplementary-material"} online). The bioinformatics analysis suggests all McdB proteins we identified share known hallmarks of phase-separating proteins: largely polar, low amino acid complexity, low hydrophobicity, and intrinsic disorder. The McdAB System Is Widespread among β-Cyanobacteria ---------------------------------------------------- Using our new criteria for McdAB proteins established from McdAB proteins that cluster near carboxysome components, we performed an additional manual search within α- and β-cyanobacterial genomes where McdAB proteins were not previously identified. This approach was necessary given that cyanobacteria possess enormously diverse genomic architectures and relatively poor operon structure in comparison to other well-studied bacteria ([@msz308-B11]). Not a single McdAB system that fit within our criteria was identified among the 205 α-cyanobacterial genomes we analyzed. Intriguingly, many α-cyanobacteria did not possess a single homolog of a ParA-type ATPase family member in general. However, in β-cyanobacteria, we identified several McdA and McdB sequences that fit within our criteria. In total, we found 248 McdA (∼75% of β-cyanobacterial genomes) and 285 McdB sequences (∼86% β-cyanobacterial genomes) from 332 β-cyanobacterial genomes analyzed ([fig. 2*A*](#msz308-F2){ref-type="fig"} and [supplementary table S1](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). We found that ∼14% β-cyanobacterial genomes lacked the McdAB system. However, we note that 16% of β-cyanobacterial genomes without candidate McdA, 8% of β-cyanobacterial genomes without McdB, and 9% of β-cyanobacterial genomes without McdA and McdB were not fully assembled genomes (still in scaffolds or contigs). Collectively, we found that McdA proteins ranged in lengths from 202 to 253 amino acids, while the lengths of McdB proteins varied far greater, ranging in from 132 to 394 amino acids. In 199 β-cyanobacterial genomes, McdAB were found next to each other genomically (∼82% of β-cyanobacterial genomes with McdAB) ([fig. 2*A*](#msz308-F2){ref-type="fig"}). Moreover, McdA and/or McdB clustered near carboxysome-related components in 75 genomes (∼31% of β-cyanobacterial genomes with McdAB) and clustered near the distant locus for the minor shell components CcmK3 and CcmK4 in 60 genomes (∼25% of β-cyanobacterial genomes with McdAB). This finding suggests a strong functional association between McdB and CcmK3/CcmK4, which is consistent with our previous bacterial two-hybrid results showing a strong interaction between McdB and these two shell components of the carboxysome ([@msz308-B56]). Lastly, McdAB sequences were found not only among all major taxonomic orders of β-cyanobacteria ([fig. 2*B*](#msz308-F2){ref-type="fig"}) but were also found in genomes across all five major morphologies ([fig. 2*C*](#msz308-F2){ref-type="fig"}). {#msz308-F2} Two Distinct McdAB Systems Exist in β-Cyanobacteria --------------------------------------------------- While conducting this bioinformatic analysis, we found that the vast majority (∼98%) of newly identified McdAB proteins was quite distinct from *S. elongatus* McdAB and only four additional cyanobacterial species, which are closely related to *S. elongatus*, had a similar McdAB system to that of *S. elongatus*. Therefore, we designated the two McdAB systems as Type 1, those similar to *S. elongatus* McdAB, and Type 2, those similar to the newly identified McdAB's in this study ([fig. 2*D* and *E*](#msz308-F2){ref-type="fig"}). Several features distinguish McdA as Type 1 or Type 2. For example, Type 1 McdA proteins lack the signature lysine residue within the Walker A motif, whereas Type 2 McdA's possess this lysine residue like classical ParA family members. The double tryptophan found in all Type 2 McdA proteins (\*) is instead a phenylalanine and a tyrosine in Type 1 McdAs ([fig. 2*D*](#msz308-F2){ref-type="fig"}). Additionally, a small ∼7 amino acid insertion (\*\*) immediately following the double tryptophan as well as a downstream phenylalanine are highly conserved in Type 2 McdA's, but absent in Type 1 ([fig. 2*D*](#msz308-F2){ref-type="fig"}). Most notably, Type 1 McdA's have a long extension between the Walker A' and Walker B motifs that is not present in Type 2 McdA's ([fig. 2*D*](#msz308-F2){ref-type="fig"}). Lastly, the Walker B motif of Type 1 McdA's has a conserved phenylalanine and cysteine that are instead small hydrophobic residues in Type 2 McdA's. We also found the two types of McdB significantly differed. For example, Type 1 McdB's have a predicted coiled coil at the C-terminus, while all Type 2 McdB's have a predicted coiled coil near the middle of the protein ([fig. 2*E*](#msz308-F2){ref-type="fig"}). Moreover, the charged N-terminal region appeared inverted among Type 1 and Type 2 McdB's, where negatively charged residues proceeded positively charged residues in Type 1 McdB's and positively charged residues proceeded negatively charged residues in Type 2 ([fig. 2*E*](#msz308-F2){ref-type="fig"}). Interestingly, although both types have a central glutamine-rich region, Type 1 McdB's are almost 10% more enriched for glutamine than Type 2 McdB's ([fig. 2*E*](#msz308-F2){ref-type="fig"}). Despite these differences, we found that both types possessed an invariant C-terminal tryptophan ([fig. 2*E*](#msz308-F2){ref-type="fig"}). Taken together, these features strongly suggest that two distinct McdAB systems are present among cyanobacteria and that the Type 2 system identified in this study is the most widely conserved. A Unique Origin for Type 1 McdAB Systems? ----------------------------------------- Since we found that α-cyanobacteria lack McdAB and that Type 1 systems were only present in ∼2% of cyanobacteria studied, we next sought to better understand the phylogenetic placement of Type 1 and Type 2 McdAB systems and determine whether the Type 1 McdAB system could have evolved independent of Type 2 systems. To explore this, we generated a Maximum-Likelihood tree inferred using a concatenation of the proteins DnaG, RplA, RplB, RplC, RplD, and RplE, which have recently been shown to be good-markers for cyanobacterial phylogenetic reconstruction ([@msz308-B38]). Notably, we found that the Type 2 McdAB system was widespread among cyanobacteria including our outgroup taxonomic order Gloeobacterales, considered the "most primitive" order among living cyanobacteria due to a lack of thylakoid membranes ([@msz308-B77]; [@msz308-B34]; [@msz308-B66]). This was an important finding that suggested the Type 2 McdAB system was present at the earliest known point of cyanobacterial evolution. Alternatively, the Type 1 McdAB system was only present in cyanobacterial species adjacent to α-cyanobacteria, which we found to lack the McdAB system, and shared a common ancestor with four β-cyanobacterial species where we could not identify an McdAB system of either type ([fig. 3](#msz308-F3){ref-type="fig"}). This was a surprising result that suggested a more recent origin for the Type 1 McdAB system. Our phylogenic inference also suggested multiple independent losses of the McdAB system with no obvious shared characteristics among species (i.e., different taxonomic orders and morphologies) ([fig. 3](#msz308-F3){ref-type="fig"}). Together, these results suggest that the Type 2 McdAB system is more ancestral and that Type 1 evolved independently of the Type 2 McdAB system, possibly via horizontal gene transfer. {#msz308-F3} McdB Proteins Possess the Hallmarks for Phase Separation -------------------------------------------------------- Why McdB proteins were so highly diverged in primary sequence was unclear. For example, while Type 2 McdA's have high mean amino acid similarity (∼64%), we found that the mean amino acid similarity among Type 2 McdB's was extremely low (∼20%) ([fig. 4*A* and *B*](#msz308-F4){ref-type="fig"}). Partially contributing to this diversity among McdB proteins, we found that the lengths of the central coiled coils and the N- and C-terminal extensions from the coiled coil greatly varied in length ([fig. 4*C*--*E*](#msz308-F4){ref-type="fig"}). N-terminal extensions were 60 ± 23 amino acids, central coiled coils were 93 ± 25 amino acids, and the C-terminal extensions were 56 ± 16 amino acids in length ([fig. 4*C*--*E*](#msz308-F4){ref-type="fig"}). ProtScale prediction (Scale: [@msz308-B48]) determined that both Type 1 and Type 2 McdB's were largely hydrophilic, and that Type 1 McdB's possessed a small hydrophobic patch toward the middle of the protein ([fig. 4*F*](#msz308-F4){ref-type="fig"}). Consistent with this finding, all three Type 2 McdB domains were biased toward hydrophilic amino acids and had obvious repetitive sequences that were rarely similar from one McdB protein to the next ([fig. 4*G*](#msz308-F4){ref-type="fig"}). {#msz308-F4} Consistent with the low hydrophobicity of McdB proteins, PONDR predicted that McdB proteins were largely disordered ([fig. 4*H*](#msz308-F4){ref-type="fig"}). Indeed, we found that the mean disorder of McdB proteins was 64%; PONDR predicted that some McdB proteins were as high as 96% disordered ([fig. 4*H*](#msz308-F4){ref-type="fig"}). Interestingly, while the consensus sequence for Type 1 McdB's was predicted to be largely ordered in a central region that partially corresponded to a hydrophobic patch and that the bulk of disorder was positioned toward the N-terminal region of the proteins ([fig. 4*F* and *I*](#msz308-F4){ref-type="fig"}), the consensus sequence for Type 2 McdB's was predicted to possess much greater disorder in the N- and C-terminal extensions from the central coiled coils and predicted to be ordered in a region that corresponded to the coiled coils ([fig. 4*I*](#msz308-F4){ref-type="fig"}). Collectively, these features were largely responsible for the diversity among McdB proteins. McdB Undergoes LLPS In Vitro ---------------------------- The shared hallmarks of McdB proteins include: 1) large regions of low complexity that greatly vary in length, 2) intrinsically disordered regions, 3) repetitive and biased amino acid compositions, 4) low hydrophobicity, and 5) extreme multivalency. All are features that are characteristic of proteins shown to undergo LLPS ([@msz308-B44]; [@msz308-B68]; [@msz308-B27]; [@msz308-B53]; [@msz308-B63]; [@msz308-B67]; [@msz308-B96]; [@msz308-B42]). LLPS refers to the ability of an otherwise homogeneous solution of macromolecules (e.g., proteins or nucleic acids) to spontaneously demix into a dilute phase and dense phase that resembles water droplets ([fig. 5*A*](#msz308-F5){ref-type="fig"}) (reviewed in [@msz308-B3]). The two liquid-like phases coexist and can in some cases undergo further reversible phase transitions to form gels and solids depending on solution conditions (i.e., macromolecule concentration, pH, salt type and concentration, and temperature); some transitions are irreversible under physiological conditions, such as amyloid-like fibers ([@msz308-B3]). This phenomenon has received increased attention in cell biology due to the discovery that this process accounts for the compartmentalization of biological functions through the formation of membraneless organelles, also recently termed biomolecular condensates ([@msz308-B87]). {#msz308-F5} Given the general features of the McdB proteins, we identified in this study ([fig. 4](#msz308-F4){ref-type="fig"}), we next wanted to test the hypothesis that *S. elongatus* McdB could undergo phase separation. We expressed and purified *S. elongatus* McdB fused at its N-terminus with a 6×-histidine tag for affinity purification and a SUMO tag to form His--SUMO--McdB. A SUMO-tag has been shown to inhibit the LLPS activity of proteins, thus simplifying the purification protocol ([@msz308-B84]). In addition, cleavage of the SUMO-tag by the addition of the Ulp1 protease allows for precise, time-controlled induction of LLPS. Under physiologically relevant pH and salt concentration (pH 7.5, 150 mM KCl), native *S. elongatus* McdB displayed phase-separation upon the addition of Ulp1 ([fig. 5*B*](#msz308-F5){ref-type="fig"}). The local environment of the carboxysome in *S. elongatus* cells has been suggested to be more acidic relative to the more basic cytosol because of bicarbonate accumulation prior to its conversion to CO~2~ by carbonic anhydrase ([@msz308-B58]). Given that McdB strongly colocalizes with carboxysomes in vivo ([@msz308-B56]), we assayed the LLPS activity of McdB across a broad pH range (5.5--8.5) ([fig. 5*B*](#msz308-F5){ref-type="fig"}). We found that McdB (30 µM) readily formed droplets following cleavage of the His--SUMO tag across a pH range of 5.5--7.5 ([fig. 5*B*](#msz308-F5){ref-type="fig"}). Between pH 5.5 and 6.5, droplets were less spherical in shape and did not readily fuse, which is suggestive of a gel-state ([@msz308-B3]). However, between pH 6.5--7.5, McdB droplets displayed liquid-like behaviors such as the fusion of two adjacent droplets into one ([fig. 5*B*](#msz308-F5){ref-type="fig"}). At or above pH 8, McdB did not undergo LLPS ([fig. 5*B*](#msz308-F5){ref-type="fig"}). Even with significantly higher concentrations of McdB (167 µM), McdB still displayed liquid-like behavior with frequent fusion events ([fig. 5*C* and *D*](#msz308-F5){ref-type="fig"} and [supplementary videos 1 and 2](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). McdB formed droplets at concentrations as low as 1 µM, but the droplets were very small and difficult to image due to their rapid diffusion in the observation well. Complete cleavage of the His--SUMO tag was verified across the assayed pH range ([supplementary fig. S3](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). The data suggest that the acidic nature of the carboxysome, relative to the cytosol of *S. elongatus* at pH 8 ([@msz308-B58]), may facilitate local phase-separation of McdB in the vicinity of carboxysomes. Both McdB types share the same hallmark features for LLPS at the amino acid sequence level. However, Type 2 McdBs differ significantly in coiled-coil location, charge distribution, and amino acid sequence length. Therefore, we next assayed whether Type 2 McdB proteins also displayed LLPS activity, in spite of these differences. As with *S. elongatus* McdB, we expressed and purified three Type 2 McdB proteins fused at their N-termini with a 6×-histidine and SUMO tag. These three Type 2 McdB proteins were strategically chosen for their broad phylogenetic distribution and presence in morphologically diverse cyanobacteria---*Gloeobacter kilaueensis* JS1, *Fremyella diplosiphon* NIES-3275, and *Fischerella* sp. PCC 9431 McdB proteins ([fig. 3](#msz308-F3){ref-type="fig"}). For all three Type 2 McdB proteins tested, we observed phase separation into droplets in vitro ([fig. 5*E*--*G*](#msz308-F5){ref-type="fig"}). Therefore, we conclude that LLPS activity is likely a universal property in both McdB types identified here. To our knowledge, this is the first direct demonstration of LLPS behavior of a ParA partner protein. Charge Distribution May Contribute to McdB Phase Separation ----------------------------------------------------------- Multivalency (charge distribution in the primary amino acid sequence) is an important feature contributing to LLPS ([@msz308-B52]; [@msz308-B35]). For example, LLPS among the most well-studied systems involves interaction between positively charged residues within an intrinsically disordered region of proteins and the negatively charged backbone of DNA or RNA ([@msz308-B3]). However, we previously showed that McdB does not interact with DNA, but instead interacts with the carboxysome shell proteins CcmK2, CcmK3, CcmK4, CcmL, and CcmO ([@msz308-B56]). We asked where the multivalency within our system could arise. Although shell proteins have surface accessible acidic and basic residue patches that McdB could interact with, shell proteins are not intrinsically disordered nor are the acidic and basic regions of all shell proteins that McdB interacts with similar ([@msz308-B88]). Therefore, given that *S. elongatus* McdB could robustly phase separate on its own, without the addition of any interacting partner (i.e., a carboxysome shell component), we next set out to determine whether Type 1 and Type 2 McdB proteins were intrinsically multivalent. Analysis of the single N-terminal extension from the C-terminal coiled coil of *S. elongatus* Type 1 McdB revealed that the isoelectric point of roughly the first half of the extension was ∼8.5 (amino acids 1--50) and the second half was ∼4.3 (amino acids 51--108). For Type 2 McdB proteins that have N- and C-terminal extensions from a central coiled coil, the average isoelectric point of both extensions combined was essentially neutral (6.9 ± 1.6) ([fig. 5*H*](#msz308-F5){ref-type="fig"}). Intriguingly, however, we found that the average isoelectric point of the vast majority of N- and C-terminal extensions was either negatively- or positively charged ([fig. 5*I* and *J*](#msz308-F5){ref-type="fig"}). Moreover, when looking at Type 2 McdB proteins individually, we found that the vast majority (∼68%) had extremely inverted charges distributed between their N- and C-terminal extensions; ∼48% had a negatively charged N-terminus and positively charged C-terminus, and ∼20% had a positively charged N-terminus and negatively charged C-terminus ([fig. 5*K*](#msz308-F5){ref-type="fig"} and [supplementary fig. S4](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Roughly 26% of McdB proteins had both extensions share a similar charge with 13% having both positively charged extensions, and ∼13% having both negatively charged extensions ([fig. 5*K*](#msz308-F5){ref-type="fig"} and [supplementary fig. S4](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Lastly, only ∼6% of McdB proteins had one charged extension and one neutral extension; ∼4% had a neutral N-terminus and positively charged C-terminus and ∼2% had a negatively charged N-terminus and neutral C-terminus ([fig. 5*K*](#msz308-F5){ref-type="fig"} and [supplementary fig. S4](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Together, these results suggest that McdB is a polyampholyte with biphasic charge distributions within intrinsically disordered regions that may contribute to its LLPS activity. Discussion ========== McdAB Systems Are Widespread among β-Cyanobacteria -------------------------------------------------- Carboxysomes are essential protein-based organelles in cyanobacteria. To ensure that each daughter cell receives an optimum quantity of carboxysomes following cell division, the McdAB system equidistantly positions each carboxysome relative to one another throughout the cell ([fig. 1*C*](#msz308-F1){ref-type="fig"}). In our previous study, we were unable to determine how widespread the McdAB system was among cyanobacteria. Although we were able to identify one McdAB system outside of *S. elongatus*, in *Gloeobacter kilaueensis* JS1, and show that its McdB protein was capable of loading onto carboxysomes in *S. elongatus*, it was not obvious why BLASTp results for *S. elongatus* McdAB were so poor and why *G. kilaueensis* JS1 McdAB were so different than those from *S. elongatus* (22.5% and 18.4% pairwise identity to *S. elongatus* McdAB) ([@msz308-B56]). As McdAB were situated near the carboxysome operon in *G. kilaueensis* JS1, we reasoned that neighborhood analysis was a better method to identify McdAB proteins in other cyanobacteria. Indeed, we were able to identify 56 McdA and 75 McdB sequences that clustered near carboxysome components throughout cyanobacteria ([fig. 1*E*](#msz308-F1){ref-type="fig"} and [supplementary table S1](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Given this much larger sample size of McdAB proteins, we were able to identify highly conserved regions and features that permitted further identification of McdAB sequences that did not cluster near carboxysome components in other cyanobacteria. In total, we identified 248 McdA and 285 McdB sequences, showing that the McdAB system is widespread among β-cyanobacteria ([fig. 2*A*](#msz308-F2){ref-type="fig"}). These results have broader implications for understanding carboxysome positioning in other species. For example, cyanobacteria can display a wide range of morphologies from unicellular to specialized multicellular ([fig. 2*C*](#msz308-F2){ref-type="fig"}). Although carboxysomes are linearly spaced or hexagonally packed along the long axis of rod-shaped *S. elongatus* cells, our prior modeling suggested that McdAB positioning is influenced by cellular geometry, but still operates within spherical cyanobacterial cells to optimally space carboxysomes from one another ([@msz308-B56]). Likewise, many heterocystous and ramified cyanobacteria, while appearing filamentous, are a chain of connected smaller cells that individually display more rounded morphologies. Transmission electron micrographs from heterocystous cyanobacteria reveal that the McdAB system likely hexagonally packs carboxysomes ([@msz308-B64]). Therefore, given the ubiquity of McdAB across these morphologies ([fig. 2*C*](#msz308-F2){ref-type="fig"}), an understanding of how this system behaves within these unique cellular geometries is of profound interest. Lastly, we found that a large clade of cyanobacteria on the right side of our phylogenic tree appeared to lack the McdAB system ([fig. 3](#msz308-F3){ref-type="fig"}). Many of these species, including *Myxosarcina* sp. GI1, *Xenococcus* sp. PCC 7305, *Stanieria* sp. NIES-3757, *Stanieria cyanosphaera* PCC 7437, *Pleurocapsa* sp. CCALA 161, and *Pleurocapsa* sp. PCC 7319, display baeocystous morphologies and some species, including *Spirulina major* and *Spirulina subsalsa*, while filamentous, have extremely spiralized morphologies. One possible reason for the absence of the McdAB system in these cellular morphologies is that their chromosome organization, cell growth, and division strategies may be incompatible. Two Types of McdAB Systems Exist in β-Cyanobacteria --------------------------------------------------- Our analysis has identified two distinct McdAB systems in β-cyanobacteria. Type 2 systems are by far the most represented among cyanobacteria (\>98% of species). This finding in particular explains why BLASTp results for the Type 1 system of *S. elongatus* McdAB were so poor. There are several features that distinguish Type 1 and Type 2 McdAB systems. For example, Type 2 McdA proteins possess the "signature lysine" residue in the Walker A box ATP-binding motif; a feature that defines this family of ATPases as "ParA-like." This result was intriguing given that Type 1 McdA proteins, like that of *S. elongatus*, lack this signature lysine in the Walker A box. Instead, Type 1 McdA proteins likely use a lysine residue (K151) distantly located on the C-terminal half of the protein to fulfill this role ([@msz308-B83]) ([fig. 2*D*](#msz308-F2){ref-type="fig"}). In addition, exclusive to Type 1 McdA proteins is a large midprotein extension adjacent to the Walker B box. Overall, it is the Type 2 McdA protein that bears the closest similarity to ParA-family ATPases. The differences in amino acid sequence in and around the ATP-binding motifs suggest these two McdA types may also have differences in ATPase activity. Indeed, we previously found that the Type 1 McdA of *S. elongatus* has voracious ATPase activity and a relatively low ATPase stimulation by McdB ([@msz308-B56]). This is in stark contrast to ParA family members involved in bacterial DNA segregation, which have feeble intrinsic ATPase activities, but are strongly stimulated by their cognate ParB protein. Therefore, it will be interesting to compare the ATPase activities of the two McdA types, as well as the stimulatory activities of their respective McdB proteins. How Type 1 and Type 2 systems differ mechanistically, and whether these differences correlate to known factors influencing carboxysome positioning such as carboxysome size and quantity ([@msz308-B32]; [@msz308-B56]), genome copy number and volume ([@msz308-B33]), redox state of cells ([@msz308-B90], [@msz308-B91]), and McdAB stoichiometry ([@msz308-B56]), as well as unknown factors including McdB LLPS behavior and/or intracellular salt/pH conditions (see below), is important for understanding the diversity and evolution of the McdAB system among cyanobacteria. Although McdA proteins differed in certain regions, they were largely conserved compared with McdB proteins that displayed extreme diversity. The lack in conservation is presumably a result of relaxed selection, as any two closely related cyanobacterial species had McdB proteins that poorly aligned ([supplementary fig. S2*B*](https://academic.oup.com/mbe/article-lookup/doi/10.1093/molbev/msz308#supplementary-data), [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Despite this lack in conservation, we were able to identify several shared features among McdB proteins. All possessed a highly charged N-terminus, a glutamine-rich region toward the center of the protein, a coiled-coil domain, and an invariant tryptophan residue within the last four amino acids of the protein ([fig. 2*E*](#msz308-F2){ref-type="fig"}). For several ParA family ATPases, the N-terminus of the partner protein is necessary and sufficient for stimulation of ParA ATPase activity ([@msz308-B72]; [@msz308-B75]; [@msz308-B9]; [@msz308-B2]). Where tested, a critical N-terminal arginine or lysine residue stimulates the ATPase activity of the cognate ParA. The charged N-terminal domains of Type 1 versus Type 2 McdB proteins appeared inverted (location of positively- and negatively charged residues) ([fig. 2*E*](#msz308-F2){ref-type="fig"}). This finding suggests a possible difference in how Type 1 and Type 2 McdB proteins interact with their McdA partner, potentially through this N-terminal region. The central glutamine-rich region likely plays a role in LLPS activity of McdB proteins (see below). A striking difference between the two McdB types was the placement of the coiled coil. Type 1 McdB proteins have a coiled coil positioned at their C-terminus, whereas the coiled coil of Type 2 McdB proteins was centrally located ([supplementary fig. S4](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Coiled coils provide an interface for stable oligomerization into defined higher order species ([@msz308-B5]). Consistent with our sequence-based predictions, a recently solved structure of the coiled-coil domain in a Type 1 McdB-like protein showed stable dimerization in an antiparallel fashion ([@msz308-B83]). As for the invariant tryptophan at the C-terminus of all McdBs, it is intriguing that many proteins involved in the assembly of viral- or phage-capsids also encode for a tryptophan residue at their C-terminus ([@msz308-B25]; [@msz308-B95]; [@msz308-B46]; [@msz308-B60]). Given the capsid-like icosahedral structure of the carboxysome, it is attractive to speculate that the C-terminal tryptophan of McdB is involved in its association with the carboxysome shell. The functions and features we have shown thus far for Type 2 McdAB systems strongly suggest that carboxysomes are positioned by Type 2 systems by a very similar mechanism to that of Type 1 systems. Type 2 McdAB genes frequently neighbor carboxysome-related components, both McdB types have similar domain architectures and undergo LLPS, and we previously showed that a Type 2 McdB from *G. kilaueensis* JS1 is capable of loading onto *S. elongatus* carboxysomes ([@msz308-B56]). It was also recently shown that carboxysomes in *Fremyella diplosiphon*, which possesses a Type 2 McdAB system, are positioned similar to that shown for *S. elongatus* with a Type 1 McdAB system ([@msz308-B49]). However, it remains speculative at this point whether Type 2 McdAB systems are indeed responsible for positioning carboxysomes in vivo. A full characterization of a Type 2 McdAB system in its native host will be required to confirm their role in positioning carboxysomes, for elucidating potential biochemical differences among the two systems, and for studying how different cyanobacterial morphologies influence McdAB dynamics. The McdAB System Is Not Present in α-Cyanobacteria -------------------------------------------------- One of our most striking findings was that the McdAB system was completely absent in α-cyanobacteria. Although there are several differences between α- and β-carboxysomes, we expected this difference would largely be reflected in the putative carboxysome interaction domains of McdB proteins. We did not anticipate a complete absence of the system. Even more surprising, we found that not just McdA, but the entire ParA family of ATPases were significantly underrepresented among α-cyanobacteria. α-Cyanobacteria lack plasmids and have small genomes (most *Prochlorococcus* genomes are smaller than 2 Mb) ([@msz308-B82]). Therefore, one possible explanation for the lack of ParA-type ATPases is that many canonical partitioning systems might not be able to function in α-cyanobacteria. In addition, the absence of plasmids in α-cyanobacteria limits one of the sources of horizontal gene transfer for inheriting the McdAB system. Alternative positioning mechanisms may exist for α-carboxysomes in α-cyanobacteria. For example, α-carboxysomes are known to tightly interact with polyphosphate bodies ([@msz308-B41]). If polyphosphate bodies are strategically positioned in α-cyanobacteria, and α-carboxysomes tightly interact with these structures, this interaction would provide a "pilot-fish" mechanism by which both are equidistantly positioned throughout cells prior to cell division; making the McdAB system unnecessary. Another possibility is that α-carboxysomes are actively positioned by a polymer-based system, such as the actin-like ATPase MamK that mediates magnetosome positioning ([@msz308-B94]) or the tubulin-like GTPase TubZ that positions viral DNA ([@msz308-B69]). As α-carboxysomes are proposed to have originated in Proteobacteria and were horizontally transferred into cyanobacteria ([@msz308-B59]; [@msz308-B73]), it was surprising that the McdAB system was absent in α-cyanobacteria largely because many carboxysome operons of Proteobacteria encode a ParA-type ATPase followed by a small coding sequence. Whether this ParA-type ATPase and the small downstream coding sequence constitute the McdAB system for α-carboxysomes in Proteobacteria is of great interest for understanding the evolution, diversity, and mechanisms of carboxysome positioning across the bacterial world. The *S. elongatus* McdAB System Might Have a Unique Origin ---------------------------------------------------------- The phylogenic placement of *S. elongatus* and Type 1 McdAB systems relative to Type 2 systems suggest the latter is more ancestral. *Gloeobacter* *kilaueensis* JS1 is one of the most primitive species among living cyanobacteria due to the lack of thylakoid membranes ([@msz308-B77]; [@msz308-B34]; [@msz308-B66]). It possesses the widely distributed Type 2 system, suggesting that the Type 1 system of *S. elongatus* was acquired more recently. We obtained strong bootstrap support for *S. elongatus* sharing an immediate common ancestor with α-cyanobacteria, which lack the McdAB system, and four cyanobacteria where the McdAB system could not be identified ([fig. 3](#msz308-F3){ref-type="fig"}). Many Type 1 McdA homologs identified here by BLASTp are plasmid encoded (the sole ParA-type ATPase on the plasmid) and the N-terminal domain of Type 1 McdB's are nearly identical to the small coding sequences following these plasmid-encoded McdA-like proteins. Therefore, the most parsimonious hypothesis is that the plasmid encoded ParA-type system used to partition plasmids was horizontally transferred into *S. elongatus*, genomically integrated, and evolved to position carboxysomes. A similar evolutionary path would explain the positioning of a diversity of cargos by ParA family ATPases ([@msz308-B107]; [@msz308-B52]; [@msz308-B105]). As the Type 2 system was present at the earliest known point of cyanobacterial evolution, it is not clear why the Type 1 system would have had a selective advantage over the Type 2 system, presumably present in the ancestors of *S. elongatus*. It is also interesting that we were unable to identify the McdAB system in several modern cousins of *S. elongatus* ([fig. 3](#msz308-F3){ref-type="fig"}). An alternative hypothesis could be that this clade of β-cyanobacteria (which now includes modern *S. elongatus*) lost the Type 2 system, similar to the β-cyanobacterial clades found on the right side of our tree, and that *S. elongatus* then acquired the Type 1 system. As more cyanobacterial genomes are sequenced, how the Type 1 McdAB system evolved will become more apparent. McdB Is a Phase Separating Protein ---------------------------------- Our finding that McdB undergoes phase separation in vitro is intriguing on multiple fronts. First, two recent studies have demonstrated that α- and β-carboxysome formation involves LLPS ([@msz308-B70]; [@msz308-B101]). With β-carboxysomes, the protein CcmM aggregates RuBisCO to form a procarboxysome ([@msz308-B18]). CcmM exists in two forms: 1) full-length CcmM (58 kDa), which contains a carbonic anhydrase-like domain followed by three RuBisCO small subunit-like domains separated by flexible linkers, and 2) short-form CcmM (35 kDa), which lacks the carbonic anhydrase-like domain ([@msz308-B54]). Binding of small subunit-like domain regions of CcmM between RbcL dimers scaffolds RuBisCO molecules and induces LLPS ([@msz308-B101]). Likewise, in α-carboxysomes of the chemoautotrophic proteobacterium *Halothiobacillus neapolitanus*, N-terminal repeated motifs within the intrinsically disordered protein CsoS2 mediate interaction with RuBisCO and induce LLPS ([@msz308-B17]; [@msz308-B70]). The pyrenoid does not have a protein shell, but serves as the functional equivalent to carboxysomes in eukaryotic algae. One study of the pyrenoid showed that its RuBisCO matrix also exhibits liquid-like properties when interacting with the intrinsically disordered protein EPYC1 in *Chlamydomonas reinhardtii* ([@msz308-B28]). Collectively, these studies suggest that LLPS is a common feature underlying the formation of a RuBisCO matrix through interactions with intrinsically disordered proteins. Given that McdB is also an intrinsically disordered protein that undergoes LLPS ([fig. 5*B*--*G*](#msz308-F5){ref-type="fig"} and [supplementary videos 1 and 2](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online) and our previous data that increased levels of McdB drastically increase carboxysome size ([@msz308-B56]), it is intriguing to speculate that the LLPS activities of McdB and the carboxysome core are related and potentially influence each other. It is possible that the McdAB system not only positions carboxysomes in the cell but also maintains homeostasis of a liquid-like carboxysome core. Our results that *S. elongatus* McdB formed droplets across a broad pH range are informative. It has been previously suggested that a pH gradient exists between the cytosol and carboxysomes ([@msz308-B62]; [@msz308-B102]). Experimental evidence in *S. elongatus* showed that while the pH of the cytosol of *S. elongatus* is ∼8.4 under light conditions, the carboxysome is predicted to be more acidic (pH ∼6.0--7.0) ([@msz308-B58]). An acidic carboxysome has been suggested to increase the maximum carboxylation rate of RuBisCO and reduce the amount of HCO~3~^−^ uptake required to saturate RuBisCO ([@msz308-B58]). Our in vitro data suggest that McdB would be soluble in the cytosol (pH ≥8) ([fig. 5*B*](#msz308-F5){ref-type="fig"}) and undergo LLPS at or near acidic carboxysomes (pH ≤ 7.5) ([fig. 5*B*](#msz308-F5){ref-type="fig"}). However, dark adapted cells of *S. elongatus* have a cytosolic pH ∼7.3 ([@msz308-B58]), so McdB behavior might differ in cells under light and dark conditions. Moreover, cyanobacteria possess a biological circadian clock that precisely operates on the 24-h rotational period of the earth. Circadian rhythms primarily enable cyanobacteria to anticipate, adapt, and respond to daily light cycles by translating environmental cues into changes in gene expression (reviewed in [@msz308-B21]). In *S. elongatus*, oscillatory patterns of gene expression are driven by phosphorylation of the master output transcriptional regulator protein RpaA. Phosphorylation of RpaA has previously been shown to bind ∼170 promoters of the *S. elongatus* chromosome ([@msz308-B61]); one of which is the promoter for the *mcdAB* operon. Therefore, it will be interesting to explore the role of McdB LLPS activity at carboxysomes, and how circadian rhythms and light-dark conditions influence McdAB expression, dynamics, and function. Lastly, McdB is the first example of a ParA partner protein shown to have LLPS activity. It is attractive to speculate that under appropriate conditions, other ParA partner proteins exhibit similar LLPS behaviors that are critical to their in vivo function. The most obvious examples are ParB proteins that bind to a centromere-like DNA sequence called *parS* and are involved in plasmid and chromosome segregation in bacteria (reviewed in [@msz308-B100]; [@msz308-B10]; [@msz308-B13]). Through ChIP approaches, it is well known that thousands of ParB dimers associate with broad regions of DNA adjacent to the *parS* site, a phenomenon known as "spreading" ([@msz308-B78]; [@msz308-B65]; [@msz308-B15]; [@msz308-B80]; [@msz308-B24]). However, the actual structure of the ParB-DNA mega-complex remains unclear. In vivo, fluorescent fusions of ParB form a massive punctate focus at the location of a *parS* site on the chromosome or plasmid. These ParB foci segregate after DNA replication in a ParA-dependent manner and have also been shown to undergo rapid fission/fusion events, whereby ParB foci segregate and then snap back together ([@msz308-B85]). FRAP measurements of ParB exchange at these foci also suggest these complexes are highly dynamic ([@msz308-B24]). Together, the data are consistent with ParB forming a biomolecular condensate with its *parS* site and its adjacent DNA. LLPS activity of ParB proteins on *parS*-containing DNA substrates has yet to be directly observed in vitro. However, in silico models have proposed that the currently known affinities of ParB--ParB and ParB--DNA associations could induce the formation of a condensate ([@msz308-B16]; [@msz308-B80]; [@msz308-B24]). The vast majority of proteins capable of LLPS do so by interacting with DNA/RNA (reviewed in [@msz308-B3]). McdB, however, does not bind DNA and is capable of LLPS activity on its own. A future direction of research is aimed at understanding how the LLPS activity of McdB, and other ParA partner proteins, is involved in the spatial organization of their cognate cargos. McdB Charge Distribution Might Contribute to LLPS ------------------------------------------------- McdB proteins across evolutionary time possess many of the features that enable LLPS including intrinsic disorder, low hydrophobicity, biased amino acid compositions, and extreme multivalency (reviewed in [@msz308-B3]) ([fig. 4](#msz308-F4){ref-type="fig"}). When analyzing the multivalent properties of McdB proteins, we found that the vast majority was polyampholytes with biphasic charge distributions between their N- and C-terminal extensions flanking the coiled-coil domain ([fig. 5*K*](#msz308-F5){ref-type="fig"} and [supplementary fig. S4](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). The reason for such a shared feature is not obvious. Although genetic drift possibly accounts for the intrinsic disorder and lack of primary sequence conservation among McdB proteins, it is unlikely that genetic drift alone could account for the patterned charge distributions. Indeed, given that the vast majority of McdB proteins exhibit an inverted charged patterning along their N- and C-terminal extensions (∼68%, [fig. 5*K*](#msz308-F5){ref-type="fig"} and [supplementary fig. S4](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online), a more parsimonious explanation could be that the charged extensions are important for McdB associations with itself or with carboxysome shell proteins. Although not a true McdB protein, the structure of the coiled-coil domain of a plasmid-encoded McdB-like protein from the cyanobacterium *Cyanothece* sp. PCC 7424 (PCC7424_5530) demonstrated an antiparallel association to form a dimer ([@msz308-B83]). Parallel or antiparallel assembly of the coiled-coil domains of McdB would either align like-or oppositely charged extensions. The role of multivalency, charge patterning, and dimerization on the LLPS activity of McdB are interesting avenues of future investigation. *mcdA* and *mcdB* Genes Often Cluster with the Minor Carboxysome Shell Genes CcmK3 and CcmK4 -------------------------------------------------------------------------------------------- Of the *mcdA* and *mcdB* genes, we found near carboxysome components, the majority clustered near the minor shell genes *ccmK3* and *ccmK4* ([figs. 1*E* and 2*A*](#msz308-F1){ref-type="fig"}); two shell genes often found together, but at a different locus than the carboxysome operon ([@msz308-B88]). The finding is consistent with our prior results showing that McdB interacts with CcmK3 and CcmK4 in a bacterial two-hybrid system ([@msz308-B56]). Moreover, carboxysomes have been shown to cluster following the deletion of CcmK3 and CcmK4 in a manner that is reminiscent of carboxysome aggregation in our *ΔmcdB* strain ([@msz308-B73]; [@msz308-B56]). How McdB and CcmK3/CcmK4 interact is still an open question. A recent study showed that CcmK3 and CcmK4 can form homohexamers, as well as heterohexamers that further assemble into dodecamers under certain conditions ([@msz308-B89]). Metabolite channeling into and out of carboxysomes is believed to occur via the central pores of these hexameric shell proteins ([@msz308-B26]; [@msz308-B45]). As the pore formed by CcmK4 homohexamers differs from those formed by CcmK3/CcmK4 heterohexamers, it is proposed that homo- and hetero-hexamic species alter carboxysome permeability, thereby modulating metabolite channeling across the protein shell ([@msz308-B89]). Interestingly, CcmK3/CcmK4 dodecamer formation was found to be influenced by pH (relative dodecamer abundance of 36% at pH 7.0 and 3% at pH 8) ([@msz308-B89]). Likewise, we find here that McdB LLPS behavior is also greatly influenced by transitions in this pH range; soluble at a pH ≥8 and formed liquid droplets at a pH ≤ 7.5 ([fig. 5*B*](#msz308-F5){ref-type="fig"}). Placing these findings in the context of current models that propose the carboxysome as more acidic (pH ∼ 7.0) than the cytosol (pH ∼ 8.4) in *S. elongatus* ([@msz308-B62]; [@msz308-B102]; [@msz308-B58]), several important questions are revealed. How does McdB interact with CcmK3/CcmK4 assemblies? Is this interaction influenced by McdB LLPS activity and by the differential pH of the carboxysome versus the cytoplasm? Finally, does McdB influence metabolite channeling into and out of carboxysomes and is this activity dependent on LLPS behavior of McdB? It is attractive to speculate that local pH changes due to the metabolic activities of the carboxysomes also serves to regulate its composition, structure, and function. Together, our results have broad implication for understanding the diversity of the McdAB system, the mechanisms of carboxysome positioning among carbon-fixing bacteria, and the role of LLPS in the biogenesis and spatial organization of carboxysomes. Materials and Methods ===================== McdAB Homolog Search and Neighborhood Analysis ---------------------------------------------- Initial searches for *S. elongatus* McdA (Synpcc7942_1833) and McdB (Synpcc7942_1834) homologs were performed via BLASTp. As BLASTp returned few hits for McdA and not a single reliable hit for McdB, we reasoned that neighborhood analysis was a better method for identifying possible homologs of these proteins. Neighborhood analysis was performed by obtaining genome assembly GenBank files from NCBI for all cyanobacteria used in this study. As genome annotations are inconsistent among cyanobacterial species and carboxysome components can be found across multiple loci, homologs of *S. elongatus* carboxysome components CcmK2 (Synpcc7942_1421), CcmK3 (Synpcc7942_0284), CcmK4 (Synpcc7942_0285), CcmL (Synpcc7942_1422), CcmM (Synpcc7942_1423), CcmN (Synpcc7942_1424), CcmO (Synpcc7942_1425), CcmP (Synpcc7942_0520), RbcS (Synpcc7942_1427), RbcL (Synpcc7942_1426), RbcX (Synpcc7942_1535), CcaA (Synpcc7942_1447), and RbcR (Synpcc7942_1980) were used as BLASTp queries to identify carboxysome components in all other cyanobacteria. We note, additional shell proteins exist among other cyanobacteria, including CcmK1, CcmK5, and CcmK6 ([@msz308-B88]). However, these proteins share high similarity to CcmK2, so they were also captured during our BLASTp search using CcmK2. Neighborhood analysis for each carboxysome component across all cyanobacterial genomes was performed manually. We defaulted to this approach as many cyanobacterial genomes were still drafts and we wanted to determine whether the contigs or scaffolds were too small to make an appropriate neighborhood determination. Moreover, we also wanted the ability to quantify the number of base-pairs and coding sequences between *mcdA* or *mcdB* and the gene of the neighboring carboxysome component. Neighborhood analysis was carried out via Biomatters Geneious v 11.1.5 by searching 10 kb up- and downstream of each carboxysome component gene across all cyanobacterial species used in this study to identify coding sequences that matched our criteria for McdA or McdB. Accession numbers of identified McdA, McdB, and carboxysome component(s) were recorded using Excel. McdAB Sequence Analysis ----------------------- Multiple sequence alignments for McdA proteins were performed using MAFFT 1.3.7 under the G-INS-I algorithm, whereas the E-INS-I algorithm was used for McdB proteins due to long gaps caused by intrinsic disorder. Coiled-coil predictions for McdB proteins were carried out using DeepCoil ([@msz308-B55]). Predictions of disorder within McdB proteins were performed using PONDR with the VL-XT algorithm ([@msz308-B109], [@msz308-B79]; [@msz308-B106]). Analysis of McdB hydrophobicity was conducted with ProtScale using the Kyte and Doolittle scale ([@msz308-B48]; [@msz308-B30]). Phylogenetic Inference ---------------------- Ortholog sequences for *S. elongatus* DnaG, RplA, RplB, RplC, RplD, and RplE were obtained via BLASTp for each cyanobacterium. Alignments for protein sequence were performed using MAFFT 1.3.7 under the G-INS-I algorithm and BLOSUM62 scoring matrix. The six resulting alignments were then concatenated into one alignment using Geneious 11.1.5. Regions of low conservation within the resulting alignment were removed using gBlocks 0.91 b ([@msz308-B20]; [@msz308-B92]). A phylogenetic tree was then estimated with maximum likelihood analyses using RAxML 8.2.11 under the LG+Gamma scoring model of amino acid substitution. Bootstrap values were calculated from 500 replicates. The resulting phylogenetic tree was then edited and visualized using iTOL3 ([@msz308-B51]). Construct Development --------------------- The His--SUMO--McdB expression plasmid was generated via Gibson Assembly ([@msz308-B31]) using synthetized dsDNA for His--SUMO--McdB that was inserted into a pET11b vector. Construct was verified by sequencing. Cloning was performed in *Escherichia* *coli* DH5α chemically competent cells (Invitrogen). His--SUMO--McdB Overexpression ------------------------------ The His--SUMO--McdB fusion was expressed from a pET11b vector containing an ampicillin resistance cassette. All *E. coli* cultures containing the vector were grown in LB + carbenicillin (100 µg/ml) at 37 °C with 220 rpm shaking unless otherwise stated. The vector was transformed into competent *E. coli* BL21-AI, and a single transformant was picked to inoculate a 25 ml overnight culture. For expression, 2 l of culture was inoculated through a 1:100 dilution of the overnight culture and grown at 37 °C until an OD~600~ of 0.5--0.7 was reached. Expression was induced with final concentrations of IPTG at 1 mM and [l]{.smallcaps}-arabinose at 0.2%. Immediately after induction, cultures were cooled in an ice bath for 5 min and shifted to 18 °C to grow for 16 h at 220 rpm. Cells were pelleted at 4000×g for 20 min at 4 °C, flash frozen in liq. N~2~, and stored at −80 °C. His--SUMO--McdB Purification ---------------------------- The following buffers were used to purify the His--SUMO--McdB fusion: lysis buffer (50 mM HEPES pH 7.6; 1 M KCl; 5 mM MgCl~2~; 10% glycerol; 2 mM BME; 20 mM imidazole; 0.05 mg/ml lysozyme; 0.05 µl/ml benzonase; protease inhibitor), Ni buffer A (50 mM HEPES pH 7.6; 1 M KCl; 5 mM MgCl~2~; 10% glycerol; 2 mM BME; 20 mM imidazole), and Ni buffer B (50 mM HEPES pH 7.6; 1 M KCl; 5 mM MgCl~2~; 10% glycerol; 2 mM BME; 500 mM imidazole). The 2-l cell pellet was resuspended in 150 ml lysis buffer and lysed using a microfluidizer at 18,000 psi equilibrated in Ni buffer A. The lysate was centrifuged at 30,000×g at 4 °C for 30 min. The clarified lysate was decanted, syringe filtered (0.2 μm), and loaded onto a 5 ml HP HisTRAP column equilibrated in Ni buffer A. The column was washed with 25 ml Ni buffer A and then 25 ml 5% Ni buffer B. Elution was done using a 5--100% gradient of Ni buffer B while collecting 2 ml fractions for a total of 60 ml. Peak fractions were flash frozen in N~2~ (liq.) and stored at −80 °C. Microscopy of In Vitro McdB Phase Separation -------------------------------------------- Prior to imaging, His--SUMO--McdB samples from *S. elongatus* 7942 were buffer exchanged into a Ulp1 reaction buffer (150 mM KCl; 25 mM HEPES; 2 mM BME) and adjusted to the indicated pH. Similarly, His--SUMO--McdB samples from each of the Type 2 systems chosen were buffer exchanged into a defined Ulp1 reaction buffer, being *Gloeobacter kilaueensis* JSI (150 mM NaCl; 25 mM HEPES, pH 7.0; 2 mM BME), *Fremyella diplosiphon* NIES-3275 (25 mM HEPES, pH 7.0; 2 mM BME), and *Fischerella* sp. PCC 9431 (150 mM NaCl; 25 mM HEPES, pH 7.0; 2 mM BME). Buffer exchange was performed using 7 K MWCO, 5 ml Zeba Spin Desalting Columns (Thermo-Fischer). All imaging was performed using 16 well CultureWells (Grace BioLabs). Wells were passivated by overnight incubation in 5% (w/v) Pluronic acid (Thermo-Fischer), and washed thoroughly with the corresponding Ulp1 buffer prior to use. For cleavage experiments, 1 µl of purified Ulp1 was added to 50 µl of His--SUMO--McdB at the indicated concentration and incubated at 23 °C for 2 h to ensure complete cleavage. Imaging of McdB droplet formation was performed using a Nikon Ti2-E motorized inverted microscope (60× DIC objective and DIC analyzer cube) with a Transmitted LED Lamp house and a Photometrics Prime 95B Back-illuminated sCMOS Camera. Image analysis was performed using Fiji v 1.0. Supplementary Material ====================== [Supplementary data](#sup1){ref-type="supplementary-material"} are available at *Molecular Biology and Evolution* online. Supplementary Material ====================== ###### Click here for additional data file. We would like to thank Lindsay Matthews from the Simmons Lab for providing the Ulp1 enzyme. This work was supported by the National Science Foundation to A.G.V. (Award No. 1817478 and CAREER Award No. 1941966), research initiation funds provided by the MCDB Department to A.G.V., University of Michigan, and by research funds from the Michigan Life Sciences Fellows Program to J.S.M. Author Contributions ==================== J.S.M. and A.G.V. conceived the project. J.S.M., J.L.B., and A.G.V. designed experiments. J.S.M. and J.L.B. performed all experiments. J.S.M. and A.G.V. wrote the article. All authors discussed results and edited the article.

Introduction {#s1} ============ *Rickettsia heilongjiangensis* is an obligate intracellular Gram-negative bacterium that belongs to the tick-borne spotted fever group (SFG) of rickettsiae [@pone.0070440-Fournier1]. It was first isolated in 1983 from *Dermacentor silvarum* ticks in Suifenhe, in the Heilongjiang Province of China [@pone.0070440-Lou1]. *R. heilongjiangensis*, which is pathogenic to humans [@pone.0070440-Wu1]--[@pone.0070440-Mediannikov2], causes the disease now formally known as Far-Eastern spotted fever (FESF). This disease has been diagnosed in patients in Northeastern China [@pone.0070440-Wu1], Siberia and the Far East of Russia [@pone.0070440-Shpynov1], [@pone.0070440-Mediannikov1], [@pone.0070440-Mediannikov2], and Japan [@pone.0070440-Ando1]. After an incubation period of 4 to 7 days, most of the patients naturally infected with *R. heilongjiangensis* experience fever, chills, headache, dizziness, myalgia, arthralgia and anorexia, after which most of the patients show signs of a macular or maculopapular rash, and some of the patients have a primary lesion (otherwise called an eschar) that appears at the site of tick attachment, as well as lymphadenopathy close to the eschar itself [@pone.0070440-Mediannikov1]. Almost half of the patients have hepatomegaly accompanied with increased alanine aminotransferase and/or aspartate aminotransferase activity, indicating that *R. heilongjiangensis* infection in these patients causes liver lesions [@pone.0070440-Mediannikov1]. It has been suggested that FESF is an important emerging infectious disease in Northeast Asia. Our previous study using a mouse model revealed that *R. heilongjiangensis* caused severe systemic infection and that the pathological lesions in the infected organs (lungs, spleen, and brain) were associated with host inflammatory responses induced by *R. heilongjiangensis* [@pone.0070440-Duan1]. Like other pathogenic SFG rickettsiae, *R. heilongjiangensis* has the ability to invade and proliferate within vascular endothelial cells and cause cell injury and death [@pone.0070440-Meng1]. A "zipper-like" invasion strategy has been proposed for rickettsia invasion of non-phagocytic host cells [@pone.0070440-Jeng1], [@pone.0070440-Heinzen1]. Zipper-like invasion is a receptor-mediated invasion mechanism, whereby a bacterial protein induces host intracellular signaling through extracellular stimulation of a membrane receptor [@pone.0070440-Chan1], which suggests that rickettsiae surface-exposed proteins (SEPs) play a fundamental role in host-rickettsial interactions. The Sca (surface cell antigen) family proteins of SFG rickettsiae are recognized as the dominant SEPs [@pone.0070440-Chan1], [@pone.0070440-Blanc1] that play important roles in rickettsial pathogenesis. Sca0 (outer membrane protein A, OmpA) and Sca1 are both involved in attachment of rickettsiae to host cells [@pone.0070440-Li1], [@pone.0070440-Riley1], whereas Sca5 (OmpB) is associated with rickettsiae entry into host cells [@pone.0070440-Uchiyama1]--[@pone.0070440-Chan2]. Sca2 functions as a formin mimic that is responsible for the actin-based motility of rickettsiae in host cells [@pone.0070440-Cardwell1], [@pone.0070440-Kleba1]. Sca4 activates vinculin and interacts with the actin cytoskeleton of host cells [@pone.0070440-Park1]. In addition, OmpA and OmpB are known to be important protective antigens in SFG rickettsiae, and have the ability to efficiently induce humoral and cellular immunity against spotted fever [@pone.0070440-DiazMontero1]. Rickettsial surface components probably play roles in pathogenicity, such as adherence to and invasion of host cells, intracellular parasite growth, and/or interactions with immune cells. Proteomics analysis of rickettsiae surface molecules has the potential to discover novel molecules involved in bacterial pathogenesis, including those required for invasion of host cells, and those required for efficient induction of specific immune responses against rickettsial infection. Such studies also have the potential to deliver reagents for serological diagnosis of rickettsiosis. In the present study, *R. heilongjiangensis* SEPs were identified by a proteome analysis of its cell surface proteins. Based on these results, recombinant SEPs from *R. heilongjiangensis* were prepared and serologically characterized by protein microarray assays and an enzyme-linked immune sorbent assay (ELISA) using sera from FESF patients or from mice experimentally infected with *R. heilongjiangensis* as well as other rickettsial pathogens. Materials and Methods {#s2} ===================== Ethics Statement {#s2a} ---------------- Specific pathogen-free male C3H/HeN or BALB/c mice (5 to 6 weeks of age) were purchased from the Laboratory Animal Center of Beijing, China. The animal experiments were approved by the Beijing Administrative Committee for Laboratory Animals and the animal care met the standard of the committee. Mice were well cared for during their stay in the facility and all efforts were made to minimize suffering. Research using samples from humans was approved by the Institutional Review Board of the Beijing Institute of Microbiology and Epidemiology. The study was performed after the receipt of informed written consent from the patients and healthy donors, or their guardians. The data analysis was performed anonymously. Mouse and Human Sera {#s2b} -------------------- Serum samples from BALB/c mice infected with *R. heilongjiangensis* (054 strain) were prepared as described previously [@pone.0070440-Duan1]. Sera from BALB/c mice experimentally infected with *R. sibirica* (246 strain), *R. rickettsii* (Sheila Smith), *R. prowazekii* (Madrid E), *R. typhi* (Wilmington), *Orientia tsutsugamushi* (Karp), or *Coxiella burnetii* (Xinqiao) were obtained from the laboratories in our Institute. All the mouse sera used were collected on day 28 post-infection. The titers of specific IgG antibodies in the sera samples were determined through use of their corresponding rickettsial antigens and indirect immunofluorescence assays (IFA) [@pone.0070440-Li2]. Nine sera from FESF patients and seven sera from healthy blood donors were obtained from hospitals in northeast China. The serum samples were collected in the hospitals as part of their routine management of patients or blood donors (i.e., it was done without any additional sampling), and data from the patients and blood donors were anonymized. The IgG antibody titers against *R. heilongjiangensis* in the human sera were determined by IFA [@pone.0070440-Li2]. Each serum sample was diluted 1∶100 in PBS (8.1 mM Na~2~HPO~4~, 1.9 mM NaH~2~PO~4~, 154 mM NaCl; pH 7.4) and then neutralized overnight using an *Escherichia coli* cell lysate at a final protein concentration of 5 mg/ml before use. Cultivation and Purification of *R. heilongjiangensis* {#s2c} ------------------------------------------------------ The *R. heilongjiangensis* (054 strain) was propagated in Vero cells (ATCC) and purified by renografin density centrifugation [@pone.0070440-Duan1]. The viability of newly purified bacteria was ∼95%, as measured using LIVE/DEAD BacLight Bacterial Viability Kits (Invitrogen, Carlsbad, CA) and the numbers of bacteria purified were estimated by quantitative polymerase chain reaction (qPCR) as described previously [@pone.0070440-Duan1]. Isolation and Separation of Surface-exposed Proteins {#s2d} ---------------------------------------------------- Newly purified *R. heilongjiangensis* (4×10^11^) bacteria were labeled with Sulfo-NHS-SS-Biotin (Thermo Scientific, Rockford, IL) and the biotinylated proteins were captured by streptavidin agarose resin (Thermo Scientific), which was performed as described previously [@pone.0070440-Ge1], [@pone.0070440-Ge2]. The captured proteins were eluted from the streptavidin resin with 500 mM dithiothreitol, and the eluted proteins were precipitated using a 2D-Cleanup Kit (GE healthcare, Waukesha, WI). The isolated proteins dissolved in rehydration buffer (7 M urea, 2 M thiourea, 4% CHAPS) were subjected to two-dimensional electrophoresis (2D-PAGE) as described previously [@pone.0070440-Xiong1]. Identification of Surface-exposed Proteins by Mass Spectrometry {#s2e} --------------------------------------------------------------- The protein spots on the silver-stained gel were excised and then subjected to in-gel digestion with trypsin [@pone.0070440-Xiong1]. The hydrolysates were analyzed by electrospray ionization tandem mass spectrometry (ESI-MS/MS) and the resultant peptides were mass fingerprinted and compared against the National Center for Biotechnology Information (NCBI) nonredundant databases using the Mascot search engine (<http://www.matrixscience.co.uk>) [@pone.0070440-Xiong1]. N-terminal signal peptides and non-classical secretion signals in the proteins were predicted by the SignalP 3.0/LipoP 1.0 [@pone.0070440-Bendtsen1], [@pone.0070440-Juncker1] and SecretomeP 2.0 [@pone.0070440-Bendtsen2] servers, respectively. SignalP is able to predict the presence and location of N-terminal signal peptides based on a combination of artificial neural networks (NN) and hidden Markov models (HMM) [@pone.0070440-Bendtsen1]. LipoP 1.0 was used to distinguish lipoproteins, SpaseI-cleaved proteins, and N-terminal membrane helices [@pone.0070440-Juncker1]. SecretomeP, which integrates various posttranslational and localizational aspects of the protein from a large number of other feature prediction servers into the final secretion prediction, is able to predict non-classically secreted proteins [@pone.0070440-Bendtsen2]. The subcellular location (SCL) of each protein was predicted by the PSORTb 3.0.2 [@pone.0070440-Yu1] and SOSUI-GramN [@pone.0070440-Imai1] servers. PSORTb combines several analytical methods [@pone.0070440-Yu1] and SOSUI-GramN uses the physicochemical parameters of the N- and C-terminal signal sequences, and the total sequence of a protein [@pone.0070440-Imai1] to predict the SCL of proteins. Both of these software packages generate prediction results for five major cellular and subcellular locations (i.e., cytoplasmic, inner membrane, periplasmic, outer membrane and extracellular) for Gram-negative bacteria. Sequence homology for each of the proteins identified herein was analyzed against proteins from other species of bacteria (available in the NCBI public database) using the Basic Local Alignment Search Tool (BLAST) (<http://www.ncbi.nlm.nih.gov/BLAST/>). The proteins identified were classified into the Clusters of Orthologous Groups of proteins (COGs) and functions were assigned by the COGnitor server available at NCBI. Preparation of Recombinant Surface-exposed Proteins {#s2f} --------------------------------------------------- The genes encoding SEPs were obtained from the genomic sequence of *R. heilongjiangensis* (GenBank accession number: CP002912) [@pone.0070440-Duan2] and were amplified by the polymerase chain reaction (PCR) with their corresponding primer pairs ([Table S1](#pone.0070440.s001){ref-type="supplementary-material"}). The amplified genes were expressed in *E. coli* and the expressed recombinant proteins were purified with Ni-NTA affinity resin as described previously [@pone.0070440-Li2]. Surface Localization of YbgF and PrsA by Immunofluorescence Assay {#s2g} ----------------------------------------------------------------- Immune sera against YbgF or PrsA proteins were prepared in C3H/HeN mice. Briefly, each mouse was subcutaneously administered 30 µg of recombinant YbgF or PrsA mixed with Freund's complete adjuvant (Sigma-Aldrich, St Louis, MO) for the primary immunization, followed by two booster immunizations with 20 µg of the homologous protein mixed with incomplete Freund's adjuvant (Sigma-Aldrich) on days 28 and 42 post primary immunization. Fourteen days after the last booster immunization, blood samples were collected from the five mice immunized with YbgF or PrsA and the separated sera were pooled together. Sera from mice immunized with *R. heilongjiangensis* whole-cell antigen (WCA) plus adjuvant and adjuvant alone were used as positive and negative controls, respectively. *R. heilongjiangensis* cells smeared onto slides were fixed with ice-cold acetone for 10 min and then incubated with the immune sera (diluted 1∶10 in PBS) for 45 min at 37°C [@pone.0070440-Li2]. After three washes with PBS, the rickettsial cells on the slides were incubated with a 1∶200 dilution of Dylight 488-conjugated goat anti-mouse IgG (Thermo Scientific) for 45 min at 37°C. After another three washes, the rickettsial cells on the slides were observed under a fluorescence microscope (Olympus BX60). Fabrication of Protein Microarrays {#s2h} ---------------------------------- Each of the purified recombinant SEPs was diluted to a final concentration of 300 µg per ml and printed onto polymer slides (Capitalbio, Beijing, China) as described previously [@pone.0070440-Xiong1]. Each protein was printed as five replicate spots, with mouse or human IgG printed as positive controls, while *E. coli* lysates from cells transformed with PET-32a plasmids were used as negative controls [@pone.0070440-Xiong1]. For quality control, the microarray slides were incubated with Cy5-labeled mouse anti-His tag IgG (SBA, Birmingham, AL) and scanned for their fluorescence intensity (FI) and the scanned images were analyzed by GenePix Pro 6.0 software (Molecular Devices, Sunnyvale, CA) [@pone.0070440-Xiong1]. Proteins with a signal-to-background ratio over 3.0 were used for further analysis [@pone.0070440-Xiong1]. Serological Analysis of Surface-exposed Proteins using Microarray Assays {#s2i} ------------------------------------------------------------------------ Recombinant SEPs on the microarray slide were analyzed using various sera according to previous descriptions [@pone.0070440-Xiong1]. Briefly, the microarray slide was blocked with PBS-BSA (PBS, 1% \[w/v\] BSA, pH 7.4) for 1 h, after which it was incubated with a 100 µl volume of each serum sample at room temperature for 1 h. After incubation, the microarray slide was washed six times in PBST (PBS, 0.05% \[v/v\] Tween 20, pH 7.4) for 5 min each time on a shaker. The microarray slide was then incubated with goat-anti-mouse IgG-Cy5 or goat-anti-human IgG-Cy5 (SBA) at a 1∶500 dilution at room temperature for 1 h. Following an additional six washes in PBST, the air-dried microarray slide was scanned with a GenePix Personal 4100A scanner (Molecular Devices) and the scanned images were analyzed by GenePix Pro 6.0 (Molecular Devices). The FI value of each protein was calculated by averaging the FI values of five replicate spots in which the backgrounds had been subtracted. The reaction was considered positive if the average FI value of any protein that had been probed with any of the serum samples from infected mice was higher than 3 standard deviations (SD) above the average FI value of the same protein that had been probed with normal mouse sera [@pone.0070440-Ramachandran1]. In addition, the reaction was considered positive if the average FI value of any protein probed with any of the sera from patients was higher than 2 SD above the average FI value of the same protein probed with sera from healthy people [@pone.0070440-Xiong1]. FI values for proteins probed with sera from infected individuals and control sera from uninfected individuals were analyzed by the Wilcoxon Two-Sample test, whereas FI values for proteins probed with sera from different bacterial infections were performed by the Kruskal-Wallis test, followed by the Student-Newman-Keuls (SNK) test using software SAS 9.1 (SAS Institute, Cary, NC). Serological Analysis of Major Surface-exposed Proteins using ELISA {#s2j} ------------------------------------------------------------------ ELISAs were performed as described previously [@pone.0070440-Fu1]. Briefly, each well of a 96-well microplate (Corning, Corning, NY) was coated overnight at 4°C with 100 µl of each of the purified recombinant proteins at 2.5 µg per ml in ELISA/ELISPOT Coating Buffer (eBioscience, San Diego, CA). Unbound sites in each well were blocked with 200 µl of 1×ELISA Diluent Solution (eBioscience) for 2 h at 37°C. After three washes with PBST, each serum sample was dispensed into three replicate wells (100 µl per well) and the plates were incubated for another hour at 37°C. After another three washes with PBST, 100 µl of a 1∶5000 dilution of horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG (SBA) was added to each well. The plates were incubated at 37°C for another hour and 100 µl of 1×TMB ELISA Substrate Solution (eBioscience) was added to each well for 5 min at room temperature. Thereafter, 50 µl of H~2~SO~4~ (2 M) was added to stop the reaction. The optical density (OD) of each well was read at 450 nm using a microplate reader (UVM 340, ASYS HitechGmbH, Eugendorf, Austria) and the mean OD~450~ of three replicate wells was calculated. The cut-off value of each protein was determined as the mean OD~450~ of normal mouse sera plus 4 SD [@pone.0070440-Stynen1]. Results {#s3} ======= Identification of Surface-exposed Proteins by ESI-MS/MS Analysis {#s3a} ---------------------------------------------------------------- The biotinylated SEPs of *R. heilongjiangensis* were isolated by biotin-streptavidin affinity chromatography and these proteins were separated by 2D-PAGE. Approximately 50 protein spots with isoelectric points (pI) ranging from 5 to 10 and molecular masses ranging from 20 to 120 kDa were visualized on the 2D-PAGE gel stained with silver ([Figure 1](#pone-0070440-g001){ref-type="fig"}). {ref-type="supplementary-material"}.](pone.0070440.g001){#pone-0070440-g001} ESI-MS/MS analysis identified 25 proteins among 50 protein spots ([Figure 1](#pone-0070440-g001){ref-type="fig"} and [Table S2](#pone.0070440.s002){ref-type="supplementary-material"}). Most of the 50 spots were identifiable as single proteins. Some spots (such as spots 34 to 37) that appeared as a chain of spots with slightly different pI values were identified as being from the same protein, while a few of the spots (such as spots 7 to 11) contained several different proteins ([Figure 1](#pone-0070440-g001){ref-type="fig"} and [Table S2](#pone.0070440.s002){ref-type="supplementary-material"}). Fourteen of the 25 proteins were predicted by the SignalP and SecretomeP servers as classical and/or non-classical secretion proteins. All proteins except Omp1 that were identified as classical secretion proteins were predicted by the LipoP server to be SPaseI-cleaved proteins ([Table 1](#pone-0070440-t001){ref-type="table"}). 10.1371/journal.pone.0070440.t001 ###### COGs classification and bioinformatics analysis of *R. heilongjiangensis* surface-exposed proteins identified by ESI-MS/MS. {#pone-0070440-t001-1} COGs and Protein annotation Locus Tag Gene Symbol Signal peptide[a](#nt101){ref-type="table-fn"} (SignalP/LipoP) Subcellular Location[b](#nt102){ref-type="table-fn"} (PSORTb/SOSUI-GramN) Report[c](#nt103){ref-type="table-fn"} --------------------------------------------------------------------------- --------------------------------------------- ------------- ---------------------------------------------------------------- --------------------------------------------------------------------------- ---------------------------------------- M: Cell wall/membrane/envelope biogenesis hypothetical protein Rh054_06965 Rh054_06965[d](#nt104){ref-type="table-fn"} *adr1* Yes/SpI EC/OM 1, 2, 4 hypothetical protein Rh054_06970 Rh054_06970[d](#nt104){ref-type="table-fn"} *adr2* Yes/SpI Non-CYT/OM 1, 2, 3 putative nucleoside-diphosphate sugar epimerase CapD Rh054_02635 *capD* No/No CYT/CYT outer membrane protein omp1 Rh054_01180 *omp1* Yes/No OM/OM 2, 3, 5, 6 190-kDa cell surface antigen Rh054_06925[d](#nt104){ref-type="table-fn"} *ompA* Yes/SpI OM/OM 3 outer membrane protein B Rh054_06005[d](#nt104){ref-type="table-fn"} *ompB* Yes/SpI OM/EC 1, 2, 3, 4 OmpW family outer-membrane protein Rh054_00645[d](#nt104){ref-type="table-fn"} *ompW* Yes/SpI Unknown/OM 2 dTDP-4-dehydrorhamnose reductase Rh054_02630 *rfbD* No/No CYT/CYT cell surface antigen Rh054_00115[d](#nt104){ref-type="table-fn"} *sca1* Yes/SpI OM/EC 1 O: Posttranslational modification, protein turnover, chaperones heat shock protease Rh054_01375 *degQ* Yes/SpI PP/EC 1, 6 molecular chaperone GroEL Rh054_05320 *groEL* No/No CYT/CYT 1, 2, 3, 4, 5, 6 Protein export protein prsA Rh054_04865[d](#nt104){ref-type="table-fn"} *prsA* Yes/SpI OM/CYT 1, 2, 3, 4 hypothetical protein Rh054_05640 Rh054_05640 *surA* Yes/SpI Non-CYT/OM thioredoxin peroxidase 1 Rh054_02600 *tdpX1* No/No CYT/CYT 1, 2, 3 J: Translation, ribosomal structure and biogenesis 50S ribosomal protein L1 Rh054_01050 *rplA* No/No CYT/Unknown 50S ribosomal protein L25/general stress protein Ctc Rh054_05115 *rplY* No/No CYT/CYT 30S ribosomal protein S2 Rh054_00685 *rpsB* No/No CYT/EC 2 Elongation factor Tu Rh054_05545 *tuf* No/No CYT/PP 1, 2, 3, 4 C: Energy production and conversion F0F1 ATP synthase subunit beta Rh054_06725 *atpD* No/No CYT, IM/CYT 1, 2, 3, 4, 6 succinate dehydrogenase iron-sulfur subunit Rh054_00425 *sdhB* No/No IM/CYT H: Coenzyme transport and metabolism bifunctional 5,10-methylene-tetrahydrofolate dehydrogenase/cyclohydrolase Rh054_03580 *folD* No/No CYT/Unknown Q: Secondary metabolites biosynthesis, transport and catabolism hypothetical protein Rh054_06655 Rh054_06655[d](#nt104){ref-type="table-fn"} No/No CYT/CYT 4 S: Uncharacterized BCR Tol system periplasmic component Rh054_01780[d](#nt104){ref-type="table-fn"} *ybgf* Yes/SpI Non-CYT/CYT 4 Not in COGs hypothetical protein Rh054_00610 Rh054_00610[d](#nt104){ref-type="table-fn"} Yes/SpI OM/EC 1, 3, 5, 6 hypothetical protein Rh054_02285 Rh054_02285 Yes/SpI Non-CYT/IM Signal peptides and signal peptide types for all of the proteins were predicted with SignalP 3.0 or LipoP 1.0 software available online (<http://www.cbs.dtu.dk/services/SignalP-3.0> and <http://www.cbs.dtu.dk/services/LipoP-1.0>. The websites were accessed on 22 January, 2013). The subcellular location of each protein was predicted with PSORTb 3.0.2 or SOSUI-GramN (<http://www.psort.org/psortb/index.html> and <http://bp.nuap.nagoya-u.ac.jp/sosui/sosuigramn/sosuigramn_submit.html>. The websites were accessed on 22 January, 2013). Proteins with homology to some *R. heilongjiangensis* surface-exposed proteins were also identified on the surfaces or in membrane extracts of some other rickettsiae. 1, 2, 3, 4, 5, and 6 refer to *R. conorii* [@pone.0070440-Riley1], [@pone.0070440-Renesto1], [@pone.0070440-Renesto2], *R. felis* [@pone.0070440-Ogawa1], *R. parkeri* [@pone.0070440-Pornwiroon1], *R. typhi* [@pone.0070440-Sears1], *Anaplasma phagocytophilum* [@pone.0070440-Ge1], and *Ehrlichia chaffeensis* [@pone.0070440-Ge2], respectively. Non-classically secreted proteins were predicted with SecretomeP 2.0 (<http://www.cbs.dtu.dk/services/SecretomeP/>. The website was accessed on 22 January, 2013). SpI: signal peptide (signal peptidase I); EC: Extracellular; OM: Outer membrane; PP: Periplasmic; IM: Inner membrane. CYT: Cytoplasmic; Non-CYT: Non-Cytoplasmic; BCR: Bacterial conserved region. Twelve proteins were predicted by PSORTb and/or SOSUI-GramN to localize within the outer membrane and/or extracellular space and five proteins were predicted to reside in the inner membrane, periplasm or other non-cytoplasmic location ([Table 1](#pone-0070440-t001){ref-type="table"}). In the database search for COGs, the 25 SEPs were classified into eight categories; nine SEPs, including OmpA, OmpB and Sca1 [@pone.0070440-Chan1]--[@pone.0070440-Riley1], [@pone.0070440-Uchiyama2], [@pone.0070440-Chan2], [@pone.0070440-Sears1] were classified as belonging to group M, which is predicted to have functions involved in cell wall, membrane or envelope biogenesis ([Table 1](#pone-0070440-t001){ref-type="table"}). To confirm the surface localization of the novel surface proteins, YbgF and PrsA, which were recognized in this study, *R. heilongjiangensis* cells were stained for IFAs with antibodies against YbgF or PrsA. Fluorescent rings were observed around the rickettsial cells stained with antibodies to YbgF ([Figure 2B](#pone-0070440-g002){ref-type="fig"}), while fluorescent spots were seen at one end of the cells stained with antibodies to PrsA ([Figure 2C](#pone-0070440-g002){ref-type="fig"}). {#pone-0070440-g002} Serological Analysis of Surface-exposed Proteins using Mouse Sera {#s3b} ----------------------------------------------------------------- All of the SEPs identified, with Sca1 being the exception, were successfully expressed in *E. coli* cells. Omp1, OmpA and OmpB were expressed as two, two, and three fragments, respectively ([Figure 3](#pone-0070440-g003){ref-type="fig"}). The purified recombinant SEPs fabricated on the microarray slide were analyzed with 16 sera from *R. heilongjiangensis*-infected mice and 16 sera from normal mice ([Tables 2](#pone-0070440-t002){ref-type="table"} and [S3](#pone.0070440.s003){ref-type="supplementary-material"}). The average FI value of each protein probed with the sera from the infected mice was significantly higher than that probed with normal sera (p\<0.05). All of the SEPs, with the exception of four proteins (i.e., Omp1-2, OmpB-1, RfbD and hypothetical protein Rh054_00610), reacted positively with at least one of the sera from the infected mice. Eleven proteins (i.e., GroEL, OmpA-2, OmpB-3, PrsA, RplA, RplY, RpsB, SdhB, SurA, YbgF and hypothetical protein Rh054_02285) reacted positively with at least half of the sera from the infected mice and were therefore considered to be the major seroreactive proteins (antigens). OmpA-2 and GroEL reacted positively with all of the sera from the infected mice and OmpA-2 had the highest average FI value ([Tables 2](#pone-0070440-t002){ref-type="table"} and [S3](#pone.0070440.s003){ref-type="supplementary-material"}). {#pone-0070440-g003} 10.1371/journal.pone.0070440.t002 ###### Average fluorescence intensity and standard deviation of each surface-exposed protein probed with the sera from *R. heilongjiangensis*-infected mice and FESF patients. {#pone-0070440-t002-2} Average fluorescence intensity ± standard deviation (positive serum No./total serum No.) ------------- ------------------------------------------------------------------------------------------ -------------------- ------------------- ------------------- Adr1 67.5±55 (4/16) 24.1±8.7 (0/16) 122.8±51.5 (0/9) 86.7±50.9 (0/7) Adr2 37.9±23.8 (2/16) 16.8±10.5 (0/16) 77±29.8 (1/9) 49±33.6 (0/7) AtpD 69.4±22.3 (6/16) 30.9±13.7 (0/16) 84.9±94.5 (0/9) 97.4±74.9 (0/7) CapD 495.7±244.4 (3/16) 308.5±163.2 (0/16) 60.3±51 (1/9) 47.3±35.3 (0/7) DegQ 298.5±248.1 (3/16) 104.8±168.1 (0/16) 64.7±68.1 (1/9) 58±42 (0/7) FolD 49.9±14 (4/16) 24.5±11 (0/16) 99.5±42.8 (0/9) 89.1±70.5 (0/7) GroEL 1085.2±606.8 (16/16) 12.9±8.2 (0/16) 522.5±879.4 (7/9) 25.4±30.3 (0/7) Omp1-1 50.5±36.4 (1/16) 23.9±30.1 (0/16) 89.1±36.8 (2/9) 45.2±30.0 (1/7) Omp1-2 233.4±206.5 (0/16) 195.7±154.1 (0/16) 130.6±146.2 (2/9) 73.9±43.7 (0/7) OmpA-1 35.3±16.7 (2/16) 17.7±10.1 (0/16) 62.3±60.2 (1/9) 71.9±55 (0/7) OmpA-2 4330.9±2886.5 (16/16) 25.6±22.9 (0/16) 1056±2055.9 (6/9) 59.7±34.9 (0/7) OmpB-1 178.7±147.6 (0/16) 122.6±119.4 (0/16) 92.7±90.9 (1/9) 65.1±57.4 (0/7) OmpB-2 397.6±183.7 (1/16) 248.3±166.4 (0/16) 182.6±115.1 (3/9) 118.9±56 (0/7) OmpB-3 379.5±768.7 (13/16) 24.1±9.6 (0/16) 225.4±170.2 (3/9) 89.2±54.7 (0/7) OmpW 340.9±295 (5/16) 123.9±147.0 (1/16) 179.4±88.1 (0/9) 142.7±79.9 (0/7) PrsA 1630.2±2833.6 (14/16) 20.7±8.4 (0/16) 87.9±46.6 (0/9) 75.4±59.4 (1/7) RfbD 273.4±265.1 (0/16) 252.8±232.9 (0/16) 65.7±66.5 (2/9) 38.9±44.2 (0/7) RplA 239.9±654.5 (10/16) 22.4±9.7 (0/16) 104.3±194.1 (1/9) 37.1±31.5 (0/7) RplY 41.4±18.9 (8/16) 20±5.2 (0/16) 106.9±37.6 (2/9) 58.3±37.7 (0/7) RpsB 79.2±28.1 (10/16) 33±13.5 (0/16) 121.6±90.4 (3/9) 69.1±44.9 (0/7) SdhB 82.5±27.3 (13/16) 27.1±11.1 (0/16) 198.7±343.4 (3/9) 67±41.7 (0/7) SurA 73.4±47.7 (8/16) 21.6±14.1 (0/16) 70.9±60.3 (1/9) 53.1±40.6 (0/7) TdpX1 359.6±216.9 (1/16) 240.9±137.5 (0/16) 82.2±84.2 (1/9) 58.4±53.9 (0/7) Tuf 591.6±291.1 (3/16) 336±182.6 (0/16) 56±61.4 (1/9) 45.1±39.2 (0/7) YbgF 186.8±256.8 (10/16) 24.6±11.7 (0/16) 79.6±56.3 (2/9) 32.5±24.4 (0/7) Rh054_00610 43.7±14.8 (0/16) 23.4±14.4 (0/16) 108.2±64.8 (1/9) 87±37.5 (0/7) Rh054_02285 182.7±200.9 (8/16) 30.2±17 (0/16) 72.3±78.9 (3/9) 49.7±49.3 (0/7) Rh054_06655 59.7±21.4 (3/16) 30.4±13.3 (0/16) 174.7±116.3 (1/9) 133.9±105.6 (0/7) FESF: Far-Eastern spotted fever. Serological Analysis of Surface-exposed Proteins using Patient Sera {#s3c} ------------------------------------------------------------------- The recombinant SEPs on the microarray slide were probed with nine sera from FESF patients and seven sera from healthy people. As a result ([Tables 2](#pone-0070440-t002){ref-type="table"} and [S3](#pone.0070440.s003){ref-type="supplementary-material"}), OmpA-2 and GroEL reacted positively against six and seven of the nine samples of patient sera, respectively; other 10 proteins reacted positively against between two and four of the nine patient serum samples, while the remaining proteins reacted with only one or none of the sera from the nine patients. Statistical analysis showed the average FI value of the GroEL protein probed with the patient sera was significantly higher than that probed with sera from healthy people (p\<0.01), and that the average FI values for all of the other proteins were also higher than for those probed with sera from healthy people, but the result was not statistically significant (p\>0.05). Specificity Analysis of Major Seroreactive Proteins using Microarray Assays {#s3d} --------------------------------------------------------------------------- The major seroreactive SEPs on the microarray slide were probed with serum samples from mice infected with different rickettsial agents. As a result ([Table 3](#pone-0070440-t003){ref-type="table"}, [Figure 4](#pone-0070440-g004){ref-type="fig"} and [Table S4](#pone.0070440.s004){ref-type="supplementary-material"}), all of the 11 major seroreactive SEPs reacted positively with at least six of the 10 sera from *R. heilongjiangensis*-infected mice; GroEL, OmpA-2, OmpB-3, SdhB, and the hypothetical protein Rh054_02285 reacted positively with at least six of the 10 sera from *R. sibirica*-infected mice; SdhB and the hypothetical protein Rh054_02285 reacted positively with five or less of the 10 sera from mice infected with *R. rickettsii*, *R. prowazekii*, *R. typhi*, *O. tsutsugamushi* or *C. burnetii*; however, all 10 sera from the *R. typhi*-infected mice were recognized by OmpA-2. {#pone-0070440-g004} 10.1371/journal.pone.0070440.t003 ###### Serological specificity of the major seroreactive surface-exposed proteins of *R. heilongjiangensis* analyzed by microarray assay using serum samples from mice infected with different rickettsial agents. {#pone-0070440-t003-3} Average and standard deviation (SD) of the FI value of each protein probed with different serum samples (positive No./total No.) ------------- ---------------------------------------------------------------------------------------------------------------------------------- ------------------- ------------------ ------------------- --------------------- -------------------- ------------------- ------------------- GroEL 727±434.5 (10/10) 281±569.9 (6/10) 29.7±35.4 (1/10) 10.2±18.5 (0/10) 25.6±18.3 (0/10) 36.6±28.8 (2/10) 15.5±9.9 (0/10) 19.1±15 (0/10) OmpA-2 3629.7±2294 (10/10) 464±218 (10/10) 78±33.5 (0/10) 80.6±47.9 (2/10) 953.2±258.2 (10/10) 38.0±22.0 (0/10) 58.9±33.5 (0/10) 61.9±25.6 (0/10) OmpB-3 1038.1±629.8 (9/10) 253±189 (10/10) 69.3±36.2 (2/10) 44±40.2 (0/10) 86.1±21 (1/10) 38.9±28.9 (0/10) 33±14.8 (0/10) 45.2±20.8 (0/10) PrsA 745.6±1296.1 (8/10) 112±48 (2/10) 45.3±34.4 (0/10) 62.1±34.0 (0/10) 66.4±18.5 (0/10) 43.1±39.2 (0/10) 30.8±27.1 (0/10) 61.8±30.1 (0/10) RplA 122.4±68.1 (6/10) 92.1±32.5 (3/10) 49.4±26.8 (0/10) 17.7±21.3 (0/10) 40.2±36.7 (0/10) 38.7±26.8 (0/10) 24.4±12.9 (0/10) 38.1±24.4 (0/10) RplY 316.2±169.2 (9/10) 114.9±76.5 (3/10) 51.8±30.4 (0/10) 35±28.6 (0/10) 59.4±33.9 (1/10) 35.3±29.9 (0/10) 21.3±10 (0/10) 40.6±27.1 (0/10) RpsB 658.8±348.5 (6/10) 196.9±84.6 (0/10) 138±73.9 (0/10) 383.4±91.7 (2/10) 233±152.6 (0/10) 204.5±230.2 (1/10) 240.1±59.6 (0/10) 333.9±54.6 (0/10) SdhB 129.6±79.6 (8/10) 101±41.8 (10/10) 51.2±37 (5/10) 31.8±28 (3/10) 32.2±21.1 (1/10) 40.4±33.3 (4/10) 18±13.7 (1/10) 22.2±8.6 (0/10) SurA 179.9±139.2 (7/10) 110.1±43.9 (4/10) 67.6±42 (1/10) 25.1±13.7 (0/10) 43.1±23.5 (0/10) 52.2±50.9 (2/10) 44.8±20 (0/10) 49.4±18.6 (0/10) YbgF 185±128.1 (8/10) 97.9±50.2 (3/10) 49.8±33.9 (1/10) 31.7±23.6 (0/10) 31.6±22.8 (0/10) 37.1±36.9 (1/10) 23±13.9 (0/10) 39.8±18.9 (0/10) Rh054_02285 182.5±99.9 (8/10) 111.3±44.1 (7/10) 69.8±39.6 (4/10) 57.3±29.8 (1/10) 48.6±25.1 (1/10) 38.7±27.1 (1/10) 98.5±124.3 (3/10) 46.8±11.7 (0/10) The major seroreactive surface-exposed proteins of *R. heilongjiangensis* fabricated on a microarray slide were probed with the sera from mice infected with *R. heilongjiangensis* (*R. hei*), *R. sibirica* (*R. sib*), *R. rickettsii* (*R. ric*), *R. prowazekii* (*R. pro*), *R. typhi* (*R. typ*), *O. tsutsugamushi* (*O. tsu*) or *C. burnetii* (*C. bur*), or normal mouse sera (Normal). The average FI values of the major seroreactive SEPs (but not RpsB and OmpA-2) probed with the sera from mice infected with *R. heilongjiangensis* or *R. sibirica* were significantly higher than those probed with sera from mice infected with any of the other rickettsial bacteria (p\<0.05) ([Figure 4](#pone-0070440-g004){ref-type="fig"}). The average FI value of OmpA-2 probed with sera from *R. heilongjiangensis*-infected mice was not significantly higher than that probed with the sera from *R. typhi*-infected mice (p\>0.05), but was significantly higher than that probed with sera from mice infected with any of the other rickettsial bacteria (p\<0.05) ([Figure 4](#pone-0070440-g004){ref-type="fig"}). Specificity Analysis of Major Seroreactive Proteins using ELISA {#s3e} --------------------------------------------------------------- The specificity of the major seroreactive SEPs was also analyzed by ELISA using sera from mice that had been infected with the different rickettsial bacteria. As shown in [Table 4](#pone-0070440-t004){ref-type="table"}, all of the SEPs (except SdhB and Rh054_02285) reacted positively with between five and 10 of the 10 serum samples from the mice infected with *R. heilongjiangensis*; all of the SEPs reacted positively with at least six out of 10 sera from mice infected with *R. sibirica* or *R. rickettsii*, and with between four and six of the 10 sera from mice infected with *O. tsutsugamushi*. We also found that only one to three out of 10 sera from mice infected with *R. prowazekii* were recognized by some of the SEPs and only one out of 10 sera from mice infected with *R. typhi* or *C. burnetii* were recognized by the SEPs, the exceptions being OmpB-3 and RplY. The concordance of the seroreactivity of the SEPs between the ELISA and microarray assay results was 75% ([Table 5](#pone-0070440-t005){ref-type="table"}). 10.1371/journal.pone.0070440.t004 ###### Serological specificity of the major seroreactive surface-exposed proteins of *R. heilongjiangensis* analyzed by ELISA using serum samples from mice infected with different rickettsial agents. {#pone-0070440-t004-4} Average and standard deviation (SD) of the optical density of each protein from ELISAs using various sera (positive No./total No.) ------------- ------------------------------------------------------------------------------------------------------------------------------------ ------------------- ------------------- ------------------ ------------------ ------------------ ------------------ ------------------ GroEL 1.31±0.37 (10/10) 0.6±0.47 (10/10) 0.75±0.43 (10/10) 0.13±0.02 (1/10) 0.14±0.02 (0/10) 0.36±0.38 (5/10) 0.14±0.02 (1/10) 0.11±0.01 (0/10) OmpA-2 1.83±0.23 (10/10) 0.73±0.2 (10/10) 0.44±0.29 (8/10) 0.14±0.03 (0/10) 0.15±0.02 (0/10) 0.28±0.22 (4/10) 0.15±0.02 (0/10) 0.14±0.03 (0/10) OmpB-3 1.01±0.59 (10/10) 0.35±0.29 (8/10) 0.31±0.11 (7/10) 0.17±0.05 (1/10) 0.23±0.04 (6/10) 0.24±0.14 (5/10) 0.18±0.04 (2/10) 0.14±0.02 (0/10) PrsA 0.85±0.76 (8/10) 0.28±0.08 (8/10) 0.29±0.09 (8/10) 0.16±0.06 (1/10) 0.16±0.03 (0/10) 0.32±0.31 (5/10) 0.13±0.02 (0/10) 0.13±0.02 (0/10) RplA 0.22±0.07 (5/10) 0.29±0.1 (8/10) 0.27±0.13 (6/10) 0.17±0.06 (3/10) 0.16±0.04 (1/10) 0.19±0.08 (4/10) 0.15±0.02 (0/10) 0.13±0.02 (0/10) RplY 0.32±0.16 (7/10) 0.54±0.15 (10/10) 0.52±0.31 (7/10) 0.19±0.05 (2/10) 0.22±0.04 (4/10) 0.38±0.26 (6/10) 0.23±0.07 (4/10) 0.14±0.02 (0/10) RpsB 0.3±0.07 (6/10) 0.43±0.11 (9/10) 0.37±0.18 (6/10) 0.22±0.07 (2/10) 0.21±0.04 (1/10) 0.24±0.13 (4/10) 0.17±0.02 (0/10) 0.15±0.03 (0/10) SdhB 0.21±0.04 (3/10) 0.38±0.1 (9/10) 0.34±0.15 (6/10) 0.17±0.05 (1/10) 0.19±0.03 (1/10) 0.27±0.17 (5/10) 0.15±0.02 (0/10) 0.14±0.02 (0/10) SurA 0.27±0.05 (6/10) 0.4±0.1 (10/10) 0.36±0.2 (6/10) 0.25±0.19 (3/10) 0.17±0.04 (0/10) 0.28±0.16 (6/10) 0.17±0.04 (0/10) 0.16±0.03 (0/10) YbgF 0.25±0.05 (8/10) 0.29±0.07 (9/10) 0.44±0.41 (7/10) 0.25±0.14 (3/10) 0.18±0.04 (1/10) 0.4±0.47 (5/10) 0.15±0.03 (0/10) 0.14±0.02 (0/10) Rh054_02285 0.22±0.04 (2/10) 0.4±0.11 (10/10) 0.31±0.14 (6/10) 0.23±0.1 (3/10) 0.19±0.04 (1/10) 0.31±0.21 (5/10) 0.15±0.02 (0/10) 0.15±0.02 (0/10) The sera from mice infected with *R. heilongjiangensis (R. hei), R. sibirica (R. sib), R. rickettsii (R. ric), R. prowazekii (R. pro), R. typhi (R. typ), O. tsutsugamushi (O. tsu) or C. burnetii (C. bur)*, or normal mouse sera (Normal) were used to test the specificity of the major seroreactive surface-exposed proteins of *R. heilongjiangensis*. 10.1371/journal.pone.0070440.t005 ###### Serological sensitivity and specificity values of the major seroreactive surface-exposed proteins of *R. heilongjiangensis* from microarray and ELISA tests. {#pone-0070440-t005-5} Microarray assay ELISA ------------- ------------------ ------- ---- ----- ------------------------------------------ ----- ---- ----- ---- ------------------------------------------ ---------------------------------------------- Proteins Se Sp Se Sp Remarks[\*](#nt110){ref-type="table-fn"} Se Sp Se Sp Remarks[\*](#nt110){ref-type="table-fn"} Concordance[\#](#nt111){ref-type="table-fn"} GroEL 100 87 57 96 Marker *R. hei* 100 61 100 86 Marker SFG 78% OmpA-2 100 69 67 76 100 69 93 92 Marker SFG 70% OmpB-3 90 81 70 98 Marker SFG 100 59 83 72 74% PrsA 80 97 33 100 Marker *R. hei* 80 69 80 88 Marker SFG 75% RplA 60 96 30 100 50 69 63 84 75% RplY 90 94 40 98 Marker *R. hei* 70 53 80 68 61% RpsB 60 96 20 94 60 69 70 86 Marker SFG 76% SdhB 80 66 77 82 30 69 60 86 86% SurA 70 90 40 96 Marker *R. hei* 60 64 73 82 76% YbgF 80 93 40 98 Marker *R. hei* 80 64 80 82 75% Rh054_02285 80 76 63 88 20 64 60 82 75% *R. hei*, *R. heilongjiangensis*; SFG, Spotted fever group; Se, sensitivity; Sp, specificity. proteins with \>70% sensitivity and \>85% specificity were selected as potential markers for *R. heilongjiangensis* or SFG rickettsia infections. concordance between ELISA and microarray assays used in the present study. Discussion {#s4} ========== The SEPs of obligate intracellular bacteria provide a crucial interface for interactions between bacteria and host cells [@pone.0070440-Sears1]. SEPs mediate the initial attachment of a bacterium to a host cell and subsequent contact with host cytosolic proteins, a process that promotes bacterial survival and replication by subverting host cellular processes [@pone.0070440-Sears1]. Definition of *R. heilongjiangensis* SEPs will provide an important starting point towards better understanding of the interactions that take place at the interface of rickettsia and host cells. In the present study, we identified 25 *R. heilongjiangensis* SEPs using biotin-streptavidin affinity chromatography coupled with 2D-PAGE and ESI-MS/MS. Some of the proteins were present in more than one spot in the gel, with the maximum number being 11 spots per protein. The theoretical molecular weights and/or pI values of the proteins did not correlate with the values obtained from the experiments. This may have been caused by the Sulfo-NHS-SS-Biotin residues that were not eluted during purification as well as post-translational modifications of rickettsial proteins [@pone.0070440-Ogawa1], [@pone.0070440-Boonjakuakul1]. OmpB was identified in eight spots (spots 13, 15, 18, 19, 22--24, and 48). Only spot 48 matched its theoretical size and pI; therefore, the other seven spots with molecular masses of ∼35 kDa were considered to be its β-peptide. Where proteins were identified in several spots in the gel, they were more likely to be highly abundant *R. heilongjiangensis* cell surface proteins. In the present study, nine proteins, including Adr1, Adr2, OmpA, OmpB, OmpW, PrsA, Sca1, YbgF, and the hypothetical protein Rh054_00610, were predicted to have both classical and non-classical secretion signals by the SignalP/LipoP and SecretomeP bioinformatics programs. Adr1 and Adr2 have been identified as important rickettsial adhesins [@pone.0070440-Renesto2], [@pone.0070440-Vellaiswamy1], [@pone.0070440-Balraj1]. Previous studies showed that antibodies to Adr1 in *R. conorii* and Adr2 in *R. prowazekii* inhibited entry of the homologous bacteria into host cells, suggesting they perform invasion-related roles [@pone.0070440-Vellaiswamy1], [@pone.0070440-Balraj1]. OmpA (Sca0), OmpB (Sca5), and Sca1 are involved in rickettsial attachment and/or entry to host cells [@pone.0070440-Blanc1]--[@pone.0070440-Riley1], [@pone.0070440-Uchiyama2], [@pone.0070440-Chan2]. While PrsA is an extracellular chaperone and membrane-bound lipoprotein that is essential for growth and protein secretion in *Bacillus* [@pone.0070440-Williams1], [@pone.0070440-Vitikainen1], YbgF is a periplasmic protein belonging to the Tol-Pal system, which maintains cell envelope integrity and also serves a role in the import of virulence factors of pathogenic bacteria [@pone.0070440-Dubuisson1]. In the present study, IFAs using antibodies against YbgF and PrsA revealed the presence of fluorescent rings and spots around the *R. heilongjiangensis* cells and at one end of the cells, respectively, suggesting they are localized on the surface of the bacteria ([Figure 2](#pone-0070440-g002){ref-type="fig"}). However, whether they are periplasmic or membrane-bound proteins requires further investigation. Roles for OmpW and hypothetical protein Rh054_00610 in bacterial pathogenesis have not been demonstrated. Four proteins, DegQ, Omp1, SurA, and hypothetical protein Rh054_02285, have classical secretion signal peptides, while hypothetical protein Rh054_06655 has a non-classical secretion signal. DegQ is a periplasmic serine protease, which is not essential for bacterial pathogenesis, but may play a small role during *salmonella* growth at systemic sites [@pone.0070440-Farn1]. SurA is either a periplasmic prolyl isomerase or chaperone that facilitates outer membrane protein biogenesis and pilus assembly in *E. coli* [@pone.0070440-Justice1]. Hypothetical protein Rh054_02285 was identified as a haloacid dehalogenase-like hydrolase. *R. heilongjiangensis* homologues of these secretion proteins (with the exceptions of SurA and the hypothetical protein Rh054_02285) reside on the surface of *R. conorii* [@pone.0070440-Riley1], [@pone.0070440-Renesto1], [@pone.0070440-Renesto2], *R. felis* [@pone.0070440-Ogawa1], *R. parkeri* [@pone.0070440-Pornwiroon1], *R. typhi* [@pone.0070440-Sears1], *Anaplasma phagocytophilum* [@pone.0070440-Ge1], and/or *Ehrlichia chaffeensis* [@pone.0070440-Ge2] ([Table 1](#pone-0070440-t001){ref-type="table"}). In addition, the classical secretion proteins in *R. heilongjiangensis* (except DegQ) are Sec-dependent extracytoplasmic proteins in *R. typhi* [@pone.0070440-Ammerman1]. The remaining 11 SEPs were not predicted to have secretion signals and are usually considered to be cytosolic proteins. However, proteins with homology to AtpD, GroEL, RpsB, TdpX1, and Tuf of *R. heilongjiangensis* were found on the cell surfaces or in the membrane fractions of *R. conorii* [@pone.0070440-Renesto1], [@pone.0070440-Renesto2], *R. felis* [@pone.0070440-Ogawa1], *R. parkeri* [@pone.0070440-Pornwiroon1], *R. typhi* [@pone.0070440-Sears1] and/or other Gram-negative bacteria [@pone.0070440-Ge1], [@pone.0070440-Ge2], [@pone.0070440-Boonjakuakul1] ([Table 1](#pone-0070440-t001){ref-type="table"}). AtpD is the inner membrane-localized β-subunit of an ATPase involved in bacterial adenosine nucleotide *de novo* biosynthesis [@pone.0070440-Gerken1]. GroEL is a heat shock protein that has been shown to be located on the surfaces of other rickettsial bacteria [@pone.0070440-Ge1], [@pone.0070440-Ge2], [@pone.0070440-Sears1]. TdpX1 (thioredoxin peroxidase 1) defends against oxidant stress in *Leishmania* [@pone.0070440-Knig1] and has been found in rickettsial organisms but not in free-living bacteria [@pone.0070440-Andersson1], indicating that its presence on the surface of rickettsial cells may afford protection against the oxidant stress encountered inside host cells. Tuf has been recognized as an adhesin-like factor with the ability to mediate attachment of *Lactobacillus johnsonii* to intestinal epithelial cells and mucins [@pone.0070440-Granato1]. Three ribosomal proteins, L1 (RplA), L25 (RplY) and S2 (RpsB), were identified among the secretion signal-lacking SEPs in the present study. RpsB is found on the surface of *R. felis* [@pone.0070440-Ogawa1] and in the present study is predicted to be an extracellular protein by SOSUI-GramN ([Table 1](#pone-0070440-t001){ref-type="table"}). Indeed, some ribosomal proteins have been recognized as membrane-associated proteins in bacteria [@pone.0070440-Ogawa1], [@pone.0070440-Pornwiroon1], while the surface proteins L12 [@pone.0070440-Spence1] and L25 [@pone.0070440-Mendum1] have been shown to be involved in bacterial pathogenicity [@pone.0070440-Spence1], [@pone.0070440-Stoll1]. Four other proteins, CapD, RfbD, SdhB, and FolD, were identified on the surface of *R. heilongjiangensis* cells in the present study. CapD, a UDP-glucose 4-epimerase [@pone.0070440-Santhanagopalan1], and RfbD, a dTDP-4-dehydroamnose reductase, both contribute to exopolysaccharide synthesis in Gram-negative bacteria [@pone.0070440-Santhanagopalan1], [@pone.0070440-Jiang1]. SdhB and FolD are succinate dehydrogenase iron-sulfur subunit and methylenetetrahydrofolate dehydrogenase or cyclohydrolase, respectively, and they may be associated with synthesis of membrane carbohydrates or energy metabolism in bacterial membranes. In previous studies, cytosol aminopeptidase has been found in the outer membrane fraction of *A. marginale* [@pone.0070440-Lopez1], while disulfide oxidoreductase has been observed on the bacterial surface of *E*. *chaffeensis* [@pone.0070440-McBride1], and elongation factor G and the ATP synthase F1 alpha and beta subunits have been detected on the cell envelope of *Staphylococcus aureus* [@pone.0070440-Gatlin1]. Therefore, these proteins with well-known functions in bacterial membranes may also be present on the surface of *Rickettsiae* and play unexpected roles in rickettsia-host interactions. To further characterize their roles in the pathogenicity and immunogenicity of *R. heilongjiangensis*, the recombinant SEPs were used to fabricate a microarray for analysis with sera from *R. heilongjiangensis*-infected mice and FESF patients. Eleven SEPs, including GroEL, OmpA-2, OmpB-3, PrsA, RplA, RplY, RpsB, SdhB, SurA, YbgF, and hypothetical protein Rh054_02285, were recognized as major seroreactive antigens by the sera from *R. heilongjiangensis*-infected mice. OmpA and OmpB were expressed as two or three fragments; however, only the C-terminal fragments (OmpA-2 and OmpB-3) reacted strongly with these sera, which could be related to the presence of a conserved outer membrane autotransporter domain located at the C-terminal fragments of both proteins. In the present study, GroEL and PrsA were identified in 10 and 13 spots on the 2D-PAGE gel, respectively, suggesting that they are both highly abundant surface proteins in *R. heilongjiangensis*. YbgF is recognized as a fairly abundant seroreactive protein, which was identified in several spots in the immunoproteomic assay of *C. burnetii* [@pone.0070440-Deringer1] and has been suggested as a serological marker for Q fever [@pone.0070440-Xiong1]. In our microarray analysis, OmpA-2 and GroEL showed the strongest seroreactivity with the sera from *R. heilongjiangensis*-infected mice and FESF patients, indicating that they may be attractive candidates for the development of serodiagnostic tests for FESF. The other major seroreactive SEPs that were identified using sera from *R. heilongjiangensis*-infected mice were poorly recognized by the FESF patient sera, suggesting that the immune profile of these SEPs in mice experimentally infected with *R. heilongjiangensis* differed markedly from that in patients that were naturally infected with this pathogen. However, some of the SEPs, including OmpB-3, RpsB, SdhB and hypothetical protein Rh054_02285, still reacted positively with 33% of the serum samples from FESF patients. This indicates that combining these SEPs with OmpA-2 and GroEL may lead to significant improvements in sensitivity and specificity for serodiagnosis of FESF. To explore their serological specificity, the major seroreactive *R. heilongjiangensis* SEPs were analyzed by microarray and ELISA using sera from mice infected with different rickettsial agents. Eleven, five and one of the 11 major seroreactive SEPs were recognized by more than 50% of the sera in the microarray assays from mice infected with *R. heilongjiangensis*, *R. sibirica* or *R. rickettsii*, respectively. The average FI values of the 11 SEPs probed with the sera from *R. sibirica*- or *R. rickettsii*-infected mice were lower or significantly lower than those probed with the sera from *R. heilongjiangensis*-infected mice ([Figure 4](#pone-0070440-g004){ref-type="fig"}). This result is consistent with the phylogenetic analysis of SFG rickettsiae, which shows that *R. heilongjiangensis* is much more closely related to *R. sibirica* than *R. rickettsii* [@pone.0070440-Fournier1]. In addition, all the major seroreactive SEPs, with the exception of OmpA-2, SdhB, and hypothetical protein Rh054_02285, reacted positively with few of the sera from mice infected with the non-SFG rickettsiae ([Figure 4](#pone-0070440-g004){ref-type="fig"}). The microarray results suggest that GroEL, PrsA, RplY, SurA, and YbgF may be potential markers of *R. heilongjiangensis* infection and OmpB-3 may be a potential marker of SFG rickettsia infection since they had over 70% sensitivity and 85% specificity in this serological analysis ([Table 5](#pone-0070440-t005){ref-type="table"}). The ELISA test showed that most of the major seroreactive SEPs had higher seroreactivity to the sera from mice infected with *R. sibirica*, *R. rickettsii*, *R. prowazekii*, or *O. tsutsugamushi* compared with the microarray assay. The higher seroreactivity of these proteins might be caused by denaturation in the ELISAs [@pone.0070440-Wilson1]. It is possible that the denatured proteins may have provided more epitopes that could cross-react with the various samples of sera tested here and that the proteins on the microarray slide may have maintained their native structures and activities [@pone.0070440-Wilson1]. Some proteins, such as OmpB-3 and RplY, exhibited high cross-reactivity with some of the sera from mice infected with non-SFG rickettsiae ([Table 4](#pone-0070440-t004){ref-type="table"}). This might be related to the highly conserved amino acid structures of these proteins in rickettsiae. OmpB is recognized as having high sequence conservation in rickettsiae [@pone.0070440-Uchiyama1]--[@pone.0070440-Chan2], while RplY also exhibits high amino acid sequence conservation as shown by sequence alignments of the seven rickettsial species used in the present study ([Table S5](#pone.0070440.s005){ref-type="supplementary-material"}). Interestingly, OmpA-2 reacted positively in the microarray assay with all 10 sera from *R. typhi*-infected mice, but with none of these sera in ELISA. The tertiary structure of OmpA-2 on the microarray slide may have exerted a steric effect that promoted non-specific absorption of IgG from the sera, an effect that would not apply to the denatured OmpA-2 in the ELISAs. This, combined with the small number of serum samples, may have affected concordance between the two assays. As shown in [Table 5](#pone-0070440-t005){ref-type="table"}, GroEL, OmpA-2, PrsA, and RpsB had a high sensitivity (\>70%) and specificity (\>85%) for SFG rickettsia (*R. heilongjiangensis*, *R. sibirica*, and *R. rickettsii*) antibodies. However, the cross-reactivity of these proteins against *O. tsutsugamushi* antibodies requires further investigation. Many studies have shown that OmpA and OmpB elicit protective immune responses against rickettsiosis in laboratory animals [@pone.0070440-DiazMontero1], [@pone.0070440-CrocquetValdes1]--[@pone.0070440-Sumner1], but the immunoprotectivity of the remaining major seroreactive SEPs of *R. heilongjiangensis* merits further investigation. In the present study, 25 *R. heilongjiangensis* SEPs were identified using biotin-streptavidin affinity purification and 2D electrophoreses coupled with ESI-MS/MS. Most of these proteins were predicted to reside on the bacterial cell surface and play roles in bacterial pathogenesis. The recombinant SEPs fabricated on the microarray slide were probed with sera from mice infected with different rickettsial agents and 11 of them were recognized as major seroreactive antigens. Among the major seroreactive SEPs, GroEL, OmpA-2, OmpB-3, PrsA, RplY, RpsB, SurA and YbgF exhibited modest sensitivity and specificity for recognizing *R. heilongjiangensis* infection and/or spotted fever in the protein microarray assay and/or the ELISAs. Our results suggest that most of the *R. heilongjiangensis* SEPs have potential roles in bacterial pathogenicity and that some of them could become candidate molecules for serodiagnosis of and subunit vaccines antigens against FESF. Supporting Information {#s5} ====================== ###### **Primer pairs designed to amplify genes encoding the surface-exposed proteins of** ***Rickettsia heilongjiangensis*** **.** (XLS) ###### Click here for additional data file. ###### **Surface-exposed proteins of** ***Rickettsia heilongjiangensis*** **identified by ESI-MS/MS.** (XLS) ###### Click here for additional data file. ###### **Average fluorescence intensity of each surface-exposed protein probed with** ***Rickettsia heilongjiangensis*** **-infected mouse sera or FESF patient sera.** (XLS) ###### Click here for additional data file. ###### **Average fluorescence intensity of each major seroreactive surface-exposed protein probed with serum samples from mice infected with different rickettsial agents.** (XLS) ###### Click here for additional data file. ###### **BLASTP amino acid sequence alignment of** ***R. heilongjiangensis*** **RplY and homologous proteins from other rickettsial agents.** (XLS) ###### Click here for additional data file. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: BW YQ XX. Performed the experiments: YQ YJ JJ. Analyzed the data: YQ CD. Contributed reagents/materials/analysis tools: XW WG. Wrote the paper: YQ BW.

Introduction {#s1} ============ Dynamic cellular processes, such as the response to a signaling event, are governed by complex transcriptional regulatory networks. These networks typically involve a large number of transcription factors (TFs) that are activated in different combinations in order to produce a particular cellular response. The macrophage, a vital cell type of the mammalian immune system, marshals a variety of phenotypic responses to pathogenic challenge, such as secretion of pro-inflammatory mediators, phagocytosis and antigen presentation, stimulation of mucus production, and adherence. In the innate immune system, the first line of defense against infection, the macrophage\'s Toll-like receptors (TLRs) play a crucial role by recognizing distinct pathogen-associated molecular patterns (PAMPs), such as flagellin, lipopeptides, or double-stranded RNA [@pcbi.1000021-Underhill1],[@pcbi.1000021-Takeda1]. TLR signals are first channeled through adapter molecules (e.g., TICAM1/TRIF [@pcbi.1000021-Hoebe1],[@pcbi.1000021-Yamamoto1] and MyD88 [@pcbi.1000021-Adachi1]) and then through parallel cross-talking signal pathways. These activated pathways initiate a transcriptional program in which over 1,000 genes [@pcbi.1000021-Gilchrist1] and hundreds of TF genes [@pcbi.1000021-Nilsson1] can be differentially expressed, and which is tailored to the type of infection [@pcbi.1000021-Nau1],[@pcbi.1000021-Jenner1]. The transcriptional network underlying macrophage activation can exhibit many distinct steady-states which are associated with tissue- and infection-specific macrophage functions [@pcbi.1000021-Sasmono1]. The transcriptional response is also dynamic and characterized by temporal waves of gene activation [@pcbi.1000021-Gilchrist1],[@pcbi.1000021-Nilsson1],[@pcbi.1000021-Jenner1], each enriched for distinct sets of gene functions [@pcbi.1000021-Nilsson1],[@pcbi.1000021-Jenner1] and likely to be controlled by different combinations of transcriptional regulators [@pcbi.1000021-Gilchrist1],[@pcbi.1000021-Nilsson1]. Long-term, elucidating the transcriptional network underlying TLR-stimulated macrophage activation, and identifying key regulators and their functions, would greatly enhance our understanding of the innate immune response to infection and potentially yield new ideas for vaccine development. Computational analysis of high-throughput experimental data is proving increasingly useful in the inference of transcriptional regulatory interaction networks [@pcbi.1000021-Basso1]--[@pcbi.1000021-Zhao1] and in the identification and prioritization of potential regulators for targeted experimental validation [@pcbi.1000021-Gilchrist1],[@pcbi.1000021-Nilsson1]. Time-course microarray expression measurements have been used to infer dynamic transcriptional networks in yeast [@pcbi.1000021-Wang1],[@pcbi.1000021-Zhao1] and static "influence" networks in mammalian cell lines [@pcbi.1000021-Basso1]. In the context of primary macrophages, expression-based computational reconstruction of the transcriptional control logic underlying the activation program is not straightforward and progress is difficult to measure, for several reasons. First, transcriptional control within mammalian networks in general [@pcbi.1000021-VanBuren1], and for key TLR-responsive genes in particular [@pcbi.1000021-Nilsson1], is combinatorial. Second, many induced TFs are subject to post-translational activation [@pcbi.1000021-Honda1] and dynamic control of nuclear localization [@pcbi.1000021-Gilchrist1]. Third, targeted genetic perturbations are presently infeasible to perform on a large scale in a mammalian animal model, and expression knockdown (RNAi) is difficult in macrophages due to the tendency of the vector to stimulate TLRs. Finally, the few transcriptional regulatory interactions that have been validated through targeted experiments in TLR-stimulated primary macrophages are not available in a single "gold standard" dataset. Therefore, in the context of transcriptional regulation in the mammalian macrophage, with presently accessible expression data sets, large-scale computational inference is primarily useful for statistically identifying potential regulatory interactions, rather than as an inference tool for predicting the transcriptional control logic for specific target genes. For the reasons described above, in order to computationally infer transcriptional regulatory interactions in a mammalian system, it is necessary to include additional sources of evidence (beyond expression data) to constrain or inform the transcriptional network model selection. Computational scanning of the promoter sequences of clusters of co-expressed genes for known transcription factor binding site (TFBS) motifs has proved particularly valuable when combined with global expression data [@pcbi.1000021-Gilchrist1],[@pcbi.1000021-Chiang1],[@pcbi.1000021-Holloway1]. Recently, Nilsson *et al.* [@pcbi.1000021-Nilsson1] used a combination of expression clustering and promoter sequence scanning for TFBS motifs to construct an initial transcriptional network of the macrophage stimulated with the TLR4 stimulus lipopolysaccharide (LPS). Their work identified two novel regulators, but the clustering was based on an expression dataset with a single stimulus, limited biological replicates, and few time points. Moreover, TFBS motif scanning of co-expressed clusters, without utilizing expression dynamics, provides only a limited and static picture of the underlying transcriptional network. Many TFBS motifs are often recognized by multiple TFs, making difficult the unambiguous identification of the regulating TF from TFBS enrichment alone. Furthermore, because of the tendency of TFBS motifs to co-occur [@pcbi.1000021-Elkon1], it is difficult to determine from among a set of co-occurring motifs which associated TF is the most relevant to the condition-specific regulation of the target cluster. In the TLR-stimulated macrophage, core transcription factors already expressed in the cell (e.g., NFkB, AP1, and CREB) are rapidly activated and initiate transcriptional regulation of "second wave" TF genes [@pcbi.1000021-Gilchrist1]. Such transcriptionally regulated TF genes are key candidates for an integrated analysis combining TF-specific dynamic expression data and sequence-based motif scanning data. This work is concerned with using computational data integration to identify a set of core differentially expressed transcriptional regulators in the TLR-stimulated macrophage and, in the form of statistical associations, the clusters of co-expressed genes that they may regulate. The clusters are differentiated based on temporal and stimulus-specific activation, and in this sense, the inferred associations constitute a preliminary dynamic transcriptional network for the TLR-stimulated macrophage. To achieve this, we used a novel computational approach incorporating TFBS motif scanning and statistical inference based on time-course expression data across a diverse array of stimuli. Our approach involved four steps. (i) A set of genes was identified that were differentially expressed by wild-type macrophages under at least one TLR stimulation experiment. (ii) These genes were clustered based on their expression profiles across a wide range of conditions and strains, grouping genes based on the similarity of the timing and stimulus-dependence of their induction. Gene Ontology annotations were used to identify functional categories enriched within the gene clusters. (iii) Promoter sequences upstream of the genes within each cluster were scanned for a library of TFBS motifs, each recognized by at least one differentially expressed TF, to identify possible associations between TFs and gene clusters. (iv) Across eleven different time-course studies, dynamic expression profiles of TF genes and target genes were compared in order to identify possible causal influences between differentially expressed TF genes and clusters. Several techniques have been developed specifically for model inference from time-course expression data, notably dynamic Bayesian networks (DBN) [@pcbi.1000021-Zou1] and ODE-based model selection [@pcbi.1000021-Bonneau1]. However, the parametric complexity of these model classes makes it difficult to apply them to infer a network underlying a specific cellular perturbation (e.g., TLR activation in the macrophage) with a limited expression dataset. Here, potential transcriptional regulatory influence is inferred from time-course expression data using the time-lagged correlation (TLC) statistic, which has been used to infer biochemical interaction networks [@pcbi.1000021-Arkin1] as well as transcriptional networks [@pcbi.1000021-Agrawal1]--[@pcbi.1000021-Raab1]. The TLC has the advantage that it accounts for the time delay between differential expression of an induced TF and differential expression of a target gene. In contrast to standard correlation-based methods that identify co-expressed genes, the TLC method uses temporal ordering of expression to determine whether the time lag between two correlated genes is consistent with a causal interaction. We developed a novel method to identify the optimal time lag for each gene pair, and used a prior probability distribution of transcriptional time delays to score possible interactions. By combining the promoter scanning-based evidence with the evidence obtained by the time-lagged correlation analysis of the expression data, we were able to identify a network of statistically significant associations between 36 TF genes and 27 co-expressed clusters. Overall, 63% of differentially expressed genes are included in the network. The network provided insights into the temporal organization of the transcriptional response and into combinations of TFs that may act as key regulators of macrophage activation. Finally, the analysis identified a potential transcriptional regulator, TGIF1 (*Tgif1*), which was not previously known to play a role in macrophage activation. As a targeted experimental validation of the inferred network, two transcriptional regulators, p50 (a component of NFkB) and IRF1, were assayed for binding to *cis*-regulatory elements in LPS-stimulated macrophages using ChIP-on-chip, and were confirmed to bind the promoters of genes in four out of five predicted target clusters at significantly higher proportions than expected for a random set of TLR-responsive genes. Results {#s2} ======= Gene selection and clustering {#s2a} ----------------------------- To probe a diverse set of transcriptional responses of Toll-like receptor (TLR)-stimulated macrophages, primary bone marrow-derived macrophages (BMMs) from five mouse strains (wild-type and four mutant strains; see [Table S1](#pcbi.1000021.s019){ref-type="supplementary-material"}) were stimulated with six purified TLR agonists representing various pathogen-associated molecular patterns (PAMPs). The TLR agonists include bacterial-associated (lipopolysaccharide, Pam~2~CSK~4~, Pam~3~CSK~4~, CpG), viral-associated (poly I:C), and anti-viral (R848) stimuli, and are listed in [Table S2](#pcbi.1000021.s020){ref-type="supplementary-material"}. The mutant strains, which were included to increase the diversity of the TLR-stimulated gene expression dataset and to increase the number of time-course measurements used, consisted of null mutations of the two key adapter molecules for the TLR signaling pathway (TRIF [@pcbi.1000021-Hoebe1] and MyD88 [@pcbi.1000021-Adachi1]) and two TFs predicted to be involved in TLR activation (ATF3 [@pcbi.1000021-Hartman1] and CREM [@pcbi.1000021-Blendy1]). Genome-wide expression measurements of 45,037 probesets, representing 23,259 annotated genes, were made for time courses of up to 48 hours post-stimulation, using oligonucleotide microarrays (see [Materials and Methods](#s4){ref-type="sec"}). In all, expression measurements were made for 95 distinct combinations of strain, stimulus, and elapsed time (hereafter, "experiments"; see [Table S3](#pcbi.1000021.s021){ref-type="supplementary-material"}). Using a spline-based multivariate regression method specifically adapted for significance testing of temporal expression datasets [@pcbi.1000021-Storey1], annotated probesets were analyzed for differential expression across seven TLR-stimulated wild-type expression time-courses. After filtering for minimum absolute expression intensity and differential expression under at least one TLR-stimulation experiment (see [Materials and Methods](#s4){ref-type="sec"}), 1,960 probesets were identified as significantly differentially expressed, with each probeset mapped to a unique gene (see [Table S4](#pcbi.1000021.s022){ref-type="supplementary-material"}). Of these, 44% were found to be upregulated in LPS-stimulated wild-type macrophages. Additionally, a set of 80 TF genes (for which corresponding position-weight matrices are available in the TRANSFAC database [@pcbi.1000021-Wingender1]) were found to be differentially expressed in the TLR-stimulated wild-type macrophage ([Table S5](#pcbi.1000021.s023){ref-type="supplementary-material"}). Those of TF families with established relevance in macrophage activation included two NFkB [@pcbi.1000021-Hayden1] component genes (*Rel*, *Nfkb1*), three AP1 [@pcbi.1000021-Foletta1] components (*Jun*, *Junb*, *Fos*), two ATF family genes [@pcbi.1000021-Gilchrist1] (*Atf1, Atf3*), six IRF family TF genes (*Irf1/2/3/5/7/9*) [@pcbi.1000021-Honda1], and four STAT family TF genes [@pcbi.1000021-Ivashkiv1] (*Stat1/3/4/5a*). The 80 TF genes were taken to constitute the set of potential regulators in the TLR-stimulated macrophage network. Clustering was used to identify cohorts of genes that were co-expressed across the diverse set of TLR-stimulation experiments, based on the assumption that genes within a cluster are likely to share common *cis*-regulatory elements such as TF binding sites [@pcbi.1000021-Chiang1]. In order to focus on TF control of the *timing* and *stimulus specificity* of gene expression, genes were clustered based on the normalized profile of expression, rather than based on the fold-change. Expression measurements were transformed based on a single universal reference experiment (wild-type unstimulated macrophages) so that the transformed measurements would all lie between −1 and 1, with zero indicating the intensity in the reference experiment. This technique, which we call the signed difference ratio (SDR), has previously proved useful in clustering genes based on temporal expression in a mammalian system [@pcbi.1000021-Stanton1]. Each log~2~ intensity measurement *y~pj~* for probeset *p* and non-reference experiment *j*, was transformed to an SDR value *x~pj~* bywhere *j~R~* is the index of the global reference experiment (*j*′ has the same range of values as *j*). By construction, −1≤*x~pj~*≤1 for all *p* and *j*. A positive SDR value indicates higher expression than in the reference experiment, and a negative value indicates lower expression. The SDR-transformed log~2~ intensities of all 1,960 target genes across all 94 non-reference experiments were clustered using an unsupervised algorithm (*K*-means with Euclidean distance), with the number of clusters chosen using the Bayesian information criterion (BIC) [@pcbi.1000021-Hastie1] (see [Materials and Methods](#s4){ref-type="sec"}, and [Figure S1](#pcbi.1000021.s002){ref-type="supplementary-material"}). The target genes were partitioned into 32 clusters (see [Table S4](#pcbi.1000021.s022){ref-type="supplementary-material"}, column 5). The differences in temporal and stimulus-specific expression between the clusters are clearly visible in a heat-map representation of the SDR-transformed expression data (hereafter, "expression data") ([Figure 1](#pcbi-1000021-g001){ref-type="fig"}; see also [Figure S2](#pcbi.1000021.s003){ref-type="supplementary-material"} for the cluster-median expression heat-map). {ref-type="sec"}, Expression Clustering.](pcbi.1000021.g001){#pcbi-1000021-g001} The clusters ([Table S6](#pcbi.1000021.s024){ref-type="supplementary-material"}), which ranged in size from 18 to 113 genes, exhibit a significant diversity of timing and TLR-specificity of response. The wild-type LPS time-course was used to characterize the time scale for each cluster to respond transcriptionally (see [Materials and Methods](#s4){ref-type="sec"}, and [Table S6](#pcbi.1000021.s024){ref-type="supplementary-material"} columns 3--4). A small number of clusters reach peak induction within the first hour (C28, C27, C25, C26), but the majority of clusters (representing 55% of genes) respond between 2--4 hours. The temporal profiles of the clusters in wild-type BMMs under stimulation by LPS, Pam~3~CSK~4~, poly I:C, and R848 are shown in [Figure S3](#pcbi.1000021.s004){ref-type="supplementary-material"}, [Figure S4](#pcbi.1000021.s005){ref-type="supplementary-material"}, [Figure S5](#pcbi.1000021.s006){ref-type="supplementary-material"}, and [Figure S6](#pcbi.1000021.s007){ref-type="supplementary-material"}, respectively. The clusters exhibit distinct temporal profiles of transcriptional activation and repression that vary in the time of initial response and the duration of differential expression. Across all four stimuli, cluster C28 is induced first (and has sustained induction), followed by cluster C27 (which undergoes transient (2--3 h) upregulation), and then by induction of C25 and C26. Induction of C27 and C28 is delayed approximately 1 h under poly I:C stimulation, while C26 fails to fully induce under poly I:C. A comparison of the responses of clusters under 8 hours post-stimulation (LPS, Pam~3~CSK~4~, poly I:C, and R848) enabled the segregation of these clusters based on the signal transduction pathway through which they are likely primarily regulated ([Figure 2](#pcbi-1000021-g002){ref-type="fig"}). Groups include those primarily induced (C11, C12, C15, C17, C21, C26) and downregulated (C7, C29) by the MyD88-dependent pathway, and those primarily induced (C6, C8, C22, C24) and downregulated (C4, C5, C10, C20) by the TRIF-dependent pathway. Although "core early response" clusters C27 and C28 appear to be inducible through either signaling pathway, a comparison of the wild-type LPS vs. poly I:C response and of the wild-type vs. *Ticam1* ^(Lps2/Lps2)^ and *Myd88* ^(−/−)^ responses under LPS (see [Table S7](#pcbi.1000021.s025){ref-type="supplementary-material"}) together indicate that the MyD88-dependent pathway is responsible for the early response (within the first hour), and the TRIF-dependent pathway is responsible for sustaining the induction of these clusters. Early induced TF genes (*Egr1/2/3, Junb, Rel, Irf1*) also appear to be inducible through either pathway, from analysis of the LPS response in *Ticam1* ^(Lps2/Lps2)^ and *Myd88* ^(−/−)^ macrophages. {ref-type="sec"}, Expression Clustering). In the heat-map, green indicates downregulation, and red indicates upregulation. Colored subtrees of the dendrogram indicate specific inferences that can be made about the likely signaling pathway (MyD88-dependent, TRIF-dependent, or a combination) on which the transcriptional regulation of the cluster depends. The legend in the lower-left corner explains the color scheme for denoting the inferred signaling pathway-dependence of the clusters. Clusters without a color bar on the right appear to respond through either signaling pathway. The regulation of clusters C7, C11, C12, C15, C17, C21, C26, and C29 appears to be primarily MyD88-dependent; regulation of clusters C4, C5, C6, C8, C10, C20, C22, and C24 appears to be primarily TRIF-dependent; and clusters C23, C30, and C32 appear to be regulated oppositely by the two signaling pathways. This plot shows only the extremal differential response to TLR agonists; the clusters also differ in temporal expression (see [Figure S3](#pcbi.1000021.s004){ref-type="supplementary-material"}, [Figure S4](#pcbi.1000021.s005){ref-type="supplementary-material"}, [Figure S5](#pcbi.1000021.s006){ref-type="supplementary-material"}, and [Figure S6](#pcbi.1000021.s007){ref-type="supplementary-material"}).](pcbi.1000021.g002){#pcbi-1000021-g002} To characterize the functional role of each gene cluster in macrophage activation, gene ontology (GO) information was used to identify GO term enrichments within the gene clusters (see [Materials and Methods](#s4){ref-type="sec"}). The 460 GO term enrichments identified within the 32 gene clusters are listed in [Table S8](#pcbi.1000021.s026){ref-type="supplementary-material"}. Many of the downregulated gene clusters are enriched for cell cycle related genes (C1, C3, C7). Clusters C15, C25, and C28 appear to be enriched for cytokines--C28 includes the pro-inflammatory cytokine *Tnf* (TNFa) as well as *Ccl3*, *Ccl4*, *Cxcl1*, and *Cxcl2*; C25 includes the cytokines *Cxcl10* and *Il10*; and C15 includes the interleukin cytokine genes *Il1b*, *Il6*, and *Il12b*. Cluster C24, enriched for signal transduction genes, also includes the important cytokine *Ifnb1* (IFNb). The early-unregulated clusters, C24--28, show a high proportion of induced TFs and are enriched for TFs relative to the genome (see [Table S6](#pcbi.1000021.s024){ref-type="supplementary-material"} and [Materials and methods](#s4){ref-type="sec"}). Across clusters, the fraction of TFs was generally found to decrease with increasing induction time ([Figure 3](#pcbi-1000021-g003){ref-type="fig"}). Subsequent analysis focused on identifying statistically significant associations between the 80 differentially expressed TF genes and the 32 co-expressed clusters. {ref-type="sec"}, Expression Clustering). The horizontal dashed line indicates the average fraction of genes that are known transcription factors, among all annotated genes in the mouse genome (0.053, see [Materials and Methods](#s4){ref-type="sec"}, Selection of Transcription Factors). The slope of the best-fit line to the scatter plot is −3.84 (Pearson\'s *R* = −0.74).](pcbi.1000021.g003){#pcbi-1000021-g003} Expression dynamics analysis {#s2b} ---------------------------- Noting the high proportion of induced TFs in early-upregulated clusters, we chose a signal processing technique, the time-lagged correlation (TLC), to assess potential transcriptional regulatory interactions using the time-course expression data [@pcbi.1000021-Arkin1], [@pcbi.1000021-Agrawal1], [@pcbi.1000021-Kato1]--[@pcbi.1000021-Wu1]. The approach is based on the observation that when an induced TF affects a target gene\'s expression through its own differentially regulated mRNA level (and through its own differential protein expression), the induction of the target gene\'s mRNA expression will occur with a time lag relative to the induction of the regulator [@pcbi.1000021-Barrio1]--[@pcbi.1000021-Zak1]. This time lag is due to the combined effects of the translation, folding, nuclear translocation, and turnover time-scales for the regulatory protein, and the time scale for elongation of the target gene mRNA. In our application of the TLC method, both the correlation magnitude and the time lag are used to assess significance, as we describe below. Let *g* ~1~ denote a differentially expressed TF gene, and let *g* ~2~ denote a differentially expressed gene. We wish to estimate our degree of confidence in the null hypothesis, that *g* ~1~ does not transcriptionally regulate *g* ~2~, given time-course expression data for both genes. In the simplest case, the alternative hypothesis could be that *g* ~1~ codes for a TF protein that binds the promoter of *g* ~2~, thereby regulating its transcriptional activity. Let *t* be a fixed time lag for which the TLC between *g* ~1~ and *g* ~2~ is to be computed. Let *T* denote a set of discrete time points at which gene expression is measured, and let *T′* denote the set of time points *T*+*t*. Let *X~T~*(*g* ~1~) denote the vector of expression measurements of *g* ~1~ at the time points *T*, and let *X~T′~*(*g* ~1~) denote the measurements of *g* ~2~ at times *T′* (which can be estimated by interpolation). The time-lagged correlation (TLC) coefficient between *g* ~1~ and *g* ~2~ with time lag *t* is defined aswhere "cov" is the standard covariance. As with the standard correlation, a TLC that is close to 1 represents a perfect correlation, and a TLC that is close to −1 represents a perfect anti-correlation. This definition is easily extended to multiple time-courses (see [Materials and Methods](#s4){ref-type="sec"}). We note that although Equation 2 is defined in terms of *g* ~1~ being a TF, it can be applied to any gene pair, for example, to obtain a background distribution of TLC coefficients of gene pairs satisfying the null hypothesis. Two examples of a TF exhibiting a high time-lagged correlation with a target gene are shown in [Figure 4](#pcbi-1000021-g004){ref-type="fig"}. Both interactions (*Rel*→*Nfkb1* [@pcbi.1000021-Cogswell1] and *Irf7*→*Stat1* [@pcbi.1000021-Barnes1]) correspond to known transcriptional regulatory interactions, and in both cases, the correlation with zero time lag is poorer than the correlation obtained with a time lag. {ref-type="supplementary-material"}, *ρ~τ~* = 0.91 and *P* = 0.011; see [Materials and Methods](#s4){ref-type="sec"}, Time-lagged Correlation, for an explanation of the statistical test). The NFκB heterodimers c-REL-p50 and c-REL-p65 are known to regulate expression of *Nfkb1* [@pcbi.1000021-Cogswell1]. The correlation at zero time lag is 0.81. (B) *Irf7* and *Stat1*. The solid line shows the expression of *Irf7* (IRF7) and the dotted line shows the expression of *Stat1* (STAT1) in LPS-stimulated *Atf3* ^(−/−)^ macrophages. The genes exhibit a high time-lagged correlation with a time delay of 20 minutes (across the ten experiments, *ρ~τ~* = 0.96 and *P* = 0.002). The transcription factor IRF7 has been shown to regulate the *Stat1* gene expression in the innate immune response to viral infection [@pcbi.1000021-Barnes1]. The correlation at zero time lag is 0.95. (C) Time-lagged correlation coefficient and time-lagged correlation significance measure (see Equation 4) as a function of the time lag *τ*, for *Irf7* and *Stat1*. The peak value of *ρ~τ~* ^2^ occurs at *τ* = 10, but the peak significance value (taking into account the lag-specific null distribution) occurs at *τ* = 20 min.](pcbi.1000021.g004){#pcbi-1000021-g004} Assessing the significance of an observed sequence of time-lagged correlations between two genes (as a function of the time lag) as an indicator of possible transcriptional regulation necessitates formulating our prior expectation (i.e., prior probability distribution) for the time lag of a true transcriptional regulatory interaction. For a TF gene *g* ~1~ and a target gene *g* ~2~, the overall transcriptional regulatory time delay *t* ~c~ (where "c" stands for the combined gene-gene delay) can be decomposed as a sum of two contributions, one for translation of the TF and post-translational processing/translocation (∼10.5±4 min [@pcbi.1000021-Monk1],[@pcbi.1000021-Yu1]), and one for transcription and post-transcriptional processing of the target gene (∼20--40 min [@pcbi.1000021-Monk1],[@pcbi.1000021-Zak1]). The total delay *t* ~c~ was modeled using the gamma distribution with a mean value of 45 min and a variance of ∼250 min^2^ (see [Text S1](#pcbi.1000021.s001){ref-type="supplementary-material"}, Section 3). Because it is conditioned on the existence of a transcriptional regulatory interaction (TRI) between *g* ~1~ and *g* ~2~, we denote this probability distribution by *P*(*τ* ~c~\|*H̅* ~0~) (the symbol *H̅* ~0~ means that the null hypothesis, i.e., that there is no TRI, is false). This distribution was discretized to the set of time lags for which the TLC was computed, to obtain an estimate of the discrete probability for observing a given optimal time lag, *P*(*τ*\|*H̅* ~0~) (see [Figure S7](#pcbi.1000021.s008){ref-type="supplementary-material"}). These probabilities were then combined with the *P* value for the squared time-lagged correlation coefficient, *ρ~τ~* ^2^(*g* ~1~, *g* ~2~), whose derivation we describe next. For each pair (*g* ~1~,*g* ~2~) for which the TLC approach was to be applied, an "optimal time lag" *θ*(*g* ~1~,*g* ~2~) was selected, so that a single representative TLC could be obtained for the pair. The set of time lags and the set of time-course experiments to use were selected according to a constraint (imposed to minimize interpolation error) that the target gene expression at maximum time lag must be interpolated from at least three measurements. Based on this constraint, and taking into account the expected precision at which the optimal time lag can be estimated (±5 min, based on the replicate variability in the expression data--see [Materials and Methods](#s4){ref-type="sec"}), the set of time lags was chosen to be *t* ∈ {0, 10, 20, 30, 40, 50, 60, 70, 80 min}. Eleven time-course experiments satisfied the criteria (combining six stimuli and three genotypes, see [Table S9](#pcbi.1000021.s027){ref-type="supplementary-material"}). The TLC *ρ~τ~* ^2^(*g* ~1~, *g* ~2~) was computed for each of the *t* values, for each pair of genes, using data from all eleven time-course experiments combined (see [Materials and Methods](#s4){ref-type="sec"}). The next step was to determine the optimal time lag for (*g* ~1~,*g* ~2~) from the squared TLC coefficient *ρ~τ~* ^2^(*g* ~1~, *g* ~2~). It is not ideal to simply select the *t* at which *ρ~τ~* ^2^(*g* ~1~, *g* ~2~) is maximal, as some studies have done [@pcbi.1000021-Agrawal1],[@pcbi.1000021-Schmitt1],[@pcbi.1000021-Arkin2], because of two competing bias effects, as we now explain. Consider a pair of genes (*h* ~1~,*h* ~2~) satisfying the null hypothesis, and let *t* ~max~≡max(*T*), where *T* is the set of time points for a single time-course. In practice the expression of *h* ~2~ cannot be extrapolated beyond *t* ~max~, so the effective number of data points for computing the TLC *ρ~τ~* ^2^(*h* ~1~, *h* ~2~) is limited to the number of time points within *T* that are less than *t* ~max~−*t*. Thus, the number of measurements that can be used to compute the TLC is *t*-dependent, and the distribution of TLCs for pairs of genes satisfying the null hypothesis depends on *t*. Therefore, one will more frequently observe (by chance) a TLC exceeding a given value (say, 0.9), by selecting the largest possible *t*. In addition, the high degree of synchronization within the transcriptional response, as well as the fact that all the SDR-transformed expression levels are zero at the initial time point, result in a second bias towards zero time lag. This effect is strengthened as the number of time points in the data set (per time-course) decreases. Therefore, selecting the optimal *t*to maximize *ρ~τ~* ^2^(*g* ~1~, *g* ~2~) introduces an unwanted bias towards the smallest and largest *t*values investigated (see [Figure S8](#pcbi.1000021.s009){ref-type="supplementary-material"}), and against *t* values in the middle of the range of time lags (i.e., 20--60 min). To avoid the above-described bias, a background cumulative distribution of squared time-lagged correlation coefficient values, denoted by (where *p~t~* is the squared correlation *ρ~τ~* ^2^) was computed separately for each time lag *t*, using a large set *H* of gene pairs such that there is no direct transcriptional regulatory interaction (TRI) for each gene pair in the set (see [Materials and Methods](#s4){ref-type="sec"}). The functions were used to select the optimal time lag *θ*(*g* ~1~,*g* ~2~),and the fractional significance of the lag-specific squared correlation coefficient *ξ*(*g* ~1~,*g* ~2~),Making use of the discretized distribution *P*(*τ*\|*H̅* ~0~) defined above, a probability ratio *R*(*τ*) was computed as the ratio of the probability of the null hypothesis (that there is no direct TRI between *g* ~1~ and *g* ~2~) given the measured optimal time lag, to the marginal probability of the null hypothesis,It should be noted that the uncertainty in *q* due to the discretization of time lags (a practical necessity in the context of microarray-derived expression data) leads to uncertainty in the estimation of *R*(*t*). However, the effect of this uncertainty on the cluster-combined *P* value (see Equation 10 below) is small, due to the fact that time lag estimation errors for genes in a cluster are not strongly correlated. The marginal probability *P*(*τ*) was estimated from the optimal time lags of all gene pairs, and the marginal probability *P*(*H* ~0~) was estimated from data in the literature (see [Materials and Methods](#s4){ref-type="sec"}). Using this probability ratio, and in analogy with Fisher\'s method, a combined score for the gene pair (*g* ~1~,*g* ~2~) was constructed, taking into account both the optimal time lag *θ*(*g* ~1~,*g* ~2~) and the fractional lag-specific significance *ξ*(*g* ~1~,*g* ~2~),Using the cumulative distribution of *s* scores for gene pairs satisfying the null hypothesis, the significance of the association between *g* ~1~ and *g* ~2~ based on expression data can be computed,This formula was applied for all pairs (*g* ~1~,*g* ~2~) where *g* ~1~ ranged over the set of 80 TFs, *g* ~2~ ranged over the set of all 1,960 differentially expressed genes, and *g* ~1~≠*g* ~2~ (see [Materials and Methods](#s4){ref-type="sec"}). The expression data for the TFs are provided in [Table S10](#pcbi.1000021.s028){ref-type="supplementary-material"} and the expression data for all 1,960 differentially expressed genes are provided in [Table S4](#pcbi.1000021.s022){ref-type="supplementary-material"}. To estimate the overall significance (based on time-course expression data) of the association between a TF gene *f* and a cluster *C*, the *P* values *P* ^tlc^(*f*,*g*) were combined into a *P* value for the cluster, *P* ^exp^(*f*,*C*). For each pair (*f*,*C*), a Fisher score *F* ^exp^(*f*,*C*) was computed,where *C*\\{*f*} means that if the TF gene *f* was a member of cluster *C*, the self-association *P* ^tlc^(*f*,*f*) was excluded. For each cluster *C*, the number of degrees of freedom, denoted by *d*(*C*), was estimated using *K*-means clustering (see [Materials and methods](#s4){ref-type="sec"}). The *d*(*C*) values were used to obtain a TF-to-cluster *P* value, *P* ^exp^(*f*,*C*), using a χ^2^ test (see [Materials and methods](#s4){ref-type="sec"}). The number of pairs for which *P* ^exp^(*f*,*C*)≤10^−3^, was 23. The differential expression levels for the strongest (TF,cluster) pairs in wild-type time-courses following stimulation by LPS (one of the time-courses used for the TLC analysis; see [Table S9](#pcbi.1000021.s027){ref-type="supplementary-material"}) are shown in [Figure S9](#pcbi.1000021.s010){ref-type="supplementary-material"}. They show a high degree of correlation between the TF gene and target cluster. The distribution of *P* ^exp^(*f*,*C*) over all TF-to-cluster pairs, and the estimated false discovery rate (FDR), are shown in [Figure S10](#pcbi.1000021.s011){ref-type="supplementary-material"}. Promoter scanning of co-expressed gene clusters {#s2c} ----------------------------------------------- To provide an independent source of evidence for association between a differentially expressed TF gene and a co-expressed gene cluster, the promoters of differentially expressed genes were scanned using position-weight matrices (PWMs) representing motifs recognized by murine TFs. A motif was selected if it is recognized by at least one TF of which at least one component protein was differentially expressed in the expression dataset, ensuring that the TF had at least one expression profile that could be compared with (potential) target genes using the TLC. For each PWM, the fraction of genes with at least one above-threshold match within the promoter was computed, within a reference set of all genes detected as expressed within the TLR-stimulated macrophage, and within each co-expressed gene cluster. A total of 150 position-weight matrices were selected from the TRANSFAC database [@pcbi.1000021-Wingender1] for motif scanning, corresponding to the 80 differentially expressed murine TF genes (see [Table S5](#pcbi.1000021.s023){ref-type="supplementary-material"}, and [Materials and Methods](#s4){ref-type="sec"}). Promoter sequences 2 kbp upstream of the transcription start site were obtained for 1,713 out of the 1,960 differentially expressed genes, and for 7,492 out of 8,788 expressed genes (used as a reference set; see [Materials and Methods](#s4){ref-type="sec"}) from the UCSC genome annotation database [@pcbi.1000021-Kuhn1]. For each PWM, a minimum match score was determined at which the PWM had a match on average once per 10 kb, within a set of 7,503 promoter sequences for genes not detectably expressed in the macrophage (to avoid biasing the match score threshold calculation with true TF targets; see [Materials and Methods](#s4){ref-type="sec"}). Using these PWM match score thresholds, promoters were scanned within the reference set of genes, and within each co-expressed cluster of genes. The distribution of distances of matches from the transcription start site ([Figure S11](#pcbi.1000021.s012){ref-type="supplementary-material"}) has a median of 537 bp. As a next step towards inferring a transcriptional network, enrichments of TFBS motifs were computed for individual gene clusters. For each cluster *C* and position-weight matrix *m*, enrichment statistics were computed based on the fraction of genes in *C* possessing at least one match for *m*. For each pair (*m*,*C*) for which the fraction of genes containing a match for *m* within the cluster *C* was greater than in the reference set of genes, a *P* value was computed using Fisher\'s exact test (see [Materials and Methods](#s4){ref-type="sec"}, and [@pcbi.1000021-Frith1]) and denoted by *P* ^scan^(*m*,*C*). This *P* value represented the significance of the enrichment of matrix *m* within the promoters of cluster *C*, relative to the reference set of promoters (expressed genes). A matrix representation of the strongest motif enrichments (56 associations with *P* ^scan^(*m*,*C*)≤10^−2^) with the clusters grouped by expression similarity ([Figure 5](#pcbi-1000021-g005){ref-type="fig"}) reveals several associations between TF motifs and patterns of differential expression. First, NFκB and IRF recognition elements are associated with upregulated clusters, while E2F and MYCMAX elements are associated with downregulated clusters. The IRF element was strongly associated with TRIF-dependent cluster C6 and STAT1 was strongly associated with C22. Many TF motifs were associated with the core early response cluster C27, including AP1, CREB/ATF, EGR, PEBP, and PPARA. The quantitative results of the cluster-wise statistical tests (numbers of matches and *P* values) are provided in [Table S11](#pcbi.1000021.s029){ref-type="supplementary-material"}. {ref-type="fig"}, and thus are grouped hierarchically by similarity of their extremal expression fold-change under the four TLR agonists LPS, Pam~3~CSK~4~, poly I:C, and R848. Each motif (row) is associated with one or more position-weight matrices (the V\$ prefix and numeric suffixes are omitted, and results for multiple position-weight matrices representing the same motif were combined for each column, by taking the matrix with the maximum number of matches within the indicated cluster). Each colored block in the matrix indicates pair of a motif and target cluster for which the fraction of genes in the cluster with a motif match, is enriched relative to the overall fraction of genes expressed in the macrophage that possess the motif (*P*≤10^−2^, Fisher\'s exact test). The color of each matrix element (block) in the interior of the figure indicates the fraction scanned of genes within the cluster containing at least one match for the indicated motif. The number of scanned genes within the cluster that contained a match for the indicated motif is shown in yellow typeface. The red/green colored blocks above the top horizontal axis shows whether each cluster is upregulated (red) or downregulated (green) at its most extremal fold-change under stimulation with the aforementioned TLR agonists. The hatched green/red pattern indicates a cluster whose extremal fold-change direction (up/down) is stimulus-dependent (see [Figure 2](#pcbi-1000021-g002){ref-type="fig"}). The colored (blue, cyan, orange, yellow, purple) blocks above the top of the matrix indicate the likely pathway through which the cluster is differentially expressed; the color scheme corresponds to that shown in the dendrogram in [Figure 2](#pcbi-1000021-g002){ref-type="fig"}.](pcbi.1000021.g005){#pcbi-1000021-g005} To enable integration of the promoter scanning evidence with the time-lagged correlation evidence, PWMs that were enriched for matches within gene clusters, were mapped to differentially expressed TF genes as follows. For each PWM *m*, a list of genes coding for TFs (or TF components) that bind the motif corresponding to *m* were obtained from a TRANSFAC-derived mapping (see [Materials and Methods](#s4){ref-type="sec"}). For each TF gene *f* and cluster *C*, a *P* value for the association between *f* and *C* based on promoter scanning evidence, *P* ^scan^(*f*,*C*), was defined as the minimum over all *P* ^scan^(*m*,*C*) for all matrices *m* that are associated with the TF gene *f*. The distribution of the resulting *P* values and the false discovery rate (as a function of *P* value) are shown in [Figure S12](#pcbi.1000021.s013){ref-type="supplementary-material"}. A total of 31 factor-to-cluster associations were identified with *P* ^scan^(*f,C*)≤10^−3^, indicating a statistical power that is slightly higher than with the TLC-based evidence. Data integration and network extraction {#s2d} --------------------------------------- To identify the set of all possible TF gene-to-target interactions consistent with motif scanning evidence, for each TFBS motif match within the promoter of a target gene, the time-lagged correlation was computed for all possible TF genes that map to the TFBS motif. The resulting list of 54,253 pairs (*f*,*g*) of TF gene *f* and target gene *g*, provided as [Table S12](#pcbi.1000021.s030){ref-type="supplementary-material"}, shows that many known transcriptional regulatory interactions have high ranking based on time-lagged correlation--for example, NFκB/*Rel* associated with *Icam1* [@pcbi.1000021-Rahman1] and *Cebpd* associated with *Il6* [@pcbi.1000021-Hu1]. Although the TLC-ranked list of motif targets has some potential utility for identifying specific transcriptional regulatory interactions, even the high-ranking elements of the list will contain many false positives (and will miss many true transcriptional regulatory interactions) due to the uncertainty in motif PWMs and the prevalence of post-translational regulation that may obscure the time-lagged correlation. Therefore, further data reduction is necessary to gain insight into the global transcriptional program of the TLR-stimulated macrophage. By using a statistical test that compares the relative frequency of motif occurrence within a cluster relative to a background set of genes, a more reliable estimate of TF association with a co-expressed cluster can be obtained. To construct a combined transcriptional network of the TLR-stimulated macrophage, *P* values for associations between TF genes and co-expressed gene clusters based on expression dynamics and promoter scanning were combined. For each pair (*f*,*C*) where *f* is one of 80 TF genes and *C* is one of 32 gene clusters, a combined *P* value *P* ^comb^(*f*,*C*) was computed from the *P* values for the scanning and expression evidences, *P* ^scan^(*f*,*C*) and *P* ^exp^(*f*,*C*). The *P* values were combined using Fisher\'s method (see [Materials and Methods](#s4){ref-type="sec"}), a standard tool for meta-analysis of independent tests of a hypothesis [@pcbi.1000021-Hwang1]. TF-cluster pairs were then ordered by increasing *P* value *P* ^comb^(*f*,*C*), and a cutoff was selected so that the estimated false discovery rate did not exceed 0.025 (resulting in a cutoff *P* ^comb^(*f*,*C*)≤0.0248). Additionally, two filtering criteria were imposed: (i) *P* ^scan^(*f*,*C*)≤0.05, to ensure that there is a minimal enrichment of TFBS; and (ii) a cluster-average optimal time lag between *f* and *C* that was greater than 10 min, i.e., 〈*θ*〉*~f,C~*≥10 min (see [Materials and Methods](#s4){ref-type="sec"}). A scatter plot of the *P* values for the two evidences is shown in [Figure S13](#pcbi.1000021.s014){ref-type="supplementary-material"}, and indicates that for the data points that were rejected based on the *P* ^comb^(*f*,*C*) cutoff, no dependency between the evidences is evident. A total of 90 interactions involving 36 TF genes and 27 clusters (comprising 86% differentially expressed genes), were accepted based on the above criteria (see [Table 1](#pcbi-1000021-t001){ref-type="table"}). If the TLC *P* values were not included, and if the same rate of false discovery were imposed, the network would be significantly less parsimonious (∼150 interactions), due to the large number of TF gene families that map to a common motif. Overall network coverage was estimated by taking the fraction of differentially expressed genes that (i) are members of the 27 clusters in the network; and (ii) possess a match for a motif recognized by one or more of the TFs associated with the cluster. From this estimate the network contains 1,232 genes, or 63% of the 1,960 genes that are differentially expressed under TLR stimulation. 10.1371/journal.pcbi.1000021.t001 ###### Network of inferred transcription factor--cluster associations {#pcbi-1000021-t001-1} Clust TF TF Clust -log~10~ *P* ^scan^ Position-Weight Matrix Name FracBind Scan Hits -log~10~ *P* ^exp^ \<*θ*\> Mean Corr ------- ---------- ---------- --------------------- ----------------------------- ---------- ----------- -------------------- --------- ----------- 1 *Cebpg* 22 5.74 V\$CEBPGAMMA_Q6 0.41 40 0.03 13.5 −0.21 1 *E2f1* 2 2.84 V\$E2F_02 0.44 43 0.82 75.4 0.72 1 *E2f7* 3 1.41 V\$E2F1_Q4_01 0.41 40 4.42 73.4 0.86 1 *Irf2* 13 1.60 V\$IRF_Q6_01 0.32 31 1.35 68.8 −0.74 1 *Irf7* 6 1.60 V\$IRF_Q6_01 0.32 31 3.26 68.7 −0.83 1 *Isgf3g* 6 1.60 V\$IRF_Q6_01 0.32 31 2.13 71.3 −0.80 1 *Mef2a* 2 2.73 V\$MEF2_Q6_01 0.33 32 1.50 75.0 0.78 1 *Mef2c* 16 2.73 V\$MEF2_Q6_01 0.33 32 2.23 67.4 0.81 1 *Nfyc* 4 6.08 V\$NFY_Q6 0.48 47 0.17 78.1 0.59 2 *E2f1* 2 3.68 V\$E2F_02 0.46 46 3.16 34.2 0.85 2 *E2f6* 10 2.97 V\$E2F_03 0.47 47 2.72 65.9 0.82 2 *E2f7* 3 2.97 V\$E2F_03 0.47 47 3.62 33.1 0.85 2 *Myc* 20 1.95 V\$MYCMAX_01 0.34 34 1.36 18.4 0.74 2 *Rxra* 14 2.55 V\$LXR_DR4_Q3 0.38 38 1.61 57.6 0.77 3 *E2f1* 2 3.86 V\$E2F_Q6_01 0.53 54 2.39 47.2 0.83 3 *E2f6* 10 3.86 V\$E2F_Q6_01 0.53 54 2.26 69.0 0.81 3 *E2f7* 3 3.86 V\$E2F_Q6_01 0.53 54 8.00 33.5 0.93 3 *Myc* 20 1.32 V\$MYCMAX_03 0.32 33 1.62 16.1 0.78 3 *Nfic* 19 3.98 V\$NF1_Q6 0.46 47 0.26 77.6 0.63 3 *Nfyc* 4 7.73 V\$NFY_Q6_01 0.54 55 0.28 58.3 0.65 3 *Rxra* 14 1.36 V\$LXR_DR4_Q3 0.33 34 1.24 64.0 0.77 3 *Stat1* 6 2.12 V\$STAT1_03 0.41 42 2.95 12.1 −0.86 4 *Cebpa* 19 2.05 V\$CEBP_Q2 0.34 34 0.43 70.0 0.65 4 *Foxm1* 3 6.18 V\$FOXM1_01 0.41 41 0.63 52.6 0.68 4 *Mef2a* 2 1.68 V\$MEF2_04 0.30 30 1.56 40.2 0.79 4 *Myc* 20 1.99 V\$MYCMAX_B 0.44 44 0.63 45.5 0.68 4 *Nfyc* 4 1.43 V\$NFY_Q6_01 0.36 36 1.06 68.0 0.74 4 *Tgif1* 27 3.12 V\$TGIF_01 0.32 32 0.16 42.5 0.10 5 *E2f1* 2 3.85 V\$E2F1_Q6_01 0.52 45 2.31 50.3 0.81 5 *E2f6* 10 2.71 V\$E2F_03 0.48 41 2.84 72.5 0.81 5 *E2f7* 3 2.71 V\$E2F_03 0.48 41 1.27 47.3 0.76 5 *Myc* 20 2.54 V\$MYCMAX_B 0.48 41 1.00 27.9 0.69 5 *Rxra* 14 1.87 V\$PPARA_02 0.30 26 1.72 63.7 0.77 6 *Irf1* 25 3.65 V\$IRF_Q6 0.40 35 0.09 79.1 0.56 6 *Irf2* 13 3.65 V\$IRF_Q6 0.40 35 3.32 44.3 0.81 6 *Irf3* 12 3.65 V\$IRF_Q6 0.40 35 0.05 35.4 −0.09 6 *Irf5* 6 3.65 V\$IRF_Q6 0.40 35 1.21 68.4 0.75 6 *Irf7* 6 3.65 V\$IRF_Q6 0.40 35 5.20 26.6 0.88 6 *Isgf3g* 6 1.84 V\$IRF_Q6_01 0.33 29 3.24 25.0 0.84 7 *Pou2f2* 9 2.10 V\$OCT_C 0.32 24 0.42 45.9 −0.64 9 *Myc* 20 1.54 V\$MYC_Q2 0.36 26 1.26 29.3 −0.67 10 *Atf1* 14 1.73 V\$CREB_Q3 0.34 25 0.86 38.1 0.73 10 *Myc* 20 2.18 V\$MYCMAX_B 0.47 35 0.43 11.2 0.68 10 *Nfyc* 4 1.67 V\$NFY_Q6_01 0.39 29 1.31 35.8 0.78 10 *Nr3c1* 6 2.79 V\$GR_Q6_01 0.34 25 1.86 48.5 −0.79 11 *Fos* 27 3.23 V\$AP1_Q2_01 0.45 29 0.11 47.0 −0.05 11 *Jun* 20 3.23 V\$AP1_Q2_01 0.45 29 0.20 41.4 −0.37 11 *Junb* 28 3.23 V\$AP1_Q2_01 0.45 29 0.05 70.3 0.24 13 *Foxo3a* 14 2.41 V\$FOXO3_01 0.37 20 0.77 14.2 −0.75 13 *Irf1* 25 7.47 V\$IRF_Q6_01 0.57 31 0.32 77.5 0.64 13 *Irf3* 12 7.47 V\$IRF_Q6_01 0.57 31 0.07 42.8 0.07 13 *Irf5* 6 7.47 V\$IRF_Q6_01 0.57 31 0.49 22.7 0.71 13 *Nfkb1* 15 2.32 V\$NFKB_Q6_01 0.41 22 2.41 31.3 0.82 13 *Rel* 25 1.65 V\$CREL_01 0.37 20 0.88 71.9 0.72 14 *Nfkb1* 15 2.14 V\$NFKB_C 0.34 19 3.37 51.3 −0.84 15 *Rel* 25 2.42 V\$CREL_01 0.42 21 1.78 29.0 0.78 16 *Cebpa* 19 1.43 V\$CEBP_C 0.32 16 1.05 73.4 0.74 16 *E2f1* 2 1.78 V\$E2F_01 0.40 20 2.52 52.3 0.80 16 *Jun* 20 2.39 V\$AP1_Q2_01 0.44 22 0.14 59.3 0.43 16 *Myc* 20 1.90 V\$MYCMAX_02 0.40 20 0.83 20.5 0.63 16 *Rxra* 14 1.32 V\$FXR_IR1_Q6 0.30 15 1.33 63.8 0.75 17 *Nfatc1* 14 1.71 V\$NFAT_Q4_01 0.36 18 1.78 33.6 −0.79 17 *Nfatc2* 14 1.71 V\$NFAT_Q4_01 0.36 18 1.60 12.2 −0.80 17 *Nfkb1* 15 2.02 V\$NFKB_Q6 0.36 18 2.25 60.3 0.80 17 *Sfpi1* 17 1.35 V\$ETS_Q6 0.42 21 1.34 14.8 0.78 18 *Pou2f2* 9 1.78 V\$OCT_Q6 0.33 16 0.68 53.9 −0.67 19 *Nr3c1* 6 1.62 V\$PR_Q2 0.33 16 1.24 37.2 −0.74 19 *Rxra* 14 1.61 V\$T3R_Q6 0.37 18 1.65 14.6 0.79 19 *Zfp161* 19 2.98 V\$ZF5_01 0.55 27 0.90 49.8 0.73 20 *E2f7* 3 2.11 V\$E2F_03 0.50 24 0.64 66.5 0.66 20 *Myc* 20 2.36 V\$MYCMAX_B 0.52 25 0.66 53.6 0.64 21 *Nfkb1* 15 2.13 V\$NFKAPPAB_01 0.38 15 2.25 28.2 0.78 22 *Stat1* 6 3.22 V\$STAT1_01 0.58 19 0.16 55.2 0.39 23 *E2f1* 2 3.24 V\$E2F1_Q4_01 0.60 21 0.08 47.6 0.54 23 *E2f6* 10 3.24 V\$E2F1_Q4_01 0.60 21 0.48 46.0 0.66 23 *E2f7* 3 3.24 V\$E2F1_Q4_01 0.60 21 0.05 50.7 0.49 25 *Irf1* 25 1.41 V\$IRF_Q6_01 0.40 10 1.42 15.1 0.79 26 *Cebpa* 19 2.50 V\$CEBP_01 0.45 14 0.01 52.6 0.14 26 *Tgif1* 27 2.14 V\$TGIF_01 0.39 12 0.28 49.4 0.36 27 *Atf1* 14 2.57 V\$CREBATF_Q6 0.58 14 0.25 69.0 0.52 27 *Cbfb* 4 2.39 V\$PEBP_Q6 0.50 12 0.45 65.3 0.52 27 *E2f7* 3 1.62 V\$E2F_Q4_01 0.54 13 1.01 61.0 0.63 27 *Egr1* 27 2.62 V\$KROX_Q6 0.58 14 1.37 16.0 0.75 27 *Egr2* 27 2.62 V\$KROX_Q6 0.58 14 1.16 11.7 0.75 27 *Jun* 20 2.63 V\$CREBP1CJUN_01 0.54 13 0.24 47.3 0.46 27 *Rxra* 14 2.46 V\$PPARA_02 0.46 11 0.61 63.3 0.53 28 *E2f1* 2 2.75 V\$E2F_01 0.54 14 0.05 37.9 −0.05 28 *Nfkb1* 15 4.48 V\$NFKAPPAB_01 0.58 15 0.07 25.7 0.27 29 *Cebpa* 19 2.48 V\$CEBP_01 0.48 12 0.04 65.9 −0.11 31 *Mef2a* 2 2.39 V\$MEF2_03 0.58 7 0.16 35.9 −0.51 Column 1 indicates the target gene cluster. Column 2 indicates the transcription factor gene that is associated with the cluster, based on the two sources of evidence. Column 3 indicates the cluster of which the transcription factor gene is a member. Column 4 indicates the -log~10~ *P* value of the promoter scanning-based evidence. Column 5 indicates the name of the position-weight matrix that had the smallest scanning-based *P* value of association with the cluster, for the indicated transcription factor gene (the "V\$" prefix is not shown). Column 6 indicates the fraction of scanned genes within the cluster that had at least one match for the indicated position-weight matrix. Column 7 contains the number of scanned genes within the cluster that had at least one match for the indicated position-weight matrix. Column 8 indicates the negative log~10~ *P* value of the time-lagged correlation evidence. Column 9 indicates the cluster-wide average time lag *θ* with respect to the indicated transcription factor gene. Column 10 contains the average optimal time-lagged correlation between the indicated transcription factor gene, and the genes within the cluster. The distribution of the number of targets regulated by TFs, the so-called out-degree distribution of the transcriptional network, is one key measure of the network\'s interconnectedness [@pcbi.1000021-Barabasi1]. For each TF that was included in the transcriptional network, the number of targets was estimated using the promoter scanning data (see [Materials and Methods](#s4){ref-type="sec"}). The out-degree varied approximately 20-fold over the set of 36 TF genes ([Figure S14](#pcbi.1000021.s015){ref-type="supplementary-material"}). The transcription factor MYC (which is involved in development and cellular differentiation [@pcbi.1000021-Mukherjee1]) was found to be the most highly connected in the network (consistent with the high out-degree for MYC found in [@pcbi.1000021-Basso1]), followed by members of the E2F family of TFs (believed to play a role in cell cycle regulation [@pcbi.1000021-Attwooll1]). Other highly connected TFs include NFYC (a repressor in the TGFβ signaling pathway [@pcbi.1000021-Chen2] and member of a TF family involved in monocyte differentiation [@pcbi.1000021-Marziali1]) and RXRA (a component of heterodimeric TFs that regulate inflammatory signaling and cholesterol metabolism [@pcbi.1000021-Castrillo1]). Also strongly connected in the network are the NFκB TF family members cREL and NFKB1/p50 (key early regulators of the immune response [@pcbi.1000021-Shakhov1]); the IRF family members IRF1, IRF3, IRF5, IRF7, and IRF9 (regulators of interferon-induced immune response [@pcbi.1000021-Honda1]); and STAT1 (a key regulator of apoptosis and mediator of interferon signaling [@pcbi.1000021-Kim1]). Both the IRF and E2F family TFs had strong *P* values for association with target clusters ([Figure S14](#pcbi.1000021.s015){ref-type="supplementary-material"}). The out degree distribution appears to be scale-free, consistent with previous reports for mammalian networks [@pcbi.1000021-Basso1],[@pcbi.1000021-Potapov1]. The number of TF genes associated each cluster (in degree) ranged from 1 to 9, with an average in-degree of 3.3. To reveal patterns among TFs that may regulate multiple clusters, the connections between the 36 TFs and the 27 clusters in the inferred network were arranged in a matrix in which each row represents an induced TF and each column represents a cluster of differentially expressed genes ([Figure 6](#pcbi-1000021-g006){ref-type="fig"}). Both the TFs and clusters were divided into subsets that are induced or repressed under LPS stimulation, and ordered within these subsets based on the time of 25% differential expression under LPS (see [Materials and Methods](#s4){ref-type="sec"}). Thus, the matrix is divided into quadrants; for example, the upper left quadrant contains connections between induced TF genes and induced clusters, and the lower-right quadrant contains connections between downregulated TF genes and downregulated clusters. The upper left and lower right quadrants contain primarily positive correlations, with most anti-correlated connections found in the upper right and lower left quadrants. In the upper left quadrant, the connections generally fall along an arc indicating the temporal sequence of TF gene activation. The anti-correlated "off arc" connections within this quadrant generally indicate the association between the falling edge of a transiently induced TF gene and the rising edge of a late-induced gene cluster. The only correlated "off arc" connections within this quadrant (*Nfkb1*→C28, and *Junb*→C11) have weak time-lagged correlation evidence, but a very significant motif scanning *P* value. In contrast, the downregulated gene clusters and TF genes are not as stratified as the upregulated clusters in terms of the time of differential expression, and thus associations appear throughout the lower-right quadrant. {ref-type="sec"}, Expression Clustering). Transcription factor genes are ordered based on the LPS response time. The vertical gray line separates upregulated clusters (left half) from downregulated clusters (right half). The horizontal gray line separates upregulated transcription factors (top) from downregulated transcription factors (bottom). An orange or blue square indicates a statistically significant association between the transcription factor gene and the cluster, based on both promoter scanning and expression dynamics. An orange solid rectangle represents a positive average time-lagged correlation with genes in the cluster; a blue solid rectangle represents a negative average time-lagged correlation. (B) The red-green matrix is a heat-map showing transcription factor gene expression. The color indicates the normalized differential expression of the indicated transcription factor gene (over time), in LPS-stimulated wild-type macrophages (SDR, see Equation 1). Red indicates upregulation relative to unstimulated macrophages and green indicates downregulation. A diamond symbol indicates the transcription factor response time. (C) Each column of the red-green matrix indicates the median normalized differential expression of the genes in the indicated cluster (over time), in LPS-stimulated wild-type macrophages. The diamond indicates the average LPS response time of the genes within the cluster.](pcbi.1000021.g006){#pcbi-1000021-g006} The network of associations between TF genes and clusters (based on combined scanning and expression evidence) directly leads to hypotheses regarding TF regulation of clusters. For example, a statistical association between any of the TF genes *Jun*, *Junb*, or *Fos* and a cluster would suggest a hypothesis that the TF AP1 regulates that cluster. The network also recapitulates several known transcriptional regulatory interactions. First, the NFκB component *Rel* is associated with C15, which is enriched for cytokines and contains many NFκB targets including *Nfkb1* [@pcbi.1000021-Cogswell1], *Il6*, and *Il12b* [@pcbi.1000021-Gilchrist1]. Second, *Jun*, a component of AP1 (a regulator of stress response such as response to ultraviolet radiation or pathogenic insult [@pcbi.1000021-Chastel1]), is associated with C27, an early-upregulated cluster that is enriched for cell cycle-related genes and genes involved in the DNA damage response. Furthermore, C27 contains *Egr1*, which is a known target of AP1 under genotoxic stress conditions [@pcbi.1000021-Chastel1]. Third, IRF1 is strongly associated with the antiviral cluster C13, which contains the validated IRF1 target gene, *Ccl5* [@pcbi.1000021-Liu1]. The network also includes the TF genes *Egr1* (a key regulator of LPS-induced cytokine signaling [@pcbi.1000021-Mostecki1]) and *Egr2* (implicated in adhesion and phagocytosis [@pcbi.1000021-Hirano1] as well as cell proliferation [@pcbi.1000021-Chavrier1]) as regulators of C27. Finally, the TF gene *Sfpi1* (PU.1) is associated with C17, an induced gene cluster enriched for endosome-associated genes (PU.1 over-expression is known to block viral escape from the endosome [@pcbi.1000021-Carey1]). Several interactions in the network were detected only through the integration of expression data with promoter scanning evidence. For example, based on scanning evidence alone, with a FDR of 0.1 (*P* ^scan^≤0.0033), the association between *Nfkb1* and C17 would not have been detected, but by including the effect of the strong TLCs between *Nfkb1* and C17 genes, an association between *Nfkb1* and C17 was detected. As a second example, the network includes an association between the TF gene *Irf1* and cluster C25; based on promoter scanning evidence alone, only a general association of the IRF family with the cluster would have been possible (see [Table 1](#pcbi-1000021-t001){ref-type="table"}). In order to investigate the possible co-operative regulation of clusters by TFs in the network, protein interactions were obtained for human orthologs of protein units associated with the 36 TF genes shown in [Figure 6](#pcbi-1000021-g006){ref-type="fig"}. Protein interactions between the TFs were obtained from the Human Protein Reference Database [@pcbi.1000021-Peri1] and the Biomolecular Interaction Network Database [@pcbi.1000021-Alfarano1] (see [Materials and Methods](#s4){ref-type="sec"}). The resulting interaction diagram, shown in [Figure S15](#pcbi.1000021.s016){ref-type="supplementary-material"}, reveals that upregulated TFs are highly interconnected at the level of protein-protein interactions [@pcbi.1000021-Gilchrist1]. Furthermore, the diagram shows 15 pairs of interacting TFs whose corresponding genes co-associate with clusters in the network. An example corresponding to a known transcriptional complex is the pair c-JUN (an AP1 component) and EGR1 [@pcbi.1000021-Barthel1]; both are associated with C27. A notable induced TF gene in the network is *Tgif1* (TGIF1, or TG-interacting factor 1, named for the core TGIF1 binding sequence, 5′-TGTCA-3′ [@pcbi.1000021-Bertolino1]), a transcriptional repressor in the TGFβ signaling pathway [@pcbi.1000021-Bartholin1]. TGIF1 has not been previously implicated in classical macrophage activation. It is associated (*P* ^scan^\<0.01) with C26, a cluster containing genes involved in immune response, ubiquitin cycle, and leukocyte activation. Specifically, C26 contains the cytokines *Csf2* (which stimulates differentiation of macrophages and granulocytes, and is pro-inflammatory [@pcbi.1000021-Hamilton1]) and *Gm1960* (a mediator of neutrophil chemotaxis [@pcbi.1000021-Nolan1]). The *Csf2* promoter appears to have a TGIF1 binding site motif match (match score\>0.96) in the region (−254,−244) relative to the transcription start site, and *Gm1960* also has three TGIF1 motif matches approximately 1.5 kbp upstream of the start site (best match score\>0.95). In humans, TGIF1 is known to interact with several protein members of the SMAD/AP1 transcriptional complex ([Figure S16](#pcbi.1000021.s017){ref-type="supplementary-material"}) [@pcbi.1000021-Bartholin1],[@pcbi.1000021-Pessah1]. To validate the microarray-based expression measurement, *Tgif1* expression was measured in murine BMMs using quantitative PCR (qPCR; see [Materials and Methods](#s4){ref-type="sec"}). Consistent with the microarray-based results, *Tgif1* expression was found to be ∼3-fold upregulated after 1 hour of stimulation by LPS or Pam~3~CSK~4~ (data not shown). Furthermore, from microarray-based measurement (Affymetrix probeset 1422286_a\_at), *Tgif1* expression is ∼2.4-fold reduced in unstimulated *Ticam1* ^(Lps2/Lps2)^ BMMs relative to wild-type (with no apparent effect in *MyD88* ^(−/−)^BMMs relative to wild-type), suggesting that basal expression of *Tgif1* is TRIF-dependent. Targeted validation using ChIP-on-chip {#s2e} -------------------------------------- Genome location analysis based on chromatin immunoprecipitation-on-chip (ChIP-on-chip) hybridization was used to validate five high-confidence associations in the transcriptional network, between NFκB/p50 and clusters C13, C17 and C28; and between IRF1 and clusters C13 and C25. This validation consisted of demonstrating a statistical enrichment of ChIP-on-chip--identified binding for a given TF among genes within a cluster with which the TF was associated through our computational method, as compared to randomly selected TLR-responding genes. A custom-fabricated oligonucleotide microarray was used, with probes tiling up- and downstream of genes that were differentially expressed under TLR stimulation in a murine macrophage-like cell line (see [Materials and Methods](#s4){ref-type="sec"}). Macrophages were stimulated with LPS and then ChIP was carried out using TF-specific antibodies at 1 and 2 h, and (for IRF1 only) 4 h. Binding of p50 was highly enriched within the genes of clusters C13 and C28 represented on the tiling array (18 out of 23 and 20 out of 21 genes were bound, respectively) but not significantly enriched for C17 (11 out of 20). IRF1 binding was enriched within the genes of C13 and C25 (18 out of 23, and 18 out of 22, respectively). In four out of five cases, the enrichment relative to the overall rate of binding to differentially expressed genes represented on the tiling array satisfied *P*\<0.01 (Fisher\'s Exact Test; see [Table 2](#pcbi-1000021-t002){ref-type="table"}). ChIP-on-chip results for individual target genes within the aforementioned clusters are provided in [Table S13](#pcbi.1000021.s031){ref-type="supplementary-material"}, and results for all clusters that were represented on the array (see [Materials and Methods](#s4){ref-type="sec"}) are shown in [Table S14](#pcbi.1000021.s032){ref-type="supplementary-material"}. For each of the two TFs assayed with ChIP-on-chip, and for those clusters that were identified as targets of the TF through the network analysis, the fraction of clusters found to have significant TF binding to their genes was higher than for clusters selected randomly from among all clusters represented on the tiling array (1.7-fold overall). Additionally, the association between IRF1 and C30 was significant (*P*\<0.05) based on scanning, but not significant based on *P* ^comb^. Consistent with the integrated analysis, C30 was not significantly enriched for IRF1 binding, based on the ChIP-on-chip assay. 10.1371/journal.pcbi.1000021.t002 ###### Validation of transcription factor-to-cluster associations using ChIP-on-chip {#pcbi-1000021-t002-2} TF Matrix Stim. Clust Time Points In Clust On Chip Bound *P*-Value ---------- ------------- ------- ------- --------------- ---------- --------- ------- ------------ NFκB/p50 NFKB_Q6 LPS C13 1 h, 2 h 64 23 18 1.1×10^−3^ NFκB/p50 NFKB_Q6 LPS C17 1 h, 2 h 58 20 11 2.5×10^−1^ NFκB/p50 NFKAPPAB_01 LPS C28 1 h, 2 h 28 21 20 1.1×10^−6^ IRF1 IRF_Q6_01 LPS C13 1 h, 2 h, 4 h 64 23 18 2.3×10^−3^ IRF1 IRF_Q6_01 LPS C25 1 h, 2 h, 4 h 37 22 18 8.8×10^−4^ Shown are five (TF,cluster) associations for which at least 30% of the genes within the cluster are represented on the tiling array, along with the results of the ChIP-on-chip assay for binding of the indicated TF to the promoters of genes within the indicated cluster. Column 1 indicates the transcription factor antibody target. Column 2 indicates the position-weight matrix that was used for scanning the promoters of genes in the cluster. Column 3 indicates the stimulus used. Column 4 indicates the gene cluster whose promoters the indicated TF is predicted to bind. Column 5 indicates the time points at which ChIP-on-chip assays were performed. Column 6 indicates the number of genes in the cluster. Column 7 indicates how many of these genes have probes tiled on the chip, in the flanking 5′ intergenic region (due to the much smaller microarray expression dataset used to select genes for the tiling array, only about 22% of the 1,960 differentially expressed genes were represented on the tiling array, as described in [Materials and Methods](#s4){ref-type="sec"}). Column 8 indicates the number of these genes that were identified positively by ChIP-on-chip as having the indicated transcription factor bound to chromatin, in the 5′ flanking intergenic region. Column 9 indicates the *P* value for the enrichment of ChIP-on-chip hits among genes within the cluster identified by promoter scanning, as compared to the set of all genes on the array (Fisher\'s exact test). The ChIP-on-chip results for individual genes are provided in [Table S13](#pcbi.1000021.s031){ref-type="supplementary-material"}. Discussion {#s3} ========== In this study we inferred a transcriptional network underlying dynamic TLR-stimulated activation of the murine macrophage. This network consists of statistical associations between differentially expressed transcription factor (TF) genes and co-expressed clusters of genes, each indicating a possible role for the associated TF in regulating the cluster. Such associations have proved useful for generating and prioritizing testable hypotheses regarding transcriptional regulation [@pcbi.1000021-Gilchrist1],[@pcbi.1000021-Nilsson1]. A novel computational approach was used that combined sequence- and expression-based evidence. Using expression data acquired under a comprehensive set of TLR stimuli (and sampled densely in time), differentially expressed genes were partitioned into clusters of co-expressed genes that revealed a diversity of induction time scales, functional enrichments, and stimulus-dependent activation patterns. The clustering enabled sensitive identification of TFBS enrichments despite uncertainty (due to limited sampling) in the position-weight matrices and in the appropriate score threshold for motif scanning. In addition, using the SDR-normalized expression data for clustering ensured that genes were clustered based on their temporal (and stimulus-dependent) activation profiles, rather than by the magnitude of fold-change. Early-upregulated clusters were found to be enriched for TFs, consistent with the idea that many regulators of the transcriptional program are themselves produced on-demand in response to TLR stimulation [@pcbi.1000021-Gilchrist1]. The early induction of a large number of TFs was an important indicator of the potential utility of analyzing temporal expression as an evidence for transcriptional regulatory interactions (TRIs). The time-lagged correlation (TLC) was used to analyze temporal gene expression for TFs and gene clusters, and in addition to the correlation strength, the biological plausibility of the estimated optimal time lag was factored into the significance assessment for the TLC. This time lag is useful for distinguishing between genes that are linked by a regulatory interaction and genes that are merely co-expressed. The TLC is efficient to compute, and in general requires fewer measurements than methods that rely on estimating the joint probability distribution of the expression of two genes (e.g., pairwise mutual information [@pcbi.1000021-Basso1]). This observation is related to the most notable drawback of the TLC, namely, that it is sensitive only to the covariance of the joint probability distribution, and not higher order moments (with significantly more expression measurements, a possible extension of this method could be to use time-lagged mutual information [@pcbi.1000021-Schreiber1]). A second limitation of the TLC (and of any evidence based solely on lagged expression comparison) is that in practice it can be difficult to distinguish between indirect transcriptional regulation through a rapid-acting intermediary, and direct transcriptional regulation. Finally, while it is not a significant issue in the cluster analysis described in this work, we note that the proposed method for estimating the significance of the expression data for a single gene pair (Equation 6) is potentially not robust with respect to noise in the data. For the purpose of single-gene analysis, it could be improved by using a polynomial fit to the *τ* dependence of , or by defining the optimal time lag to be the time lag that minimizes *σ*. The specific implementation of the TLC approach used in this study has two key advantages. First, by selecting the optimal time lag for a TF--gene pair based on minimizing the lag-dependent *P* value rather than maximizing the squared correlation coefficient, the inherent bias of the TLC technique in selecting time lags was avoided. This made it possible to include the contributions of (i) the magnitude of the correlation, and (ii) the probability of observing the optimal time lag, to the significance of a pairwise association. Second, the probability distribution for time lags among true interactions was incorporated as a prior in the significance calculation. This enabled taking into account the biological plausibility of the time lag in computing the significance. This significance test for the TLC has not, to our knowledge, been previously reported. With any network inference method based on pair-wise comparison of the expression profiles of a regulator and a possible target (including the TLC method), it is difficult to accurately resolve the multi-factorial control of a target gene. This is particularly true when the effect of one TF is simply to modulate (amplify or dampen) the time-varying influence of another TF on a target gene. Several additional mechanisms can confound or eliminate the correlation between the expression level of a TF gene and the chromatin-bound activity of the corresponding TF, including multimeric TF assembly from protein products of several genes, post-translational activation of the TF, dynamically regulated nuclear translocation, and dynamically regulated TF protein turnover. For example, in the case of ATF3, there is little correlation between differential expression and nuclear localization [@pcbi.1000021-Gilchrist1], and as a result, this TF is not strongly implicated in the network via TLC. However, we note that the CREB/ATF binding motif was identified as enriched within the core early response cluster C27. Additionally, we note that given that the expression data set used in this work is densely sampled at early times (1--2 hours) and sparsely sampled at late times, our ability to leverage expression data as an evidence for TRI is reduced for very late-responding TF genes (e.g., *Lmo2*). In summary, with a limited expression dataset, a high-significance TLC by itself should not be regarded as sufficient evidence to infer a TF-to-target association, underscoring the importance of incorporating additional sources of evidence. In the present work, promoter sequence scanning was used to identify TFBS motifs enriched within co-expressed gene clusters. Due to the often one-to-many mapping between TFBS motifs and TFs, the scanning-based evidence often identifies multiple candidate TFs with a gene cluster, of which perhaps a single TF may be the relevant regulator in the given condition. The TLC approach described here provides an objective statistical framework for evaluating the suitability of a proposed TF-to-target association based on a large set of time-course expression measurements. In particular, the approach enabled the preferential identification of TF-to-target associations for which the optimal time lag is biologically plausible, and the rejection of associations with a biologically implausible (e.g., zero) time lag. Four (TF,cluster) associations were validated using ChIP-on-chip assays, in which enriched binding of the relevant TF was shown among genes within the relevant cluster. The ChIP-on-chip enrichment *P* values are conservative estimates of the genome-wide binding enrichment, due to the fact that genes were selected for inclusion in the tiling array based on differential expression under LPS stimulation in a macrophage-like murine cell line (RAW 264.7). We note that for each of the two TFs assayed, two (TF,cluster) pairs were found to be enriched for binding based on ChIP-on-chip, but not based on the network analysis. Such false-negative predictions may be the result of binding sites sometimes occurring upstream of the 2 kbp region selected for TFBS motif scanning, the target TF being cross-linked to a DNA-bound co-regulator recognizing a different motif than the TF, or due to the TF recognizing a TFBS motif variant not represented in the motif database. The inferred transcriptional network resulting from our analysis associates at least one TF with 27 of the 32 clusters. The 27 clusters comprise 86% of all differentially expressed genes, with an overall network coverage (including motif matches for individual targets) of 63%. An average of 3.3 TF genes were associated with each cluster, which may reflect the prevalence of combinatorial control in the transcriptional network. The TFs implicated in the network are also highly interconnected at the level of protein-protein interactions, and interacting TFs are found to co-associate with clusters in the network. Many TFs known to play a role in macrophage activation were strongly associated with clusters in the inferred network (e.g., NFκB, AP1, IRF family members, and STAT1). NFκB and AP1 appear to be the most prolific activators in the network. EGR family members appear to be associated with early-induced clusters, and IRF family members are associated with later-induced clusters. In particular, the network associated specific TFs with immunologically important gene clusters (e.g., EGR1/2 and AP1 regulating cluster C27; and NFκB and IRF1 regulating cluster C13). Finally, incorporating expression data enabled identifying a specific TF from among members of a large TF family recognizing a motif enriched within a target cluster; for example, the predicted interaction between IRF1 and C25 was validated by ChIP-on-chip. However, we note that more ChIP-on-chip data, with a variety of TF targets, would be required to quantitatively assess the performance of the combined network analysis compared to single-evidence analysis using sequence data or expression data alone. We note that by including in the analysis only TFBS motifs for which at least one associated TF gene was differentially expressed, the inferred network does not include TFs for which there is *no* transcript-level differential expression; this trade-off enabled network inference based on *dual* criteria of motif match enrichment and the estimated time lag prior probability. Work is in progress to extend the analysis to include all 208 TFBS motifs corresponding to TFs that are transcriptionally expressed in the TLR-stimulated macrophage. Another limitation related to sequence scanning is that the promoter sequence data set used is purely upstream of the annotated transcription start site (TSS); recent evidence suggests that some TFs may be equally likely localized downstream of the annotated TSS [@pcbi.1000021-Birney1]. In future work, it could be productive to scan for TFBS motifs both upstream and downstream of the annotated TSS. In addition to recapitulating known regulators, the analysis identified a potential transcriptional regulator not previously known to play a direct role in TLR-stimulated macrophage activation, TGIF1. TGIF1 is a three-amino acid loop extension homeobox protein that acts as an obligate repressor through either direct binding to the retinoic acid responsive element on DNA, or through its interaction with SMAD2 in the TGFβ pathway [@pcbi.1000021-Bartholin1]. Its associated TFBS motif is enriched within the promoters of genes within cluster C26 (*P*\<10^−2^) and cluster C4 (*P*\<10^−2^), and *Tgif1* is strongly (11-fold) upregulated in murine macrophages in response to *Streptococcus pyogenes* infection [@pcbi.1000021-Goldmann1]. Particularly intriguing is the possibility that, in light of motif scanning evidence, TGIF1 may act as a transcriptional repressor of the cytokines *Csf2* and *Gm1960.* The approach of combining promoter scanning-based evidence with expression dynamics-based evidence enabled more specific identification of the TF gene(s) regulating a cluster than would have been possible using promoter scanning alone. Time-course expression data allowed, in some cases, the disambiguation of which TF gene (out of a family of TF genes associated with a given TFBS motif) is the likely regulator of a cluster enriched for the corresponding TFBS motif. Inclusion of expression data provided a second source of evidence to indicate the relevance of a given TF gene for predicting the condition- and time-specific expression of a target gene cluster. In total, these results validate the strategy of computationally integrating two distinct large-scale data sources (expression and genomic sequence) to infer a murine macrophage transcriptional network. In a future study, additional sequence-based data sources, such as evolutionarily conserved elements in the *cis*-regulatory region, could be incorporated into the method. Materials and Methods {#s4} ===================== All data were analyzed in MatLab (MathWorks, Natick, MA) unless otherwise stated. In all cases where Fisher\'s exact test was performed, the test was one-tailed, using the cumulative distribution function (CDF) of the hypergeometric distribution. Microarray expression measurements {#s4a} ---------------------------------- Mutant strains (see [Table S1](#pcbi.1000021.s019){ref-type="supplementary-material"}) were generated in the 129SVJ background and backcrossed to C57BL/6 (Jackson Laboratories), ten times. Femurs from the C57BL/6 and mutant strains were flushed with complete RPMI (RPMI 1640 supplemented with 10% FBS, 2mM L-glutamine, 100 IU/mL penicillin and 100 µg/mL streptomycin, all from Cellgro, Mediatech, except the FBS which was from Hyclone). Bone marrow cells were plated on non-tissue culture treated plastic in complete RPMI supplemented with recombinant human M-CSF (rhM-CSF) at 50 ng/mL (gift from Chiron). On day 4 the cells were washed two times with RPMI with no additions and then grown 2 more days in complete RPMI supplemented with 50 ng/mL of rhM-CSF. On day 6 the cells were lifted from the non-tissue culture treated plastic, counted and plated at a density of 1.04×10^5^ cells/cm^2^ (1×10^6^ cells per well in a 6-well dish) on tissue culture-treated plastic. On day 7 cells were stimulated with TLR agonists at the concentrations indicated in [Table S2](#pcbi.1000021.s020){ref-type="supplementary-material"}, without changing the media. Stimulus reagent sources are shown in [Table S15](#pcbi.1000021.s033){ref-type="supplementary-material"}. Stimulation of the cells was verified by the presence of TNFα in the culture supernatants detected by ELISA (Duoset ELISA Assay Development System, R&D Systems). Total RNA was isolated using TRIzol (Invitrogen) and analyzed for overall quality using an Agilent 2100 Bioanalyzer. mRNA was labeled using the Affymetrix One-Cycle Target Labeling protocol and reagents for eukaryotic target preparation. The labeled cRNA was hybridized to an Affymetrix GeneChip Mouse Genome 430 2.0 array using standard protocols and reagents from Affymetrix. Probe intensities were measured using the Affymetrix GeneChip Scanner 3000 and processed into CEL files using Affymetrix GeneChip Operating Software. Microarray data processing {#s4b} -------------------------- Expression data were acquired from 216 microarray hybridization experiments comprising 95 combinations of strain, stimulus, and time point (hereafter, "experiments"; see [Table S3](#pcbi.1000021.s021){ref-type="supplementary-material"}), of which 41 were in mutant strains, and 54 in wild-type. Data in the form of CEL files were background-subtracted and normalized with the Robust Multi-chip Average (RMA) method [@pcbi.1000021-Irizarry1] using the software Bioconductor [@pcbi.1000021-Gentleman1], then exported to MatLab for further analysis. For each of the 95 experiments, normalized expression measurements for each probeset were averaged across biological replicates using the log~2~ intensities [@pcbi.1000021-Irizarry1] to obtain the replicate-combined probeset intensity. Differential expression testing {#s4c} ------------------------------- Significance testing was performed using mean log~2~ intensities from 7 wild-type TLR-stimulation time-course experiments comprising 54 assays (where "assays" refers to a specific combination of strain, stimulus, and elapsed time; see [Table S3](#pcbi.1000021.s021){ref-type="supplementary-material"}) for which at least two replicates were available, relative to the mean log~2~ intensities of unstimulated wild-type macrophages (hereafter, the "reference experiment"). For each probeset and for each of the wild-type TLR-stimulation time-course experiments, a differential expression test was performed using a spline-based multivariate regression method [@pcbi.1000021-Storey1] to obtain a *P* value for the difference in the sum-squared residuals under the alternative and the null hypotheses. A fourth-order polynomial basis was used, with 1,000 iterations for the bootstrap resampling. For each time-course experiment, a separate *P* value threshold was selected based on a maximum Benjamini-Hochberg false discovery rate (FDR) [@pcbi.1000021-Benjamini1] as described below. Probeset selection {#s4d} ------------------ A probeset selection algorithm was carried out to select a representative probeset for each gene, eliminating probesets that are annotated as cross-hybridizing to transcripts from different genes. Representative probesets from among the 45,037 probesets (excluding on-chip control probesets) on the Affymetrix Mouse GeneChip 430.2 were selected based on four criteria. A probeset was selected if and only if: (i) it possessed an Entrez GeneID in the Affymetrix probeset annotation database [@pcbi.1000021-Affymetrix1]; (ii) it had a log~2~ intensity exceeding a fixed cutoff, in at least one replicate-combined experiment; (iii) it had a *P* value less than a fixed cutoff, for at least one experiment; and (iv) its probeset name did not contain "\_x\_" or "\_s\_", and was not associated (by GeneID annotation) with transcripts of two distinct genes. Criterion (iv) was imposed in order to eliminate probesets containing probes that cross-hybridize to transcripts from different genes [@pcbi.1000021-Affymetrix2]. Whenever multiple probesets mapped to the same GeneID (or the same collection of GeneIDs), the probeset with the smallest minimum *P* value, across all non-reference experiments, was selected as the "representative probeset" for the GeneID (or GeneID list). This selection procedure was applied with four different cutoffs for log~2~ intensity and *P* value, as summarized in [Table S16](#pcbi.1000021.s034){ref-type="supplementary-material"}. (i) To generate a set of differentially expressed genes suitable for expression clustering (hereafter, the "target" genes), a log~2~ intensity cutoff of 6 was used, and a *P* value cutoff of 10^−4^ was used. The resulting number of representative probesets for target genes was 1,960. The complete list of the 1,960 target genes, and their expression measurements, are provided in [Table S4](#pcbi.1000021.s022){ref-type="supplementary-material"}. (ii) To generate a set of differentially expressed TF genes, the algorithm was re-run for probesets that are annotated as TFs, and for which a TRANSFAC matrix is available (see Selection of Transcription Factors), with a FDR cutoff of 0.05. A total of 80 differentially expressed TF genes were identified, as described in Selection of Transcription Factors below. (iii) To generate a set of all genes that were expressed in the macrophage, in at least one experiment, the probeset selection was run with a log~2~ intensity cutoff of 6 and no filtering for differential expression. The 8,788 resulting genes were used as the reference set for applying Fisher\'s exact test to the promoter scanning results (see Promoter Scanning below). (iv) To generate the set of all genes represented by "\_at" or "\_a_at" probesets on the GeneChip, the algorithm was run with no filtering for minimum intensity or differential expression. This generated a list of 20,905 genes that constituted the genome-wide set used in the gene ontology enrichment analysis (see Functional Enrichment Analysis below). Selection of transcription factors {#s4e} ---------------------------------- A set of 388 position-weight matrices (PWMs) corresponding to murine TFs was obtained from the TRANSFAC Professional database version 10.3 [@pcbi.1000021-Wingender1]. These PWMs were mapped using TRANSFAC as well as literature searching, to 273 mouse genes that code for corresponding TFs or TF components. Of these, 80 TF genes were identified as differentially expressed (FDR≤0.05) as described in Probeset Selection above (see [Table S5](#pcbi.1000021.s023){ref-type="supplementary-material"}). The 80 TF genes are represented by 150 TRANSFAC position-weight matrices. [Table S10](#pcbi.1000021.s028){ref-type="supplementary-material"} contains the microarray expression measurements for these TF genes. To estimate the fraction of genes in the mouse genome that are TFs, a genome-wide list of 1,245 murine TF genes (and probable TF genes) was assembled by mapping a list of 1,800 human TF genes from the literature [@pcbi.1000021-Roach1] to mouse orthologs present on the Mouse GeneChip and integrating the set of genes possessing GO annotations for transcription factor activity (GO:0003700). Expression clustering {#s4f} --------------------- The SDR values *x~pj~* for log~2~ intensity, where *p* indicates the probeset and *j* indicates the experiment, were clustered using a fast implementation of the *K*-means algorithm [@pcbi.1000021-Dollar1], with a minimum cluster size of 1. The number of clusters *K* was chosen to minimize the Bayesian Information Criterion (BIC) [@pcbi.1000021-Hastie1]. The BIC is a function of *K* represented as *BIC*(*K*),where *k~p~* is the cluster to which the *p* ^th^ probeset is assigned, is the *j* ^th^ coordinate of the centroid of the *k* ^th^ cluster in the SDR-transformed space of expression measurements, *N* = 1,960 (the number of target genes), *M* = 94 (the number of non-reference experiments), and *σ~ε~^2^* is the average intra-cluster variance evaluated at *K* = 3. The *K*-means clustering was carried out for integer values 18≤*K*≤50, for 1,000 iterations at each value of *K*; the optimal clustering (lowest average BIC over the 1,000 iterations) occurred at *K* = 32 (see [Figure S1](#pcbi.1000021.s002){ref-type="supplementary-material"}). The cluster expression profiles were characterized using the within-cluster median of the SDR; as a result, the *cluster* expression profile will not necessarily have a maximum value of 1 across all data points. This is because, in general, the genes within a cluster will not all reach a maximum value at the same time point. The induction time scale for the median SDR expression within each cluster was estimated using linear interpolation between the time points for the wild-type LPS time-course, and finding the time at which the absolute value of the SDR first exceeded 0.25. Clusters were displayed (in [Figure 1](#pcbi-1000021-g001){ref-type="fig"} and [Figure S2](#pcbi.1000021.s003){ref-type="supplementary-material"}) in the cluster order that minimized the sum of Euclidean distances between adjacent clusters, obtained using simulated annealing [@pcbi.1000021-Press1] with 5000 iterations and a cooling rate of 0.5. The cluster expression profiles in [Figure 2](#pcbi-1000021-g002){ref-type="fig"} were ordered for display using hierarchical agglomerative linkage using the Euclidean distance of extremal SDR expression level in time-course microarray experiments under the four indicated TLR agonists. Functional enrichment analysis {#s4g} ------------------------------ Jackson Laboratory Mouse Genome Informatics GO annotations [@pcbi.1000021-Jackson1] were added to the Affymetrix Mouse GeneChip GO annotations [@pcbi.1000021-Affymetrix1] by string matching on the gene symbol field for each annotated probeset. For each of the 20,945 GO term IDs [@pcbi.1000021-Gene1], the number of occurrences of the GO term ID (or a descendent of the GO term ID) in the GO hierarchy was computed for all 20,905 genes represented on the Affymetrix Mouse GeneChip (see Probeset Selection above) as well as for each co-expressed gene cluster. For each GO hierarchy (process, component, and function) the total number of genes possessing at least one GO annotation for the hierarchy was computed (see [Table S17](#pcbi.1000021.s035){ref-type="supplementary-material"}). The *P* value for GO enrichment was computed for each pair (*i*,*C*) of a GO term ID *i* and gene cluster *C*, using Fisher\'s exact test (under-occurrences of a GO term relative to the reference set were discarded). Any pairs (*i*,*C*) in which less than 5% of the genes within *C* possess GO term ID *i*, or with a term level in the GO hierarchy less than 3, were discarded. The resulting 629 (*i*,*C*) pairs were ordered by *P* value, and a *P* value cutoff was selected by demanding that the estimated false discovery rate be 0.02 (*P*≤0.0148, or −log~10~ *P*≥1.83). The resulting 460 GO term enrichments are shown in [Table S8](#pcbi.1000021.s026){ref-type="supplementary-material"}. The list of 32 TLR-regulated murine cytokines was obtained by screening for all differentially expressed genes possessing an annotation for cytokine or chemokine activity, and by refining the list by using NCBI PubMed searches to determine whether each gene is a cytokine. Selection of genes for null distribution {#s4h} ---------------------------------------- To form the null distribution of time-lagged correlation, a set of non-TF genes was generated. From the set of 1,960 differentially expressed genes, a set *Q* of 484 genes were selected such that each gene: (i) does not correspond to a TRANSFAC transcription factor as described above; (ii) has at least two GO process and two GO function annotations; (iii) is not annotated as "regulation of transcription, DNA-dependent" (GO:0008015); (iv) does not have a gene name with the prefix "Zfp" (zinc finger protein); and (v) is not listed among the 1800 TF genes (see Selection of Transcription Factors). The time-lagged correlations between genes within this group were taken as the null distributions of time-lagged correlations, for the purpose of computing the *P* value of a time-lagged correlation between a TF and a gene (see Time-lagged Correlation below). Constructing the prior distribution of time lags {#s4i} ------------------------------------------------ Given the time resolution of the expression data (for which the smallest Δ*t* is 20 min), the set *L* of time lags was chosen to be 0--80 min (inclusive), at 10 min intervals. The precision at which the optimal time lag can be estimated, at \|*ρ* ~τ~\|≥0.9, was determined to be ±5 min, based on simulated independent Gaussian noise added to the replicate-combined array data with standard deviation given by the measured replicate-standard deviation of the log~2~ intensity in each experiment. The upper limit of 80 min was selected to ensure that in each time-course with time points *T*, the target gene expression evaluated at time points {*t*+*τ*\|*t*∈*T* and *t*+*τ*≤max(*T*)} would always be based on measurements from at least three time points. The conditional probability density *P*(*τ* ~c~\|*H̅* ~0~) of the overall transcriptional time delay *τ* ~c~, for true interacting TF--target gene pairs, was defined using the gamma distribution (see [Text S1](#pcbi.1000021.s001){ref-type="supplementary-material"}, Section 3). This probability density was integrated for bins of *τ* ~c~ centered at the discrete time lags *τ*∈*L*, to obtain an estimate of the discrete probability for observing an optimal time lag, where Δ*τ* = 10 min. Using the distribution *P*(*τ* ~c~\|*H̅* ~0~), the upper limit of 80 min for *τ* included approximately 97% of transcriptional delays. Time-lagged correlation {#s4j} ----------------------- The time-lagged correlation (TLC) was computed for all possible triples (*f*,*g,τ*) of TF gene *f*, potential target gene *g*, and time lag *τ* ∈ *L*. There were 80 TFs and 1,960 target genes. The TLC was computed as follows, for a given (fixed) time lag *τ*. Let the vectors *X~T~*(*f*) and *X~T~*(*g*) represent the log~2~-transformed, SDR-normalized expression measurements for *f* and *g* in a time-course, where *T* is the set of time points, and let *t* ~max~≡max(*T*). Let *T~τ~*≡{*t*∈*T*\|*t*≤*t* ~max~−*τ*}. Let and represent the measurements of *f* and *g,* respectively, at the times *T~τ~*. We now define the set of shifted time points *T\'~τ~*≡*T~τ~*+*τ* = {*t*+*τ*\|*t*∈*T~τ~*}. The expression values were computed using linear interpolation between the adjacent time points. Expression values for each time course were concatenated together to obtain a combined multi-experiment vector of measurements for *f* and a combined vector of time-boosted measurements for *g*. The TLC *ρ~τ~*(*f*,*g*) was then computed using Equation 2 and using and . The criteria for inclusion of a time-course experiment in the TLC calculation were (i) a minimum of three points in the set *T~τ~*, and (ii) a minimum of three measurements contributing to the interpolated values . A total of eleven time-course experiments comprising 72 independent time points were included in the TLC analysis, as shown in [Table S9](#pcbi.1000021.s027){ref-type="supplementary-material"}. To build the background (null) TLC distribution (as defined in [Text S1](#pcbi.1000021.s001){ref-type="supplementary-material"}, Section 2) for each time lag *τ*, the TLC was computed for a set *H* consisting of all non-identical pairs of genes (*h* ~1~,*h* ~2~), where *h* ~1~ and *h* ~2~ are drawn from the set *Q* of non-TF genes (see Selection of Genes for Null Distribution above). The background distributions were constructed from the *ρ~τ~* ^2^(*h* ~1~,*h* ~2~) values, using Gaussian kernel density estimation [@pcbi.1000021-Hastie1] (see also [Text S1](#pcbi.1000021.s001){ref-type="supplementary-material"}, Section 4) with a smoothing length of 0.005 (chosen to maximize the number of pair-wise associations in the non-background set for which *P* ^tlc^≤10^−3^). For each *τ* and each *ρ~τ~*(*f*,*g*), the complementary CDF was computed by integration of using the extended Simpson\'s Rule (closed interval) [@pcbi.1000021-Press1] with 200 bins. The TLC was then analyzed for the set *G* of gene pairs (*g* ~1~,*g* ~2~), where *g* ~1~ was drawn from the set of 80 TFs (see Selection of Transcription Factors above), *g* ~2~ was drawn from the set of 1,960 differentially expressed ("target") genes (see Probeset Selection above), and *g* ~1~≠*g* ~2~ (the inequality avoids perfect zero-time-lagged correlations that would bias the significance test). For each pair (*g* ~1~,*g* ~2~), the time lag that maximized was selected as the optimal time lag for the pair, and denoted by *θ*(*g* ~1~,*g* ~2~). The probability ratio *R*(*τ*) was computed using Equation 5. The marginal probability *P*(*H* ~0~) was estimated to be ∼0.94 based on an analysis of the transcriptional network of [@pcbi.1000021-Nilsson1], taking the average out-degree of the TFs in [Fig. 4B](#pcbi-1000021-g004){ref-type="fig"} and dividing by the number of differentially expressed genes in that study (1,784 genes). The marginal probability *P*(*τ*)was obtained from *θ*(*H*). The combined, cumulative, TLC-based *P* value for (*f*,*g*), denoted by *P* ^tlc^(*f*,*g*), was computed according to Equation 7 (for which a detailed mathematical derivation is given in [Text S1](#pcbi.1000021.s001){ref-type="supplementary-material"}). Empirical evidence showing the approximate independence of *ξ* and *R* under the null hypothesis is shown in [Figure S17](#pcbi.1000021.s018){ref-type="supplementary-material"}. For each pair (*f*,*C*) of TF gene *f* and gene cluster *C* (see Expression Clustering above), an overall *F* score, *F* ^exp^ (*f*,*C*) was computed using Equation 8, combining the \|*C*\\{*f*}\|*P* values. Because the genes within a cluster are grouped by expression similarity, their TLCs with respect to *f* are not independent, even under the null hypothesis that *f* does not regulate any of the genes within the cluster. Thus, among a large collection of pairs (*f*,*C*) satisfying the null hypothesis, the *F* scores *F* ^exp^ (*f*,*C*) will not be distributed according to the χ^2^ distribution with 2\|*C*\\{*f*}\| degrees of freedom. Instead, the number of intra-cluster degrees of freedom was computed for each cluster by clustering the SDR expression profiles of the genes within a cluster (across all 94 non-reference experiments) using the *K*-means algorithm. For a range of numbers *k* of sub-clusters, the BIC was computed using the variance at *k* = 3 for normalizing the bias term [@pcbi.1000021-Hastie1]. The number of sub-clusters *k* at which the BIC was minimized was doubled to obtain the effective number of degrees of freedom, *d*(*C*), within each cluster. The average over all clusters was 〈*d*(*C~k~*)〉~k~ = 11.03, where *C~k~* denotes the *k* ^th^ cluster. The χ^2^ test was applied with *d*(*C*) degrees of freedom, to obtain an overall *P* value for the association between *f* and *C*:where *F* ^exp^(*f*,*C*) is defined in Equation 8, and *γ* is the incomplete gamma function [@pcbi.1000021-Press1]. A second statistic, the average time lag, was computed for each pair (*f*,*C*),and used as an additional criterion in the network inference (see Network Inference below). Promoter scanning {#s4k} ----------------- Mouse position-weight matrices (150 in total) corresponding to the 80 differentially expressed TF genes, were obtained from TRANSFAC Professional (see Selection of Transcription Factors above, and [Table S5](#pcbi.1000021.s023){ref-type="supplementary-material"}) [@pcbi.1000021-Wingender1]. Promoter sequences of 2 kbp upstream of 17,254 mouse genes were obtained from the UCSC genome database [@pcbi.1000021-Kuhn1] (UCSC annotation build mm8, based on the NCBI mouse genome assembly m36), each identified by NCBI RefSeq ID. The \[−2 kb, 0\] coordinate range relative to the transcription start site was selected based on Figure 2c from [@pcbi.1000021-Harbison1]. For each representative probeset (see Probeset Selection above), the corresponding RefSeq ID (if available) was obtained from the Affymetrix GeneChip annotation file [@pcbi.1000021-Affymetrix1]. The 8,788 expressed genes mapped to 7,492 unique promoter sequences (hereafter, the "reference" set, denoted by *μ* ~exp~ = 7,492). Of the 12,117 genes that were not expressed in any of the microarray experiments, 7,503 were mapped to UCSC promoter sequences (hereafter, the "background" set). The 1,960 differentially expressed genes were mapped to 1,713 unique promoter sequences. Low-complexity repeats were masked from all promoter sequences prior to motif scanning, using RepeatMasker [@pcbi.1000021-Smit1]. Scanning was performed using MotifLocator version 3.2 [@pcbi.1000021-Thijs1], using a first-order background model with frequencies computed from the first 496 genes on chromosome 17, obtained from the 5 kbp upstream promoter sequence file from NCBI mouse genome assembly 32 (UCSC build mm4), and using motif matrix score thresholds selected as described below. The background sequences were scanned with all matrices with no cutoff. For each matrix, the score threshold was computed at which an above-threshold match would occur on average in one out of every 5 promoter sequences (i.e., once per 10 kb). The motif match score thresholds are given in [Table S11](#pcbi.1000021.s029){ref-type="supplementary-material"}. The reference promoter set was scanned using these score thresholds, and for each matrix *m*, the number of promoter sequences in the reference set that had at least one above-threshold match was denoted by *ν* ~exp~ (*m*). For each cluster *C*, the mapped promoter sequences for the genes within the cluster (the number of which was denoted by *μ*(*C*)) were scanned, and the number of sequences with at least one above-threshold match was denoted by *ν*(*m,C*). For each matrix *m* and cluster *C*, a *P* value *P* ^scan^ (*m*,*C*) was computed from the values *μ* ~exp~, *ν* ~exp~ (*m*), *μ*(*C*), *ν*(*m,C*), using Fisher\'s exact test. Let Φ denote the mapping between the 80 TF genes and subsets of the 150 TRANSFAC matrices (see [Table S5](#pcbi.1000021.s023){ref-type="supplementary-material"}), so that Φ(*f*) is the set of TRANSFAC matrices associated with the TF gene *f*. For each TF gene *f* and cluster *C*, a *P* value representing the association between *f* and *C* was computed as follows,The values of *μ* ~exp~, *ν* ~exp~ (*m*), *μ*(*C*), and *ν*(*m,C*) for all clusters, are provided in [Table S11](#pcbi.1000021.s029){ref-type="supplementary-material"}. Network inference {#s4l} ----------------- For each pair (*f*,*C*) of TF gene *f* and co-expressed gene cluster *C*, an overall combined *P* value, *P* ^comb^ (*f*, *C*) for the significance of the association between *f* and *C* based on both promoter scanning and expression time-course data, was computed using Fisher\'s method,The set of all pairs (*f*,*C*) were selected, satisfying the following criteria: (i) *P* ^comb^ (*f*, *C*)≤0.0248 (or −log~10~ *P* ^comb^ (*f*,*C*)≥1.61, where the *P* value cutoff was obtained using an FDR of 0.025); (ii) *P* ^scan^ (*f*,*C*)≤0.05 (or −log~10~ *P* ^scan^ (*f*,*C*)≥1.3); and (iii) 〈*θ*〉*~f,C~*≥10 min. Criterion (iii) was used to ensure that a pair (*f*,*C*) would not be accepted based solely on a very low *P* ^scan^ (*f*,*C*) value; the average optimal time lag must be biologically plausible. A total of three TF-cluster associations were rejected, that passed criteria (i) and (ii), but not criterion (iii). A total of 90 TF-cluster associations were identified based on these criteria, involving 36 TF genes. The out-degree of a TF gene *f* within the network was estimated by summing (over all clusters for which (*f*,*C*) was accepted) the product *z*(*f*,*C*) \|*C*\\{*f*}\|, where *z*(*f*,*C*) is the fraction of genes within *C* that have at least one binding site for any matrix *m* ∈ Φ(*f*). The diagrams shown in [Figures S15](#pcbi.1000021.s016){ref-type="supplementary-material"} and [S16](#pcbi.1000021.s017){ref-type="supplementary-material"} were generated using Cytoscape [@pcbi.1000021-Shannon1] version 2.5.0. Protein interactions were obtained from the Human Protein Reference Database [@pcbi.1000021-Peri1], Release 6 (2007/01/01) and the Biomolecular Interaction Network Database [@pcbi.1000021-Alfarano1] (2007/10/14). The 36 differentially expressed TF genes were mapped to human orthologs using NCBI Entrez Gene. For the protein network diagram shown in [Figure S16](#pcbi.1000021.s017){ref-type="supplementary-material"}, a minimum log~2~ microarray probeset intensity cutoff of 6.5 was required in at least one array experiment (with the exception of *Smad6,* whose human ortholog protein is expressed in HL60 macrophage differentiation [@pcbi.1000021-Glesne1]). Quantitative PCR {#s4m} ---------------- Total RNA was isolated from bone marrow-derived macrophages using TRIzol (Invitrogen), treated with DNAase (Ambion), and used as template for reverse transcription (Superscript II, Invitrogen) according to the manufacturers\' instructions. qPCR was performed using Applied Biosystems ABI 7900 HT. Expression units were computed relative to the housekeeping gene *Eef1a1* [@pcbi.1000021-Gilchrist1],[@pcbi.1000021-Flo1]. Primer reagents for *Tgif1* and *Eef1a1* were obtained as described in [Table S15](#pcbi.1000021.s033){ref-type="supplementary-material"}. ChIP-on-chip validation {#s4n} ----------------------- Five (TF,cluster) pairs were selected for ChIP-on-chip validation based on several criteria: (1) the gene members of the cluster needed to be well-represented on the tiling array (at least 30% of the genes in the cluster must be represented on the ChIP-on-chip array); (2) a correlation between TF gene and cluster expression consistent with known function (activator or repressor) for the TF; (3) the availability of a high-quality polyclonal murine antibody for a relevant TF protein; (4) demonstrated specificity of the antibody based on Western blot analysis; (5) a successful ChIP assay for several known targets of the TF. Genome location was assayed using ChIP-on-chip hybridization as described in [@pcbi.1000021-Gilchrist1], with polyclonal antibodies for murine IRF1 and p50 (*Nfkb1*) ([Table S15](#pcbi.1000021.s033){ref-type="supplementary-material"}). A custom Affymetrix GeneChip microarray was used, consisting of 25-mer oligonucleotides selected to densely tile 20 kbp upstream and 20 kbp downstream (and selectively, the coding regions) of genes selected based on differential expression in preliminary microarray expression studies involving murine RAW 264.7 cells stimulated for 60 minutes by LPS, Pam~3~CSK~4~, or Pam~2~CSK~4~ [@pcbi.1000021-Innate1]. Of the 1,960 differentially expressed genes identified in Probeset Selection, 517 are represented on the tiling array. Hybridization to the custom tiling array was carried out using standard protocols and reagents from Affymetrix. ChIP-on-chip microarray scans were background-adjusted and quantile normalized as described in [@pcbi.1000021-Gilchrist1]. ChIP-on-chip data were processed as follows. First, probes were sorted based on chromosomal location. The sample/control absolute intensity ratio was computed for each probe, where the control intensity was taken from an experiment with antibody, but without LPS stimulation. A smoothed intensity profile was then generated using a sliding window algorithm based on Tukey\'s biweight kernel [@pcbi.1000021-Affymetrix3] with a 100 bp window size (as was used in [@pcbi.1000021-Gilchrist1]). Probes were then selected for which the intensity ratio was higher than a statistical cutoff (*P*≤0.01). If there were multiple significant probes within a 200 bp region, the combined statistical significance of region was computed by performing a *t*-test in which the distribution of probe intensities within the 200 bp region is compared to a background region of probe intensities. For each identified chromosomal region, the annotated gene nearest to the region in the 5′ direction was recorded, along with the distance to the nearest flanking gene. Significance testing of the enrichment of ChIP-on-chip binding among genes within a specific cluster was carried out using Fisher\'s exact test with a background set consisting of all 520 differentially expressed mouse genes (see Differential Expression Testing above) for which at least one probe on the array is located within 20 kbp upstream of the TSS. Accession numbers {#s4o} ----------------- All microarray expression data from this study have been deposited into the ArrayExpress [@pcbi.1000021-Parkinson1] public database under accession number E-TABM-310. NCBI Entrez Gene identifiers can be found for all differentially expressed genes considered in this study, in [Tables S4](#pcbi.1000021.s022){ref-type="supplementary-material"} and [S5](#pcbi.1000021.s023){ref-type="supplementary-material"}. Mouse Genome Informatics Allele accession numbers are provided for each mutant strain, in [Table S15](#pcbi.1000021.s033){ref-type="supplementary-material"}. Supporting Information {#s5} ====================== ###### Mathematical Derivations. This document provides a complete mathematical description of the significance test used for the time-lagged correlation. In addition, it provides background information on the Gaussian kernel density estimation method and some key theorems supporting the derivation of the method. (0.23 MB PDF) ###### Click here for additional data file. ###### The optimal number of clusters was determined using the Bayesian Information Criterion (BIC). The horizontal axis indicates the number of clusters *K* used for *K*-means clustering. The cluster analysis was repeated for *K* varying between 18 and 50, with the BIC computed for each number of clusters. The optimal number of clusters, for which the BIC is minimized, was found to be *K* = 32 (see [Materials and Methods](#s4){ref-type="sec"}, Expression Clustering). (0.15 MB TIF) ###### Click here for additional data file. ###### Differential expression profiles of gene clusters, in TLR-stimulated macrophages, across all microarray expression experiments. Each row represents an experiment (a specific combination of strain, stimulus, and time point), and each column represents a cluster. Clusters are displayed in the order that minimizes the sum of pairwise distances between adjacent clusters (see [Materials and Methods](#s4){ref-type="sec"}, Expression Clustering). Each colored rectangle within the heat-map indicates the centroid of the expression levels for genes within the indicated cluster, for the indicated experiment. The differential expression level (SDR, see Equation 1) is indicated in red/green color, and varies between -1 (bright green) and 1 (bright red), with 0 (black) indicating no change from the expression level in the unstimulated wild-type macrophage. The shaded light gray/charcoal regions in the far left column indicate the genotype. The color-coding in the second-to-left column indicates the stimulus (color code legend in lower right; and see [Table S2](#pcbi.1000021.s020){ref-type="supplementary-material"} for the concentrations). The four-digit numbers to the right of the color-code column, indicate the elapsed time (min) post-stimulation, for each experiment. (1.38 MB TIF) ###### Click here for additional data file. ###### Cluster-median differential expression profiles in wild-type macrophages stimulated with LPS show a diversity of time scales. Each data point shown is the median of the SDR-transformed (see Equation 1) differential expression levels of the genes within the indicated cluster, at the indicated time after stimulation. (0.34 MB TIF) ###### Click here for additional data file. ###### Cluster-median differential expression profiles in wild-type macrophages stimulated with Pam~3~CSK~4~ show a diversity of time scales. Each data point shown is the median of the SDR-transformed (see Equation 1) differential expression levels of the genes within the indicated cluster, at the indicated time after stimulation. Cluster C26 shows sustained activation under this stimulus, as opposed to the case of stimulation with LPS (see [Figure S3](#pcbi.1000021.s004){ref-type="supplementary-material"}). (0.32 MB TIF) ###### Click here for additional data file. ###### Cluster-median differential expression profiles in wild-type macrophages stimulated with poly I:C show a diversity of time scales. Each data point shown is the median of the SDR-transformed (see Equation 1) differential expression levels of the genes within the indicated cluster, at the indicated time after stimulation. The core response Clusters C27 and C28 induce later in this time-course experiment than in the case of stimulation with LPS ([Figure S3](#pcbi.1000021.s004){ref-type="supplementary-material"}). (0.32 MB TIF) ###### Click here for additional data file. ###### Cluster-median differential expression profiles of wild-type macrophages stimulated with R848 show a diversity of time scales. Each data point shown is the median of the SDR-transformed (see Equation 1) differential expression levels of the genes within the indicated cluster, at the indicated time after stimulation. Cluster C26 shows sustained activation under this stimulus, as opposed to the case of stimulation with LPS (see [Figure S3](#pcbi.1000021.s004){ref-type="supplementary-material"}). (0.34 MB TIF) ###### Click here for additional data file. ###### Discretized prior probability distribution *P*(τ\|*H0*) of observing an optimal time-lag τ, for a gene pair that have a transcriptional regulatory interaction. Here, the symbol ∼*H0* denotes the complement of the null hypothesis, i.e., that there is a transcriptional regulatory interaction (this is denoted by an overbar in the main text and in the [supporting text](#pcbi.1000021.s001){ref-type="supplementary-material"}). The symbol τ denotes the optimal time lag. For a discussion and derivation of the prior probability distribution of transcriptional time lags, see [Materials and Methods](#s4){ref-type="sec"} (Constructing the Prior Distribution of Time Lags) and [Text S1](#pcbi.1000021.s001){ref-type="supplementary-material"} (Section 3). (0.26 MB TIF) ###### Click here for additional data file. ###### Histogram of time lag values that maximize the absolute time-lagged correlation coefficient, for randomly drawn pairs of non-transcription factor genes. The non-uniformity of the histogram (the highest counts appear at high and low values of the time lag) shows the inherent bias in the standard method of selecting the optimal time lag, i.e., maximizing the absolute lagged correlation coefficient. Time-lagged correlations could not be reliably estimated for time lags greater than 80 min, due to limited effective sample size for higher time lags (see [Materials and Methods](#s4){ref-type="sec"}, Constructing the Prior Distribution of Time Lags). (0.22 MB TIF) ###### Click here for additional data file. ###### Differential expression levels (SDR, see Equation 1) in wild-type macrophages stimulated with LPS, for 38 pairs of transcription factor genes and gene clusters. The pairs all show high-significance time-lagged correlation based on the significance criterion *P* ^exp^ ≤ 5×10^-3^, and all satisfy the minimum average time lag criterion \<*θ*\> ≥ 10 min. Differential expression levels are relative to wild-type unstimulated macrophages, with positive/negative values indicating upregulation/downregulation. The names of the TF gene and the correlated cluster are shown above each plot. The cluster expression level, shown in green, is the centroid from the *K* -means clustering algorithm (see [Materials and Methods](#s4){ref-type="sec"}, Expression Clustering). Of the pairs, 23 have a positive time-lagged correlation coefficient, and 15 have a negative time-lagged correlation coefficient. (0.52 MB TIF) ###### Click here for additional data file. ###### Combined plot showing (i) the histogram of -log~10~ *P* ^exp^ values for the significance of the time-lagged correlation; and (ii) the estimated false discovery rate, as a function of the -log~10~ *P* ^exp^ value. The *P* ^exp^ values were computed for all possible pairs of (*f*,*C*) of transcription factor gene *f* and coexpressed gene cluster *C*. The histogram was generated using 40 bins. (0.27 MB TIF) ###### Click here for additional data file. ###### Histogram of positions of transcription factor binding site motif matches relative to transcription start site. The median distance from the transcription start site is 537 bp. The density of motif matches can be seen to peak at −20 bp relative to the start site. (0.25 MB TIF) ###### Click here for additional data file. ###### Combined plot showing (i) the histogram of -log~10~ *P* ^scan^ values for enrichment of TFBS motifs within co-expressed gene clusters; and (ii) the estimated false discovery rate as a function of the -log~10~ *P* ^scan^ value. The *P* ^scan^ values were computed for all possible pairs pairs (*f*,*C*) of transcription factor gene *f* and cluster *C*, using the position-weight matrix associated with *f* that had the smallest enrichment *P* value for the promoters of the genes in cluster *C*. The histogram was generated using 40 bins. (0.28 MB TIF) ###### Click here for additional data file. ###### Integrating the two sources of evidence using Fisher\'s method. Each blue circle represents a unique (TF,cluster) pair. The solid line indicates the cutoff for the combined *P* value, at FDR = 0.1. Data points to the lower left of the line have a *P* ^comb^ value smaller than the cutoff (see [Materials and Methods](#s4){ref-type="sec"}, Network Inference). The dotted green line indicates the cutoff for the promoter scanning-based *P* value, *P* ^scan^ = 0.05. Pairs that fall below the green dotted line and to the lower-left of the solid magenta line and for which the average time lag \<*θ*\> ≥ 10 min, were included in the final network. (0.59 MB TIF) ###### Click here for additional data file. ###### The set of transcription factor genes has a 20-fold variation in out-degree (number of target genes), within the transcriptional network. (a) Estimated out degree of transcription factor genes. The out degree of a transcription factor gene is the number of genes estimated to be regulated by the transcription factor(s) associated with that TF gene (i.e., of which that TF gene is a component). For each gene cluster with which a TF gene was associated, the number of genes within the cluster for which a motif match was found (corresponding to the TF gene), was tabulated. The number of target genes was summed over all clusters with which the TF was associated, based on the combined expression and promoter scanning data (see [Materials and Methods](#s4){ref-type="sec"}, Network Inference). Among the 36 TF genes in the network, the estimated out degree had a median of 49, and a maximum value of 285. (b) Estimated significance of the association of the TF gene in the network. For each TF gene *f* implicated in the network, the minimum *P* value *P* ^comb^(*f*,*C*) of association with any cluster *C*, was used as a measure of the overall significance of the association of TF gene in the transcriptional network. Transcription factor genes are displayed in decreasing order of estimated out degree (number of target genes). Transcription factors associated with larger clusters are seen to correlate with higher significances in the network, as a consequence of the sample size-dependence of the statistical tests used for the motif scanning and expression dynamics evidences. (0.51 MB TIF) ###### Click here for additional data file. ###### Transcription factors involved in macrophage activation are highly interconnected in the protein interaction network, and the interacting TFs co-associate with clusters. Nodes indicate TF genes whose transcript levels are differentially expressed in LPS-stimulated macrophages, and that are associated with the transcriptional network through the combination of scanning- and expression-based evidences. Node labels are gene names. A red node indicates upregulated gene expression under LPS, and green indicates downregulation, and a purple node indicates transient up- and downregulation. A blue arc indicates that the human orthologs of the murine proteins associated with the murine TF genes connected by the arc, have an interaction in the Human Protein Reference Database [@pcbi.1000021-Alfarano1] or in the Biomolecular Interaction Network Database [@pcbi.1000021-Barthel1]. A thick black arc indicates that the two connected TF genes co-associate with one or more clusters within the network, *and* share a protein interaction (suggesting a possible transcriptional complex). A purple arrow indicates a known protein-DNA interaction between the source node\'s human ortholog protein and the promoter of the human ortholog of the gene indicated by the target node. Brown ellipses denote the core transcription factor complexes NFκB and AP1. (0.64 MB TIF) ###### Click here for additional data file. ###### TGIF1 interacts with many members of the SMAD/AP-1 transcription complex. Shown here is a network diagram of 16 proteins that interact with the SMAD family of transcription factors SMAD1/2/3/6, the histone deacetylaces HDAC1/2, and the TG-interacting factors TGIF1/2. Nodes indicate proteins, and a blue line between two nodes indicates that the human orthologs of the two proteins have an interaction, in either the Human Protein Reference Database (HPRD) [@pcbi.1000021-Alfarano1] or in the literature [@pcbi.1000021-Hamilton1],[@pcbi.1000021-Schreiber1]. Red arrows indicate human protein-DNA interactions annotated in the TRANSFAC database [@pcbi.1000021-Hayden1]. The diagram includes nearest-neighbors of the SMAD, HDAC, and TGIF families in the protein interaction network. Each node shown in the diagram corresponds to a transcript that is likely expressed in murine bone marrow-derived macrophages, based on having an above-threshold microarray intensity within at least one experiment (see [Materials and Methods](#s4){ref-type="sec"}, Probeset Selection). (2.02 MB TIF) ###### Click here for additional data file. ###### Histogram of the cumulative density function of *ω*, for the *ω* values for all sample points with *ψ* = 80 min. Strict uniformity of this distribution (for each and every outcome *ψ* = *τεL*) would imply that ω is totally independent of *ω*\|*ψ*. Here, conditioning on *ψ* is seen to not introduce a significant bias in the distribution of *ω* values (see [Supporting Text](#pcbi.1000021.s001){ref-type="supplementary-material"}, Section 2). (0.30 MB TIF) ###### Click here for additional data file. ###### Summary of mutant mouse strains used in this study. Expression data from available mouse strains with mutations of known TLR signaling adapter molecules or known transcriptional regulators were included in the cluster analysis, in order to maximize the diversity of expression patterns in the data set used for clustering. Column 1 is the mutant strain name. Column 2 is the name of the molecule affected by the mutation. Column 3 gives the gene title. Column 4 briefly summarizes the relevance of the molecule in TLR-stimulated macrophages. (0.03 MB DOC) ###### Click here for additional data file. ###### Stimuli used for macrophage gene expression experiments. Column 1 indicates the purified TLR agonist. Column 2 gives the description of the agonist. Column 3 indicates the receptor(s) that are stimulated by the agonist. Column 4 indicates the adapter molecule(s) associated with the receptor. Column 5 indicates the concentration used for *in vitro* stimulation of macrophages. (0.04 MB DOC) ###### Click here for additional data file. ###### List of microarray experiments included in this study. Each row indicates a microarray experiment. Column 1 indicates the mouse strain, with "Wild-type" indicating C57BL/6. Column 2 indicates the stimulus (or combination of stimuli, separated by a slash "/"). Column 3 indicates the elapsed time post stimulation. Column 4 indicates the number of biological replicates combined in the experiment. Column 5 indicates whether the expression measurements for the experiment were used in identifying differentially expressed genes. Column 6 indicates if the experiment was used for the clustering analysis. Column 7 indicates if the experiment was used for time-lagged correlation (TLC) analysis. The alternating shaded pattern for rows is used to visually distinguish between experiments from different genotypes. (0.21 MB DOC) ###### Click here for additional data file. ###### Target genes with microarray expression data. This spreadsheet contains the replicate-combined probeset intensities for all 1,960 differentially expressed genes (see [Materials and Methods](#s4){ref-type="sec"}, Probeset Selection) across all 95 microarray experiments (see [Table S3](#pcbi.1000021.s021){ref-type="supplementary-material"}). Column 1 indicates the NCBI gene symbol of the gene. Column 2 indicates the NCBI Entrez Gene ID. Column 3 indicates the probeset selected as representative for the gene. Column 4 provides a brief gene description, obtained from the Affymetrix Mouse GeneChip annotations file. Column 5 indicates the co-expressed gene cluster to which the gene was assigned (see [Materials and Methods](#s4){ref-type="sec"}, Expression Clustering). Columns 6--8 provide listings of the gene\'s Gene Ontology annotations in the process, component, and function GO hierarchies, respectively (see [Materials and Methods](#s4){ref-type="sec"}, Functional Enrichment Analysis). Column 9 indicates the maximum log~2~ intensity observed, across all experiments. Columns 10-104 provide the log~2~ intensity measurements of the probesets across all 95 microarray experiments. (4.30 MB XLS) ###### Click here for additional data file. ###### Differentially expressed transcription factor genes considered as possible regulators of co-expressed gene clusters in this study. Column 1 contains gene symbol. Column contains the NCBI Entrez GeneID for the gene. Column 3 contains the representative Affymetrix probeset selected for the gene. Column 4 contains the co-expressed gene cluster of which the transcription factor is a member. Column 5 contains the TRANSFAC position-weight matrices that are associated with the transcription factor (or TF component) coded for by this gene (see [Materials and Methods](#s4){ref-type="sec"}, Selection of Transcription Factors). The "V\$" prefixes on TRANSFAC matrices are not shown. (0.13 MB DOC) ###### Click here for additional data file. ###### Summary of co-expressed gene clusters. Column 1 indicates the cluster name. Clusters were numbered in order of decreasing size. Column 2 indicates the number of genes in the cluster. Column 3 is a heat-map representation of the within-cluster median of the normalized differential expression intensity (SDR, see Equation 1), over time, in wild-type macrophages stimulated with LPS. The color red indicates upregulation relative to wild-type unstimulated macrophages, and green indicates downregulation (see color bar in [Figure S2](#pcbi.1000021.s003){ref-type="supplementary-material"}). Column 4 indicates the cluster response time under LPS stimulation, defined as the time scale (in minutes) for the log~2~ fold change to reach 25% of its extremal value (see [Materials and Methods](#s4){ref-type="sec"}, Expression Clustering); the time scale uncertainty is ± 5 min. Column 5 lists the known (excluding those solely inferred from electronic annotation, i.e., "IEA" evidence code) transcription factor genes that are *members* of the cluster (these are *not* the inferred transcriptional regulators of the cluster). Column 6 lists the known cytokines and chemokines that are members of the indicated cluster. (0.13 MB DOC) ###### Click here for additional data file. ###### The timing of induction of core response clusters C27 and C28 is adapter molecule-dependent. Column 1 indicates the stimulus. Column 2 indicates the microarray conditions compared, for example, fold-change (stimulated relative to unstimulated) in *Myd88* ^(−/−)^ macrophages vs. the fold-change in wild-type. Column 3 indicates the time post-stimulation. Columns 4 and 5 are the within-cluster medians of the log~2~ of the ratios for the condition comparison indicated in column 2, for the clusters C27 and C28, respectively. The data indicate that the early response of these clusters is largely dependent on the MyD88 signaling pathway, and that the later response (2 hours) is more strongly dependent on the TRIF signaling pathway. (0.03 MB DOC) ###### Click here for additional data file. ###### Gene Ontology enrichments in co-expressed gene clusters. Column 1 indicates the cluster. Column 2 contains the Gene Ontology ID (GOID) for the GO term. Column 3 contains the GO term. Column 4 indicates the GO hierarchy (process, component, or function) to which the GO term belongs. Column 5 contains the -log~10~ *P* value (significance) for the enrichment of the GO term in the indicated cluster. Column 6 contains the level of the GO term in the gene ontology hierarchy. Column 7 indicates the number of genes within the cluster that possess this GO term. Column 8 indicates the frequency at which this GO term appears in the set of all annotated genes in the genome (see [Materials and Methods](#s4){ref-type="sec"}, Functional Enrichment Analysis). Column 8 indicates the frequency at which the GO term appears among genes in the indicated cluster. (0.54 MB XLS) ###### Click here for additional data file. ###### Time-course macrophage stimulation microarray experiments used for time-lagged correlation analysis. Only time-course expression studies with a sufficient number of time points to admit time-lagged correlation analysis are shown (see [Materials and Methods](#s4){ref-type="sec"}, Time-lagged Correlation). Column 1 indicates the genotype from which macrophages were derived. Column 2 indicates the stimulus used. Column 3 indicates the times post-stimulation, at which gene expression was measured. (0.04 MB DOC) ###### Click here for additional data file. ###### Transcription factor genes with microarray expression data. This spreadsheet contains microarray probeset intensities for all 80 differentially expressed transcription factor genes (see [Materials and Methods](#s4){ref-type="sec"}, Selection of Transcription Factors) across all 95 microarray experiments (see [Table S3](#pcbi.1000021.s021){ref-type="supplementary-material"}). Column 1 indicates the NCBI gene symbol of the gene. Column 2 indicates the NCBI Entrez Gene ID. Column 3 indicates the probeset selected as representative for the gene. Column 4 provides a brief gene description, obtained from the Affymetrix Mouse GeneChip annotations file. Column 5 indicates the co-expressed gene cluster to which the gene was assigned (see [Materials and Methods](#s4){ref-type="sec"}, Expression Clustering). Columns 6--8 provide listings of the gene\'s Gene Ontology annotations in the process, component, and function GO hierarchies, respectively (see [Materials and Methods](#s4){ref-type="sec"}, Functional Enrichment Analysis). Column 9 indicates the set of TRANSFAC matrices associated with this transcription factor gene (see [Materials and Methods](#s4){ref-type="sec"}, Selection of Transcription Factors). Column 10 indicates the maximum log~2~ intensity observed, across all experiments. Columns 11--105 provide the log~2~ intensity measurements of the probesets, across all 95 microarray experiments. (0.21 MB XLS) ###### Click here for additional data file. ###### Transcription factor binding site (TFBS) motif position-weight matrices, threshold scores, and number of matches for promoter TFBS motif searching. This spreadsheet contains the results from scanning the promoters of all genes in the reference set and in each co-expressed cluster, for transcription factor binding site motifs from TRANSFAC (see [Materials and Methods](#s4){ref-type="sec"}, Promoter Scanning). Column 1 contains the TRANSFAC matrix name. Column 2 contains the minimum MotifLocator match score required for the given PWM to be identified as matching the sequence at a given chromosomal location. Column 3 contains the number of matches within the set of 7,492 reference promoter sequences. Columns 4--35 contain the number of matches for the PWM for each of the 32 co-expressed gene clusters. Section 2 contains the *P* values of the enrichments of the PWM matches within each of the 32 clusters (see [Materials and Methods](#s4){ref-type="sec"}, Promoter Scanning). Row 2 indicates the number of genes whose promoters were scanned, for each cluster. The number of matches for each motif within each of the clusters is shown in a second section of the spreadsheet, starting at row 154). (0.15 MB XLS) ###### Click here for additional data file. ###### Time-lagged correlation data for all (TF,target) gene pairs in which a motif associated with the TF gene was found to match within the promoter region of the target gene. Column 1 contains the transcription factor gene symbol. Column 2 contains the transcription factor gene\'s Affymetrix probeset ID. Column 3 contains the target gene symbol. Column 4 contains the target gene\'s Affymetrix probeset ID. Column 5 indicates the co-expressed gene cluster (1-32) of which the target gene is a member. Column 6 indicates the time-lagged correlation coefficient between the TF and the target genes, at the optimal time lag. Column 7 indicates the optimal time lag selected for the gene pair. Column 8 contains the score assigned to the motif match by MotifLocator. (7.84 MB XLS) ###### Click here for additional data file. ###### ChIP-on-chip data. Results of five ChIP-on-chip assays for predicted (TF,cluster) pairs. Each row in the table shows integrated data sources for a specific gene target. Column 1 indicates the TF gene predicted to regulate the target cluster. Column 2 gives the probeset of the TF gene. Column 3 indicates the gene symbol of the target gene. Column 4 gives the target gene probeset. Column 5 gives the co-expressed cluster of which the target gene is a member. Column 6 gives the score for the best motif match for the indicated TF, within the promoter of the target gene (a blank cell indicates that no above-threshold motif match was found, at the 1 match per 10 kbp level of stringency). Column 7 indicates the *P* ^tlc^ from time-lagged correlation. Column 8 indicates whether the gene\'s promoter region was represented on the promoter array. Column 9 indicates the ChIP-on-chip *P* value; a blank cell in this column indicates that no significant ChIP-on-chip binding was found (see [Materials and Methods](#s4){ref-type="sec"}, ChIP-on-chip Validation). (0.05 MB XLS) ###### Click here for additional data file. ###### ChIP-on-chip enrichment results for co-expressed gene clusters that are well-represented on the promoter array. Each row in the table gives results for the ChIP-on-chip assay for a particular cluster and for a particular TF target. Each row in the table is associated with a particular cluster and a particular TF target, for all pairings of p50/NFKB1 and IRF1 with the nine clusters for which at least 30% of the member genes were represented on the tiling array. The first column indicates the TF target. The second column gives the cluster number. The third column gives the number of genes on the ChIP-on-chip array for which binding was observed upstream of the transcription start site. The fourth column gives the number of genes within the cluster, that were represented on the ChIP-on-chip array. The fifth column gives the number of genes within the cluster that showed evidence of TF binding in the upstream region, in the ChIP-on-chip assay. The sixth column gives the fraction of genes in the cluster that are represented on the array. The seventh column gives the enrichment *P* value for the ChIP-on-chip hits within the cluster (see [Materials and Methods](#s4){ref-type="sec"}, ChIP-on-chip Validation). The eighth column gives the motif match enrichment *P* value based on sequence scanning (see [Materials and Methods](#s4){ref-type="sec"}, Promoter Scanning). The ninth column gives the *P* value based on the time-lagged correlation of expression profiles of the TF gene and the genes within the target cluster. The tenth column gives the average time lag, between the TF gene and the genes within the target cluster. The eleventh column gives the combined *P* value based on motif match enrichment and time-lagged correlation (see Equation 13). (0.02 MB XLS) ###### Click here for additional data file. ###### List of key materials and reagents. Column 1 indicates the type of material (mouse strain or stimulus reagent). Column 2 indicates the specific strain or reagent. For mutant mouse strains, the Mouse Genome Informatics accession number of the allele is provided. Column 3 indicates the source laboratory from which the mouse strain or reagent was obtained. (0.05 MB DOC) ###### Click here for additional data file. ###### Summary of probeset selection criteria. Each row describes a set of data selection criteria, for a specific purpose. For a detailed explanation of each set of criteria, see [Materials and Methods](#s4){ref-type="sec"}, Probeset Selection. Column 1 states the purpose of the set of selection criteria. Column 2 indicates the minimum log~2~ absolute probeset intensity that must have been recorded in at least one experiment, for the gene to be included in the selection described in Column 1. Column 3 indicates the false discovery rate used to determine the *P* value cutoffs for each of the seven time-course experiments used for differential expression testing (see [Materials and Methods](#s4){ref-type="sec"}, Differential Expression Testing); "n/a" means that no differential expression test was applied, for genes in the indicated row. Column 4 gives the number of probesets resultant from the indicated selection criteria. (0.03 MB DOC) ###### Click here for additional data file. ###### The total numbers of genes that possess gene ontology (GO) annotations, from each GO term hierarchy. Representative genes are selected from the set of annotated Affymetrix probesets as described in [Materials and Methods](#s4){ref-type="sec"}, Probeset Selection. (7.84 MB XLS) ###### Click here for additional data file. The authors are grateful to the following researchers for kindly providing mutant mouse strains (see [Table S15](#pcbi.1000021.s033){ref-type="supplementary-material"}): Shizuo Akira, Bruce Beutler, Tsonwin Hai, and Günther Schütz. SR thanks Matti Nykter, Jeroen de Ridder, and Elain Fu for helpful discussions and advice. The authors thank Dave Campbell, April Clark, Bridget Coila, Kerry Deutsch, Kristi Hamilton, Robert Hubley, Sarah Killcoyne, Martin Korb, Aaron Lampano, William Longabaugh, Bruz Marzolf, Alex Nachman, Irina Podolsky, David Rodriguez, and Tetyana Stolyar for technical assistance. The authors have declared that no competing interests exist. This work was supported by NIH grant U54-AI54253 and NIH contract HHSN272200700038C from the National Institute of Allergy and Infectious Diseases. [^1]: Conceived and designed the experiments: SR DZ KK. Performed the experiments: DZ KK MG EG CJ VL GN CR NY. Analyzed the data: SR BL. Contributed reagents/materials/analysis tools: SR VT BL JR AR. Wrote the paper: SR. Helped develop the statistical test used in this work: SK. Carried out qPCR and microarray expression experiments: DZ. Generated the SMAD/TGIF protein interaction network: VT. Wrote the ChIP-on-chip data processing program and processed the ChIP-on-chip data: BL. Developed a hand-curated list of murine cytokines and chemokines: NY. Supervised the project and provided advice on the writing of the manuscript: AA IS.

Introduction {#sec0005} ============ Baroreflex index and hemodynamics {#sec0010} --------------------------------- Baroceptors, baroreceptors, mechanoreceptors or even mechanoceptors are receptors responsible for the contraction and dilation of blood vessels.[@bib0315] They are located mainly in the carotid sinus[@bib0320] and aortic arch[@bib0325], [@bib0330] and have the function of detecting variations in blood pressure and transmitting this information to the central nervous system. This information generates responses from the autonomic nervous system, modulating the functioning of the blood circulation (Baroreceptor reflex).[@bib0335] The more the baroceptors are activated, the more functional oxygen is transported to the lungs and the brain.[@bib0340] This functioning with active and constant movement of these mechanoceptors also seems to indicate that there is a good flow in the blood displacement at the level of blood microcirculation, also called hemodynamics.[@bib0345] Adequate hemodynamics depends on the state of cardiovascular health. According to Houston (2014), the blood vessel has a finite number of responses to an infinite number of insults.[@bib0350] Finite vascular responses are oxidative stress (Reactive Oxygen Species and Reactive Nitrogen Species),[@bib0355] inflammation[@bib0360] and exaggerated immune reaction of the organism[@bib0365] and are related to the pathophysiology of vascular disease, which lead to an abnormal vascular biology, such as endothelial dysfunction (ED) and vascular smooth muscle dysfunction (VSMD).[@bib0355] In this infinite estimate of insults, there is hypertension, dyslipidemia,[@bib0370] diabetes,[@bib0375] hyperglycemia,[@bib0380] obesity,[@bib0385] smoking,[@bib0390] insulin resistance,[@bib0395] increased homocysteine,[@bib0400] among others, which enable vascular responses with cytokines,[@bib0405] chemokines, [@bib0410] adhesion molecules,[@bib0415] heat shock proteins[@bib0420] and antibodies,[@bib0425] leading to inflammation,[@bib0430] oxidative stress[@bib0435] and autoimmune dysfunction.[@bib0440] Atherosclerosis and vascular disease are postprandial phenomena, that is, they arise after eating allergenic and inflammatory foods,[@bib0445] associated with hyperglycemia and hypertriglyceridemia, with endothelial dysfunction being the earliest stage of vascular disease, which can be detected through the baroreflex index, hemodynamics[@bib0445] and biomarkers already studied in the literature, linked to cardiovascular diseases.[@bib0450] In this sense, this study aimed to report the values of the baroreflex and hemodynamic index related to biomarkers that signal obstruction of the blood microcirculation found in the 57-year-old male patient, who died after being affected by COVID-19. Methods {#sec0015} ======= Assessments were performed by functional neurometry (FN) and laboratory tests. The evaluation with functional neurometry was performed in an air-conditioned room at a temperature of 22 ± 2 °C. The capacity, functionality and elasticity of the blood vessels are evaluated indirectly in 3 (three) positions. Thus, the FN was performed in 3 positions: dorsal decubitus, standing up and orthostatic, called DSO analysis by the author, organizer and creator of the method, Nelson Alves Pereira Júnior.[@bib0445] The sympathetic autonomic nervous system represents the vasoconstrictions measured at the frequencies of 0.01--0.04 Hz and from 0.04 to 0.20 Hz and the parasympathetic represents the vasodilatations measured at the frequencies of 0.20--0.50 Hz.[@bib0445] The five categories of this evaluation protocol of DSO analysis were named by the author as: anxiety control; physiological response; baroreflex index; hemodynamics (blood flow velocity); and brain neurometry. Detailed descriptions of each category, which captures biological signals by sensors calibrated in Series and/or Fourier Transform algorithms, have been published in the Journal of Psychology and Psychotherapy Research (DOI: <https://doi.org/10.12974/2313-1047.2020.07.1>).[@bib0445] In this case report, the categories were approached briefly with an emphasis only on the expected values to facilitate the reader\'s understanding of the interpretation of the results. In Functional Neurometry, anxiety control is measured through the galvanic skin response.[@bib0455] Thus, the electrodermal resistance is indirectly measured by means of the sweat gland.[@bib0460] In this context, as sweat on the skin allows greater electrical conduction,[@bib0465] it can be estimated that with a greater amount of sweat there seems to be less energy stored in the liver (glycogen) for a person\'s adrenal to enable the fight and/or flight process or for that adrenal to control anxiety through the hormones aldosterone, cortisol, adrenaline and dehydroepiandrosterone.[@bib0445] The anxiety control category is calculated using a scale with agreed values from 0 to 100%, in which the level of energy provided by the functional reserve already absorbed by the enterocytes and already stored in the liver can be measured indirectly,[@bib0470] awaiting the need to pass into the bloodstream by glycogenolysis[@bib0475] to make the fight and/or flight process feasible.[@bib0480] The minimum expected in this category is 75%. The physiological response is measured by varying the peripheral temperature,[@bib0485] demonstrating whether or not there is adequate functionality of blood vessel elasticity.[@bib0490] This elasticity refers to an expected variation in vasoconstriction and vasodilation,[@bib0490] legitimizing the sympathetic (fight and flight) and parasympathetic (relaxation).[@bib0495] The signal origin is made through the peripheral temperature sensor, which scales the unit of measurement in degrees Celsius. The originated values generate 2 more important results, as a percentage of variability of the sympathetic and parasympathetic systems and the temperature values in the thermoregulation. Both are expressed on a scale ranging from 0 to 100 (%) or (°C). The location of this sensor is in the proximal phalanx of the ring finger.[@bib0500] The expected values for an ideal thermoregulation are between 31.5 °C and 32.5 °C.[@bib0485] Baroreflex behavior is the ability of blood vessels to contract and dilate, which allows the neuroscientist to observe the level of functional oxygen that will better signal red blood cell displacement,[@bib0505] in the process of transporting oxygen within the blood vessels. In functional neurometry, the balance of these contractions and dilations (baroreflex index) must present a percentage above 90%.[@bib0510] Some studies have been warning about the importance of this measure of the low baroreflex index found in patients with obstructive sleep apnea,[@bib0515] with slight respiratory changes[@bib0520] asymptomatic or imperceptible, but which gradually cause mitochondrial suffering.[@bib0525] Other studies also demonstrate a reduction in the baroreflex index associated with anemia, inflammation,[@bib0445] to a lack of vitamin C,[@bib0530], [@bib0535] folic acid[@bib0540] and sedentary lifestyle[@bib0540] and may contribute to the manifestation of possible negative effects in the prevention of atherosclerosis. Hemodynamics[@bib0545] may be directly related to endothelial dysfunction and, consequently, to the risk of cardiovascular disease.[@bib0550] In functional neurometry, the ideal hemodynamics should be below 10%. When above this percentage, it indicates a reduction in blood flow speed caused by peripheral vascular resistance, cardiac output and/or inflammatory response.[@bib0445] Case report {#sec0020} =========== A fifty-seven-year-old male patient, assessed on February 10, 2020 at the Brain Institute of Pernambuco -- *Instituto do Cérebro de Pernambuco* (*ICerPE*). Before the assessment by functional neurometry, the following measurements were taken: weight \[100 kg\], height \[1 m 78 cm\], waist circumference \[117 cm\] and body mass index \[31.6\], fulfilling the classification criteria of mild obesity, with 23.3 kg overweight. The patient reported that he did not frequently do physical activity (sedentary). The patient went to *ICerPE*, taken by his daughter, with the objective of undergoing a complete assessment with functional neurometry and quantitative electroencephalography, after learning that there was treatment for mood disorder and generalized anxiety disorder. When the evaluation with functional neurometry ended, the *ICerPE* partner physician and researcher asked him to perform a blood test of the following biomarkers: blood count: \[red blood cells (RBCs); hemoglobin (Hb); hematocrit (Hct); mean corpuscular volume (MCV); mean corpuscular hemoglobin (MCH); mean corpuscular hemoglobin concentration (MCHC); total leukocytes; red cell distribution width (RDW)\]; serum ferritin (SF); transferrin saturation (TS); vitamin B12; homocysteine; C-reactive protein (CRP); fibrinogen; 25-hydroxy-vitamin-D3 (D3); apolipoprotein A-1 (Apo-A1); apolipoprotein B (Apo-B); glutamic oxalacetic transaminase (GOT); glutamic-pyruvic transaminase (GPT); gamma-glutamyl transpeptidase (Gamma GT); total cholesterol (TC); triglycerides (TG), and; high-density lipoprotein (HDL) and low-density lipoprotein (LDL). After this stage, the patient went on to confinement due to the COVID-19 pandemic. However, on April 23, 2020, the patient\'s daughter sent us a message, saying that her father had not respected the confinement recommendations and that, unfortunately, he had died on April 18, 2020. Results of the assessment by functional neurometry and biomarkers {#sec0025} ================================================================= The result of the patient\'s anxiety control assessed indirectly by the functional reserve, in percentage, was 7.27% \[expected value: 75%\] ([Table 1](#tbl0005){ref-type="table"}). This result is usually compatible with exhaustion of the adrenal gland and/or fatty liver. In a semiological analysis, as the patient was 23.3 kg overweight and with an abdominal circumference of 117 cm, diagnostic compatibility increased in probability. Even though, some thin patients also have hepatic steatosis. The gamma GT level found was 67 \[Reference value: 7--45 U/L\] ([Table 2](#tbl0010){ref-type="table"}). The high level of gamma GT, exceeding the laboratory\'s maximum reference value, is related to insulin resistance, preventing diabetes, and can also signal toxic and/or free radical intoxication, in addition to indicating an affected liver.Table 1Synthesis of the results of the functional neurometry exam of the male patient who died after being affected by COVID-19.Table 1Functional neurometryAnxiety control/Functional reserveBaroreflex indexHemodynamicsPhysiological responseExpectedResultExpected minimumResultExpected maximumResultExpected rangeResult75%7.27%90%60.92%10%26.07%31.5--32.5%33.77%Table 2Synthesis of the results of the biomarkers of the male patient who died after being affected by COVID-19.Table 2Blood biomarkers**RBCHbHctMCV**Reference valueResultReference valueResultReference valueResultReference valueResult4.4--6.6 million/μL4.4213--1813.839--54%42.3%76--9695.7

**MCHMCHCTotal leukocytesRDW**Reference valueResultReference valueResultReference valueResultReference valueResult27 - 3331.232--3632.64.000--10.000/μL5.43011--1513.6

**Serum ferritinTransferrin saturationB12 vitaminHomocysteine**Reference valueResultReference valueResultReference valueResultReference valueResult23.90 and 336.20 ng/mL161.220--50%20.8210.0--980.0 pg/mL1535--15 μmol/L16.5

**C-reactive proteinFibrinogen25-Hydroxy-vitamin D3Apolipoprotein A-1**Reference valueResultReference valueResultReference valueResultReference valueResult\<0.3 mg/dL1.74200--400 mg/dL395\>3021.6104--202 mg/100 mL142

**Apolipoprotein BGlutamic oxalacetic transaminaseGlutamic-pyruvic transaminaseGamma GT**Reference valueResultReference valueResultReference valueResultReference valueResult66--133 mg/100 mL113\<34 U/L1710--49 U/L117--45 U/L67

**Total cholesterolTriglyceridesHDLLDL**Reference valueResultReference valueResultReference valueResultReference valueResult\<190 mg/dL200\<150 mg/dL611\>40 mg/dL42\<130 -- Low Risk36

**CT/HDL ratioTG/HDL ratioLDL/HDL ratioAPO B/APO A1 ratio**Reference valueResultReference valueResultReference valueResultReference valueResult≤3.34.76≤2.514.4≤2.30.85≤0.50.7957 The gamma-GT catalyzes a transpeptidation reaction that is involved in the glutathione metabolism. Glutathione is abundant in the epithelial fluid of the lung lining. However, little is known about gamma-GT expression in pulmonary alveoli epithelial cells. There is, however, a relevant study[@bib0555] showing that the pulmonary alveolar epithelial cell type 2 expresses the gene for gamma-GT. The gamma-GT expression in the type 2 pulmonary alveolar epithelial cell is via mRNA III, a transcript that was initially cloned from the liver. This cell synthesizes the gamma-GT protein and releases enzyme activity in a pool associated with surfactant in the pulmonary alveolus.[@bib0555] The results of this study[@bib0555] suggest that the surfactant may play an expanded role in the biology of lung cells as a vehicle for the redistribution of proteins anchored in amphipathic signals across the gas exchange surface of the lung. The gamma GT transports the glutathione that is outside the cell into the cell and thus keeps the levels of reactive oxygen species low. Some epidemiological studies[@bib0560], [@bib0565] have shown that gamma GT levels, within laboratory reference values, closest to the highest reference value, are associated with cardiovascular risks, with predictors of future cardiovascular diseases, hypertension, stroke and type II diabetes mellitus. There is still much to be researched and studied to better understand the relationship between the liver and the lung during deaths caused by COVID-19. However, lung injuries have been considered as the main damage caused by SARS-CoV-2 infection and there are already findings reporting that liver damage occurred during the course of the disease in severe cases.[@bib0570] Regarding the baroreflex index, the patient presented 60.92% \[expected value: \>90%\] ([Table 1](#tbl0005){ref-type="table"}) and hemodynamics of 26.07% \[expected value: \<10%\] ([Table 1](#tbl0005){ref-type="table"}), values found in patients with endothelial dysfunction usually in outpatient consultations. The endothelial dysfunction in the case of the patient in this study seems to have been confirmed by the level of his gamma GT, reported in the above-mentioned paragraphs, of 67 \[reference value: 7--45 U/L\] ([Table 2](#tbl0010){ref-type="table"}). There is an epidemiological study in the literature that corroborates this interpretation. This study was carried out with 4.266 participants, demonstrating the association of gamma GT with C-reactive protein in men and women and, mainly, with the stiffening of the brachial artery in men.[@bib0565] The endothelial dysfunction in this case report can still be identified by several biomarkers, among which are: low 25-hydroxy-vitamin D3 of 21.6 \[reference value: \>30\] ([Table 2](#tbl0010){ref-type="table"}), high CRP of 1.74 \[reference value: \<0.3 mg/dL\] ([Table 2](#tbl0010){ref-type="table"}) and the physiological response of 33.77% \[reference value: 31.5--32.5\] ([Table 2](#tbl0010){ref-type="table"}). The digestive overload gives meaning to the low absorption of vitamin B12, found to measure 153 in the patient\'s blood test \[reference value: 210--980 pg/mL\] ([Table 2](#tbl0010){ref-type="table"}), The decrease in vitamin B12 also causes an increase in homocysteine 16.5 \[reference value: 5--15 μmol/L\] ([Table 2](#tbl0010){ref-type="table"}), which further increases the probability of endothelial dysfunction. The patient in this study furthermore had a low 25-hydroxy-vitamin D3 and high C-reactive protein. However, an important finding in the literature[@bib0575] analyzed the relationship of vitamin D3 and C-Reactive protein with asymptomatic individuals supplemented with 21 ng/mL vitamin D3, and found no significant reduction specifically in C-reactive protein. However, in different contexts from the previous study, another very relevant study[@bib0580] demonstrated that patients supplemented with vitamin D3 of 10 ng/mL for 3 months showed improvement in endothelial function, reduction of oxidative stress and improvement in insulin sensitivity. This finding may encourage the scientific community to improve respiratory microcirculation, as obstruction and vascular resistance have been a major concern for scientists during the COVID-19 pandemic.[@bib0585] And to further encourage researchers in relation to the benefits of vitamin D, an epidemiological study by Melamed et al. (2008) found that low levels of vitamin D3 increase the risk of mortality. This study used NHANES III data, collected from 1988 to 1994, studying 13,331 adult individuals of both genders with an average follow-up of 8.8 years. The result was 1,806 deaths, 43% from cardiovascular disease, 26% from cancer, 6% from infectious diseases and 5% from external causes. It was found that people with serum vitamin 25-OH-vitamin-D3 levels between 40-49 ng/mL had a lower mortality rate.[@bib0590] The indication of digestive overload found in the patient in this study using the functional neurometry[@bib0445] may be related to parasites, but it may also have been due to sedentary habits, which seems to explain his low vitamin B12 as well. This result seems to reinforce the hypotheses about the obstructions of the respiratory microcirculation.[@bib0585] Kwok et al. (2012)[@bib0595] conducted a study using vitamin B12 in 50 vegetarian subjects who had serum B12 \< 150 pmol/L (\<202 pg/mL) and the results showed a significant serum increase in B12, a decrease in homocysteine and a significant improvement in dilation mediated by brachial flow from 6.3 ± 1.8% to 7.4 ± 1.7%. The conclusion of this study was that B12 supplementation improved the arterial function of vegetarians with subnormal B12 levels, thus proposing a new strategy for the prevention of atherosclerosis. In this context of importance of vitamin B12, another study[@bib0600] recently published reinforces the importance of vitamin B12 in the treatment of COVID-19, the authors point out that the virus of the Coronaviridae family (COVID-19) has a single-stranded positive RNA genome. This genome encodes the nsp12 protein, which hosts the RNA polymerase-dependent RNA (RdRP) activity responsible for replicating the viral genome. Thus, the authors[@bib0600] report that a nsp12 homology model was prepared using the SARS nsp12 (6NUR) framework and that it was used to screen silica to identify molecules among FDA-approved natural products or drugs that could potentially inhibit nsp12 activity. However, this experiment showed that vitamin B12 (methylcobalamin) can bind to the active site as an inhibitor of the nsp12 protein.[@bib0600] Regarding total cholesterol (TC), triglycerides (TG), HDL and LDL, this study highlighted the results of TG and HDL of the patient who died after acquiring COVID-19. In addition to the TG being well above 611 mg/dL ([Table 2](#tbl0010){ref-type="table"}) of the expected reference value \[\<150 mg/dL\], there are findings in the literature demonstrating ideal values expected from CT/HDL ratios ≤3.3, TG/HDL ≤2.5 and LDL/HDL ≤2.3,[@bib0605] called atherogenicity indexes, functioning as a safe predictor of high lipoprotein density with the risk of coronary disease. Data from the Framingham study show that each 1 mg/dL increase in HDL decreases the coronary disease mortality rate by 2% for men and 3% for women.[@bib0605], [@bib0610] However, Tian and Fu (2010)[@bib0605] establish that the greater the result of the relationship between CT/HDL and TG/HDL, the less will be the protective effect of HDL, because it will indicate a greater quantity of small HDLs (HDL Pre 1) and lesser quantities of large HDLs (HDL2b) and add that when these relationships worsen, not only the HDL amount worsens, but the type of HDL as well.[@bib0605] The HDL is responsible for much more than transporting cholesterol from arteries and peripheral tissues to the liver; it has antioxidant action, anti-inflammatory action, pro-fibrinolytic action, antithrombotic action, and even reverse cholesterol transport.[@bib0615] Thus, when the relationships based on the above-mentioned study were replicated in this case report, the following results were found: CT/HDL = 4.76, TG/HDL = 14.4 and LDL/HDL = 0.85, indicating that the patient in this case report had a high level of atherogenicity. It also seems relevant that, if there is a decision to initiate early care with the population to improve the microvascular state in order to have better conditions to face the COVID-19 virus, it will be necessary to better understand other relationships, such as apolipoprotein A1 (APO-A1) and apolipoprotein B (APO-B). O'Keefe et al. (2008) measured APO-A1 and saw that APO-A1 does not become a risk factor when in high concentrations.[@bib0620] A large HDL can be an atherogenic HDL because it can donate cholesterol in peripheral tissues through the scavenger receptor class B1 (SR-B1), unless the APO-A1 is also large.[@bib0620] Thus, it also seems relevant to investigate the relationship between Apo B and Apo A1. In this case report, when the studied patient\'s APO B \[113 mg/100 mL\] was divided by his APO A1 \[142 mg/100 mL\], a result of 0.7957 was found. However, a study by Tian and Fu (2010)[@bib0605] showed that people with an APO B/APO A1 ratio with a value equal to or less than 0.5 die less from cardiovascular disease. Conclusion {#sec0030} ========== The assessment of the baroreflex and hemodynamic index by functional neurometry by health professionals can anticipate the identification of people who will need to have a blood test to investigate the biomarkers that indicate subclinical inflammation, hepatic steatosis and endothelial dysfunction. This study found subclinical inflammation and endothelial dysfunction. However, as this study is only a case report, it is suggested to conduct a clinical study with several patients to ensure statistical significance. Conflicts of interest {#sec0035} ===================== The authors declare no conflicts of interest.

Introduction ============ *Sclerotinia sclerotiorum* is a plant pathogenic fungus with a remarkably broad host range. An early literature review by [@evx030-B11]) identified 408 species of plants in 278 genera and 75 families that are susceptible to infection by *S. sclerotiorum*. A number of these host species are economically important crops, for example, *Brassica napus* (canola/oilseed rape), *Glycine max* (soybean), *Beta vulgaris* (sugar beet), *Arachis hypogaea* (peanut), and *Lactuca sativa* (garden lettuce) ([@evx030-B24]; [@evx030-B72]; [@evx030-B67]; [@evx030-B48]; [@evx030-B18]). The broad host range of *S. sclerotiorum* is in contrast to most other well-studied plant pathogenic fungi that infect only a single or small number of host species. Illustrating this point, a recent article listed the "top ten plant pathogenic fungi," based on economic destructiveness and scientific importance ([@evx030-B23]). This article was produced following a survey of 495 international experts on fungal diseases of plants. The list contained eight narrow host range plant infecting fungal species or genera and only two broad host range pathogens, *Botrytis cinerea* and *Fusarium graminearum*. Although the fungus *Fusarium oxysporum* was also included, this species is actually a species complex, which comprises numerous *formae speciales* with different host specificities ([@evx030-B80]; [@evx030-B40]). Thus, much of our current knowledge of plant infecting fungi is derived from narrow host range pathosystems, which often exhibit particular characteristics resulting from a pronounced evolutionary arms race between host and pathogen. This arms race often results in pathogens evolving virulence genes that are counteracted by corresponding resistance loci in the host, which are themselves counteracted by new or variant virulence genes in the pathogen, *ad infinitum*. This is termed a "gene-for-gene" interaction, and was first observed by Flor in the interaction between the fungal parasite *Melampsora lini* and its host, flax (*Linum usitatissimum*) ([@evx030-B33]). Plant pathogenic fungi that coevolve with their hosts in a gene-for-gene manner often exhibit compartmentalized genomes that contain alternating nonrepetitive, gene-rich regions and repetitive, gene sparse regions. Virulence-associated genes are often clustered in repetitive, gene sparse regions, and it is thought that this clustering and compartmentalization allows for their rapid evolution through transposable element (TE)-mediated duplication and mutation, without affecting housekeeping genes. This theory is known as the "two-speed-genome" hypothesis ([@evx030-B44]; [@evx030-B77]; [@evx030-B64]; [@evx030-B79]; [@evx030-B76]; [@evx030-B20]; [@evx030-B69]; [@evx030-B26]; [@evx030-B31]). Enhanced mutation rates in repeat-rich regions in fungi may arise through a process termed repeat-induced point mutation (RIP). This process is a genomic defense against excessive TE activity, which could be deleterious if left uninhibited. It causes point mutations in duplicated sequences through 5′-methylation of cytosine followed by deamination to thymine. As a result, RIP-affected genomic regions often exhibit pronounced A/T content concomitant with a decrease in G/C content ([@evx030-B85]; [@evx030-B45]; [@evx030-B86]). Virulence-associated proteins are often found to have been affected by this process, which may be important for their rapid adaptation ([@evx030-B79]). Many of the virulence-associated genes of plant pathogenic fungi studied to date encode small secreted proteins, whose sole purpose is to manipulate plant defenses to advance fungal infection ([@evx030-B90]; [@evx030-B60]). These proteins, termed "effectors," have been discovered in multiple plant pathogenic fungi and exhibit numerous different functions depending on fungal lifestyle. For example, necrotrophic fungi, which require dead tissue on which to feed, often produce effectors that promote cell death, whereas biotrophic fungi, which require living tissue, produce effectors that prevent cell death ([@evx030-B37]; [@evx030-B19]; [@evx030-B22]; [@evx030-B61]; [@evx030-B103]; [@evx030-B63]). Hemibiotrophic fungi, which require both living and dead tissue at different life cycle stages, may produce different effectors at different time points during infection ([@evx030-B65]; [@evx030-B52]; [@evx030-B55]; Rudd et al. 2015). Many effector genes studied are species or even isolate specific and interact directly or indirectly with the products generated by particular intraspecifically polymorphic loci in the host, thus conforming with the gene-for-gene hypothesis ([@evx030-B59]; [@evx030-B15]; [@evx030-B14]; [@evx030-B36]; [@evx030-B38]). However, several effectors are also conserved across various fungal species and have seemingly similar roles in pathogenesis on plants, interacting with loci conserved within a species and/or between species ([@evx030-B94]; [@evx030-B22]; [@evx030-B83]; [@evx030-B78]; [@evx030-B49]; [@evx030-B21]; [@evx030-B8]). In *S. sclerotiorum*, thus far, there have been several proteins described with effector-like properties and/or effector-like activities *in planta*. These proteins include a polygalacturonase enzyme, sspg1d, that interacts with a *B. napus* C2 domain containing protein ([@evx030-B100]); an integrin-like protein, SSITL, that suppresses defense responses in *Arabidopsis thaliana* (although, in addition to the activity of SSITL *in planta*, deletion mutants for the cognate gene exhibited abnormal hyphal tip branching, slower growth *in vitro* and abnormally small sclerotia) ([@evx030-B106]); two necrosis and ethylene-inducing proteins, SsNep1 and SsNep2, that cause necrosis when heterologously expressed in *Nicotiana benthamiana* ([@evx030-B21]); a cutinase, SsCut, that causes cell death in a range of plant species, including the nonhost, wheat (*Triticum aestevum*) ([@evx030-B105]); a polygalacturonase protein that induces calcium signaling and host cell death in soybean ([@evx030-B107]); a hypothetical secreted protein that, when disrupted, attenuates virulence in *S. sclerotiorum* ([@evx030-B58]); a small, cysteine-rich secreted protein with a cyanoviron-N homology (CVNH) domain, which attenuates virulence when deleted ([@evx030-B62]); and, most recently, a secreted protein with no known functional domains, SsSSVP1, seemingly restricted to *S. sclerotiorum* and the closely related species *B. cinerea*, which causes necrosis in *N. benthamiana.* This protein was shown to interact with *N. benthamiana* QCR8, a subunit of the cytochrome b--c~1~ complex of the mitochondrial respiratory chain. It is thought that this activity is what caused necrosis in host tissue ([@evx030-B63]). The starting point for investigations into the evolution and molecular activity of fungal effector genes often begins with analysis of the fungal genome. Numerous studies, including two studies on *S. sclerotiorum*, have attempted to identify effector-like or secreted proteins potentially involved in infection (e.g., [@evx030-B1]; [@evx030-B68]; [@evx030-B42]; [@evx030-B47]). These studies have used various criteria to differentiate effector-like from noneffector-like genes. Primarily, the focus has been on small, cysteine-rich proteins with a predicted secretion signal, as many experimentally characterized effectors exhibit these properties ([@evx030-B87]). Recently, a machine learning approach based on numerous criteria derived from properties of known effectors was applied to effector prediction. This approach was shown to be more sensitive and specific than the traditional, and to some degree arbitrary, approach of filtering through manually selected criteria ([@evx030-B88]). Predicted secretome and effector studies have been facilitated by the huge increase in availability of fungal genome sequences in recent years. However, although there are currently 1,509 fungal genomes that have been sequenced (as of May 26, 2016, source: <http://www.ncbi.nlm.nih.gov/genome/browse/>), many are fragmented and poorly annotated. The regions missing from these genomes are largely composed of repetitive sequence. This is because the technologies used to sequence them produce reads up to a maximum of 1,000 bp (Sanger) and in many cases only 300 bp (Illumina) ([@evx030-B39]). Sequence assemblies based on reads that are substantially shorter than repeated elements do not produce a reliable contiguous sequence as the repeats are often collapsed into a single, short sequence during graph-based assembly ([@evx030-B27]). This poses a particular problem to researchers interested in fungal effectors, which, as aforementioned, often cluster within repetitive genomic regions ([@evx030-B95]). Recently, Pacific Biosciences developed the single molecule real-time sequencing (PacBio) platform, which produces reads of up to 60 kb ([@evx030-B39]). Thus far, two gapless fungal plant pathogen genomes have been assembled using PacBio reads, in combination with optical mapping---a process that uses fluorescent staining in conjunction with restriction enzymes to infer genomic structure based on restriction site patterns across chromosomes ([@evx030-B30]; [@evx030-B98]). The existing *S. sclerotiorum* genome, although of high quality compared with many other fungal genomes, is not complete, with an estimated 1.6 Mb of missing sequence, based on comparison of the assembly with an optical map ([@evx030-B2]). To improve upon the previous genome, its annotations, and subsequent effector analyses, we used the existing optical map in combination with PacBio and Illumina sequencing. Using this improved genome sequence we identified a number of novel effector-like protein encoding genes, several of which were significantly upregulated during infection. We also recharacterized TEs in the *S. sclerotiorum* genome and demonstrated that the five most abundant TEs have been subjected to RIP and exhibit an increased RIP index relative to a randomized set of control sequences. Additionally, comparative analyses with the genomes of several host-specific filamentous plant pathogens revealed a lack of strong association between secreted or effector-like protein encoding genes and repetitive elements or RIP in the genome of *S. sclerotiorum*. We conclude that the broad host range necrotroph *S. sclerotiorum* does not exhibit a classic "two-speed genome" as found in host-specific biotrophic and hemibiotrophic fungi and oomycetes. Although a more subtle signature of potential enhanced selection of secreted proteins was observed, predicted effector-like protein encoding genes specifically were not found to be localized to repeat-rich or RIP-affected regions more than other genes. Materials and Methods ===================== Growth, nucleic acid extraction, and sequencing of *S. sclerotiorum* Isolate 1980 --------------------------------------------------------------------------------- Various procedures were used to grow *S. sclerotiorum* and extract and sequence nucleic acids for the various analyses presented in this article. All experiments used WT 1980 or a mutant derivative thereof. Five data sets in total were used. Data set 1 consisted of one *in vitro* sample and six *B. napus* infection time points (1, 3, 6, 12, and 24 h postinoculation \[hpi\]) in triplicate; data set 2 consisted of several *in vitro* samples; and data set 3 consisted of one *in vitro* sample and two *A. thaliana* infection time points for both WT 1980 and the oxaloacetate acetyle hydrolase *S. sclerotiorum* deletion strain Δ*oah* ([@evx030-B57]). These three data sets were used for RNASeq-based gene calling. Data set 1 was also used for differential expression analysis of effector candidates. Data sets 4 and 5 were single *in vitro* samples and were used for Illumina genomic and PacBio sequencing, respectively. Experimental procedures used in the generation of all these data sets are detailed in the [supplementary file 1](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online. Assembly of PacBio Long Reads and Correction with Illumina Short Reads ---------------------------------------------------------------------- A *de novo* genome assembly of *S. sclerotiorum* strain 1980 was generated using MHAP version 1.5b1 ([@evx030-B10]) with default settings. To assess contiguity of the assembled sequences, they were aligned to the previously generated optical map with MapSolver version 3.2 (OpGen, Gaithersburg, MD). Overlapping or adjacent sequence contigs were manually joined and gap-filled with PBJelly2 version 15.2.20 ([@evx030-B29]) with default settings, except for the assembly stage where "-maxTrim" and "-maxWiggle" were set to 15 kb. A single region on chromosome 9 was not completely assembled *de novo* using MHAP, and this region was inferred from the previous *S. sclerotiorum* genome assembly ([@evx030-B2]). The final assembly was error corrected after manual gap filling using Quiver ([@evx030-B17]). Illumina short reads were first trimmed using Trimmomatic version 0.36 ([@evx030-B12]) then mapped to the *de novo* assembly using Bowtie version 2.2.6 ([@evx030-B54]). Modules from the Genome Analysis Toolkit version 3.5-0-g36282e4 ([@evx030-B66]) and Picard Tools version 2.1.0 (available at <http://broadinstitute.github.io/picard/>) were then used to call variants between the *de novo* assembly and the Illumina reads. After mapping to the version two genome sequence, GATK was used to create a consensus sequence from the Illumina reads and PacBio *de novo* assembly. Substitutions, of which there were very few, were ignored and only insertions and deletions (InDels) were considered. Polymorphisms that were identified based on a depth of less than 30× or a mapping quality of less than 30 were also excluded. Alignment of RNA Sequencing Data for Use in Gene Predictions ------------------------------------------------------------ RNASeq data obtained from the experimental conditions enumerated in the previous section were used for gene prediction. Reads were trimmed using Trimmomatic version 0.36 ([@evx030-B12]). Reads from *in planta* samples were first binned based on whether they mapped better to the *S. sclerotiorum* genome or the host genome using BBSplit version 34.86 (available at <https://sourceforge.net/projects/bbmap/files/>). For this purpose, the *B. napus* genome was downloaded from <http://www.genoscope.cns.fr/brassicanapus/> and the *A. thaliana* genome was downloaded from <https://www.arabidopsis.org/> ([@evx030-B16]; 2000). Reads were then mapped to the *S. sclerotiorum* genome using Top Hat version 2.1.0 ([@evx030-B97]) with the following nondefault settings: "--min-intron-length 10" and "--max-intron-length 5000." Cufflinks version 2.2.1 ([@evx030-B97]) was used with default settings to assemble spliced transcripts from read mappings. Transcripts from all experimental conditions were merged for use in gene prediction using the cuffmerge module of Cufflinks. Gene Prediction Using *Ab Initio* and RNASeq-Guided Softwares, Protein Homology, Transcript, and Expressed Sequence Tag Evidence -------------------------------------------------------------------------------------------------------------------------------- The following gene prediction software packages were used to generate predictions for the version two *S. sclerotiorum* genome: Augustus version 2.5.5 ([@evx030-B89]), Coding Quarry version 1.2 ([@evx030-B92]), and GeneMark-ES version 3.1 ([@evx030-B13]). These softwares were used with default settings and gene predictions were supplied to Evidence Modeler ([@evx030-B43]) alongside assembled transcripts, available expressed sequence tags (ESTs) and protein homology data, and weighted accordingly. A detailed description of methods used in gene calling is supplied in the [supplementary file 2](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online. Genomes used for protein homology information are detailed in the [supplementary table 1](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online. Comparison of Version One Gene Predictions with Version Two Gene Predictions ---------------------------------------------------------------------------- To determine what genomic loci had changed or remained the same, two approaches were used: 1) Bedtools version 2.17.0 ([@evx030-B75]) was used to determine which new gene predictions overlapped annotations that had been transferred from the previous assembly to infer fusions or splits and 2) nucleotide sequence predictions from both genome assemblies were subjected to a BLAST analysis in both directions using BLASTn version 2.2.31 to determine differences in gene sequence length and content. Approach two was used to generate a table of version two loci and corresponding version one loci. The script used for the reciprocal BLAST analysis is available at: <https://github.com/markcharder/GeneralBioinformaticsScripts/blob/master/reciprocalBestHitsBlast.pl>. Validation of Version Two Gene Annotations ------------------------------------------ To determine overall concordance with RNASeq data, CDS sequences from both genome versions' gene predictions were compared with Cufflinks transcripts (derived from the previously mentioned read mapping) using BLASTn. Overall concordance with RNASeq data was inferred from coverage of query sequence with the best BLAST hit from this analysis. Sequencing depth for each site in all exonic regions for both genome assemblies was calculated using Bedtools version 2.17.0. Both sets of gene predictions were also tested against the NCBI nonredundant (nr) database using BLASTp, excluding previous *S. sclerotiorum* gene models. Percent identity and query coverage of subject sequence based on best BLAST hits were used to infer quality of gene models. Finally, a random set of 30 sequences selected from the 300 loci that differed the most between the two genome versions were manually reinspected to check which version's prediction most closely matched supporting evidence and why there might have been a discrepancy. Protein Domain Prediction in the Version Two Annotations -------------------------------------------------------- Protein domains in amino acid sequences were predicted using Interpro Scan version 5.17-56.0 ([@evx030-B50]), which was also used to retrieve Gene Ontology (GO) ([@evx030-B5]) terms. GO terms were mapped to "GO-slim" terms using Goatools version 0.5.9 (available at <https://libraries.io/github/tanghair/goatools>). Effector Prediction in the Version Two Annotations -------------------------------------------------- Putative effector prediction was carried out using the following pipeline: 1) SignalP version 4.1 ([@evx030-B73]) was used to identify proteins with secretion signals, 2) TMHMM version 2.0 ([@evx030-B53]) was used to filter out those that contained transmembrane domains, 3) GPIsom (available at <http://gpi.unibe.ch/>) ([@evx030-B32]) was used to filter out proteins that harbored a putative glycophosphatidylinositol membrane-anchoring domain, and 4) EffectorP version 1.0 ([@evx030-B88]) was used to predict potential effector sequences among those remaining after the previous filtering steps. Analysis of Differential Expression of Putative Effectors --------------------------------------------------------- A subset of the RNASeq data, data set 1, that were used for gene prediction were aligned to the version two assembly using the previously described methods. This subset included samples taken from the following conditions: *In vitro*, 1, 3, 6, 12, 24, and 48 hpi of *B. napus* with *S. sclerotiorum* strain 1980. Samples were replicated three times as per previously mentioned. Aligned reads were then used in conjunction with the version two gene annotations to determine differential expression using the cuffdiff module of Cufflinks ([@evx030-B97]). All sequences that exhibited homology to ribosome-related sequences were specified to cuffdiff at run-time. Fold-change in expression was considered significantly different if it was at a *P*-adjusted value of below 0.05 indicating a false discovery rate of below 0.05. Identification of Putative Effector Families and Their Association with Segmental Duplications and TEs ------------------------------------------------------------------------------------------------------ To identify repetitive sequences in the version two *S. sclerotiorum* genome, the REPET pipeline ([@evx030-B74]) was run with default settings. *De novo* repeat annotations produced by REPET were used in all downstream analyses. TEs were divided into nine subclasses (automatically by REPET) based on the nomenclature system devised by ([@evx030-B104]). These subclasses included noCat, DXX, RIX, RXX, RLX DTX, RPX, XXX and RYX, corresponding to no category, unclassified DNA element, unclassified long interspersed nuclear element, unclassified retroelement, long terminal repeat (LTR) element, DNA terminal inverted repeat (TIR) element, Penelope retroelement, unclassified TE, and DIRS-like element, respectively. To identify paralogous effectors in *S. sclerotiorum*, OrthoMCL version 2.0.9 ([@evx030-B56]) was used with default settings. The *Botrytis cinerea* reference genome version 3.0 ([@evx030-B98]) was downloaded from the ENSEMBL Fungi database (<http://fungi.ensembl.org/Botrytis_cinerea/Info/Index>) for this purpose ([@evx030-B71]). For effector sequences that were determined to be recent paralogs by OrthoMCL, regions of 4--5 kb encompassing the gene models were extracted and aligned using ClustalW version 2.1 ([@evx030-B96]). Up to 20-kb regions surrounding paralogous genes were manually inspected for the presence of LTRs using LTRharvest ([@evx030-B28]), TEs predicted by REPET, and overlapping gene models containing typical retro-element Pfam domains (based on the previously mentioned InterProScan analysis). Consensus TE sequences were used to search the Dfam database version 2.0 ([@evx030-B102]) to identify homologous TE families. Assessment of RIP in the Most Abundant Repeats and at the Whole-Genome Scale in *S. sclerotiorum* ------------------------------------------------------------------------------------------------- Before identifying RIP in repeat families, the whole genomes of *S. sclerotiorum* and the fungal and oomycete species *Blumeria graminis* f. sp. *hordei* (herein referred to as *B. graminis*), *Leptosphaeria maculans*, and *Phytophthora infestans* were scanned to determine whether there was a bimodal GC content. This was done using OcculterCut version 1.1 with default settings. The additional genome sequences were used to provide a context for genome bimodality against which *S. sclerotiorum* was compared. These additional genomes were downloaded from GenBank. After identifying repeat families *de novo* using REPET, the five TE sequences with the highest copy number were subjected to analyses to identify RIP. For each repeat, only the 50 longest sequences were used. First, the most numerous repeat was subjected to analysis using RipCal version 1.0 ([@evx030-B45]) through the alignment method to show dominant forms of RIP. Second, the ApT/TpA RIP index was calculated for each repeat family member and a random set of sequences from the *S. sclerotiorum* genome of the same size and lengths. Student's *t*-tests were used to determine whether the repeat sequences had a significantly higher ApT/TpA index than an equivalent random set of sequences. Random sequences were produced and RIP indices were calculated using a custom Perl script available at: <https://github.com/markcharder/GeneralBioinformaticsScripts/blob/master/getRandom.pl>. Comparative Genomic Analysis of Association of Secreted and Effector Proteins with Repeat-Rich, Gene Sparse Genomic Regions and RIP-Affected Sequence ----------------------------------------------------------------------------------------------------------------------------------------------------- To identify secreted and effector-like proteins, the same secreted protein and effector discovery pipelines were run on the genomes of the three fungi *S. sclerotiorum, B. graminis* and *L. maculans*, and the oomycete *P. infestans*. The three additional species were chosen as their genomes have been well characterized and exhibit typical features described in the "two-speed" genome hypothesis ([@evx030-B79]; [@evx030-B44]; [@evx030-B4]). Genomes used in this analysis, apart from *S. sclerotiorum*, were downloaded from GenBank. Before predicting secreted proteins, all gene sequences were filtered based on homology to TEs through blasting against the RepBase database version 20.05 ([@evx030-B9]). Those sequences with more than 30% amino acid identity and alignment coverage and an *e*-value of \<1e^−10^ with subjects from RepBase were discarded. This was to ensure that TE genes were not included in downstream analysis, as this would affect CDS content and proportion of secreted proteins in repeat-rich regions. The secreted protein prediction pipeline followed the same procedure as detailed in the "Effector prediction in the version two annotations" above but used Pfam instead of GPI-som to predict GPI-anchoring domains. To identify a subset of the secreted proteins that were potential effectors, EffectorP was run. The whole pipeline for secreted protein discovery is available at: <https://github.com/markcharder/GeneralBioinformaticsScripts/blob/master/effector_pipeline.sh>. Repeats in all of these genomes were identified for the purpose of this analysis using RepeatMasker; simple sequence repeats were discarded. To identify potentially RIP-affected sequence, RipCal version 1.0 was run on whole genomes with default settings to identify regions with a high RIP index. To determine whether there was an association between the presence of secreted and effector-like proteins and repeats, RIP-affected sequence and gene-sparse sequence in the four genomes, two approaches were taken. First, for each set of secreted proteins, each set of effector-like proteins and each total gene set, the "closest" module of Bedtools was used to determine the distance from nearest repeats and RIP-affected sequence. Then, from the distances generated for the total set of genes, random sets the same size as the secreted protein sets and the effector-like sets were generated. The secreted protein sets and effector-like sets were compared against the random sets using Wilcoxon's test in R version 3.3.0 beta. The script used for this analysis is available at: <https://github.com/markcharder/GeneralBioinformaticsScripts/blob/master/testingAssociations.r>. Second, percentage of bases covered by CDSs, GC content, and number of genes and number of secreted proteins were determined across a sliding window of 100 kb for each genome, incrementing by 1 kb. The R scripts used for this analysis are available at: <Https://Github.Com/Markcharder/GeneralBioinformaticsScripts/Blob/Master/Test_Two-Speed.r>. Spearman's rank was used to assess correlation between proportion of secreted protein content (defined as number of secreted proteins/total number of genes) per sliding window with CDS and GC content. GC content was used to infer the likelihood of extensive RIP in sliding windows. Results ======= The *S. sclerotiorum* Version Two Assembly Is Near Complete, with a Single Gap in the Nucleolus Organizer Region ---------------------------------------------------------------------------------------------------------------- To produce a more complete and accurate genome for *S. sclerotiorum*, PacBio reads were assembled *de novo* and scaffolded using the previously generated optical map and a portion of the Sanger assembly ([@evx030-B2]). To further improve accuracy, Illumina reads that mapped to the new assembly were used to create a consensus sequence. The final assembly was based on 36× coverage of PacBio reads and consisted of 38,806,497 bp distributed across 16 scaffolds, representing the 16 chromosomes predicted for the *S. sclerotiorum* strain 1980 ([@evx030-B2]). This is an improvement in both sequence content---an additional 805,146 bp have been sequenced---and contiguity, as the version one assembly contained 38,001,451 bp (excluding Ns) distributed across 36 scaffolds that contained numerous gaps ([fig. 1](#evx030-F1){ref-type="fig"}; [supplementary fig. 1*a* and table 2](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). In the version two assembly, a single gap of unknown size was present within the nucleolus organizing region (identified through homology of repeating sequence to ribosomal DNA of other fungi; data not shown). This gap was given an arbitrary size of 100 Ns; this is contrasted to the 328,761 N-bases of the version one assembly ([supplementary table 3 and fig. 1*a*](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online; [fig. 1](#evx030-F1){ref-type="fig"}). The new version of the S. sclerotiorum genome and its annotations are deposited in GenBank under bioproject number PRJNA348385.F[ig]{.smallcaps}. 1.---Circos plot depicting features of the new *S. sclerotiorum* assembly and positions of corresponding version one contigs. From outer to inner circular tracks: The 16 version two assembly (Ss1980 version 2) contigs colored to distinguish neighboring contigs; the contigs from the version one assembly (Ss1980 version 1) colored to distinguish neighboring contigs; circles depicting positions of secreted proteins and SsPEs---clear circles = secreted proteins, colored circles = SsPEs; log fold change in expression of secreted proteins from *in vitro* to each of the *B. napus* time points tested, from blue (downregulated) to red (upregulated); regions with a high RIP index as predicted by RipCal; repeat sequences from REPET *de novo* annotation of more than 5 kb. The lines crossing the center of the Circos plot join sequences of the same repeat element. Mapping of Illumina reads to the version two assembly scaffolds revealed 653 insertions (In), 222 deletions (Del), and 8 substitutions ([supplementary table 3](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online); all 875 InDels were subsequently corrected. The version two assembly was congruent with the version one assembly ([fig. 1](#evx030-F1){ref-type="fig"}; [supplementary fig. 1](#sup1){ref-type="supplementary-material"}*b*, [Supplementary Material](#sup1){ref-type="supplementary-material"} online); although upon mapping Illumina reads to the version one assembly, it became apparent that there were more discrepancies between the Illumina reads and this assembly than the version two assembly, totaling 751 insertions, 1,828 deletions, and 1,764 substitutions ([supplementary table 3](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Gene Predictions in the *S. sclerotiorum* Version Two Genome Assembly Exhibit Significant Divergence from the Version One Predictions ------------------------------------------------------------------------------------------------------------------------------------- To determine how individual gene models had changed from version one to version two, a reciprocal best hits BLAST analysis in conjunction with Bedtools-based analyses were conducted. The Bedtools analyses were used to determine fusion and splitting events between the version two predictions and the transferred version one predictions. The BLAST analysis was used to identify version one loci corresponding to version two loci. The total number of gene models in the version two assembly is 11,130. This is 3,392 less than the previous total of 14,522, and 730 less than the previous "nondubious" total of 11,860. Furthermore, a total of 435 predictions in the version two assembly did not have a one-to-one BLAST correspondence with any version one loci. From version one to version two, 1,301 loci were fused to neighboring loci. A total of 279 version one loci were split to form separate loci in the version two assembly. The mean length of version two models was 1,439.88 bp, whereas the mean length of version one models was 1,088.47 bp. Based on the reciprocal BLAST analysis, 6,891 sequences were identical at the nucleotide level ([supplementary table 4](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Gene Predictions in the Version Two Genome Assembly Are More Accurate than Version One -------------------------------------------------------------------------------------- To assess the accuracy of the new gene models, both the version one predictions and the version two predictions were 1) tested against the NCBI nr database, 2) scanned for protein domains using InterProScan, 3) assessed for congruence with RNASeq data by comparison to Cufflinks transcripts and determining RNASeq depth for all CDS regions, and 4) manually reinspected randomly to determine which version's models were best supported by accompanying evidence. Based on the BLAST analysis against the NCBI nr database, the version two predictions exhibited a higher mean percent identity (86.7%) with their best BLAST hits than the version one predictions (81.78%). Furthermore, the version two sequences exhibited a higher mean subject coverage per alignment with their best BLAST hits (90.39%) than the version one sequences (84.9%) ([fig. 2*a*](#evx030-F2){ref-type="fig"}). The InterProScan analysis showed that the version two sequences contained more predicted protein domains than the version one sequences ([fig. 2*b*;](#evx030-F2){ref-type="fig"}[supplementary table 5](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online).F[ig]{.smallcaps}. 2.---BLAST and RNA sequencing validation of version two annotations. (*a*) Alignments of best BLASTp hits of gene annotations were higher quality for version two annotations than version one. On the *x* axis is the percent identity of the best BLAST hits and on the *y* axis is the coverage of the subject sequence by the query sequence. Blue contour lines represent kernel density. (*b*) More GO slim terms were present in version two gene predictions than version one. Each GO slim term identified is plotted as a separate colored block in the stacked histogram for each version's annotations. On the *y* axis is the number of GO slim numbers associated with each annotation set for each GO slim term. (*c*) Version two annotations were more concordant with RNASeq data, based on comparison to assembled Cufflinks transcripts. The *y* axis represents percent coverage of the gene annotation CDS with best BLAST hit. Blue lines represent median values, boxes and whiskers represent second and third quartiles, and interquartile range, respectively. (*d*) Version two gene annotations showed a greater depth of RNASeq coverage than version one annotations. Depth of RNASeq alignment across exonic regions is plotted on the *y* axis. Blue lines represent median values and boxes and whiskers represent second and third quartiles, and interquartile range, respectively. Comparing CDS sequences from both sets of gene predictions with assembled Cufflinks transcripts demonstrated that 10,523 version two sequences had RNASeq-based transcript support. A total of 7,506 of these were fully supported, that is, gene predictions were fully covered by an assembled Cufflinks transcript which was 100% identical. Furthermore, the version two sequences were more concordant with RNASeq data than the version one sequences. Mean query coverage per high scoring segment pair (HSP) for best BLAST hits was 91.88% for the version two sequences and 83.53% for the version one sequences, whereas median coverage per HSP was moderately increased to 100% in the version two sequences from 99% in the version one sequences. However, interquartile range for the version two sequences was 1, whereas interquartile range for the version one sequences was 32. Additionally, mean RNASeq coverage for all CDS regions of the version two predictions was 535.187×, whereas for the version one predictions it was 456.359×, median values were 132 and 124, respectively ([fig. 2*c* and *d*](#evx030-F2){ref-type="fig"}). Thirty of the 300 most divergent sequences from version one to two were randomly selected. Reinspection of these genes (after initial manual curation in a previous step) confirmed that the version two models were more in accordance with supporting evidence than version one models ([supplementary fig. 2](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). TE Content Is Increased in the *S. sclerotiorum* Version Two Genome Assembly ---------------------------------------------------------------------------- To assess the amount of repetitive sequence in the new *S. sclerotiorum* genome the REPET pipeline ([@evx030-B74]) was run. This demonstrated that the version one assembly contained 4,329,439 bp (11.39%) of repetitive sequence, whereas the version two assembly contained 5,041,697 bp (12.96%); this is an increase of 712,258 bp (1.57%). As a proportion of the total additional sequence content (905,046 bp), this is 78.7%; thus, most of the additional sequence was repetitive. Both class I (retro) elements and class II (DNA) elements were identified in the version two genome. The first and second-most abundant TEs were unclassified retroelements and DNA TIR elements, respectively. Overall proportions of TE subclasses did not markedly differ between the two genome versions, though the total number of bases within TEs did ([fig. 3](#evx030-F3){ref-type="fig"}).F[ig]{.smallcaps}. 3.---Transposable element content of the version one and version two assemblies. On the *y* axis is the number of bases covered by TEs in each family, represented by different colored blocks in the stacked histogram for both genome versions. The figure is based on *de novo* prediction by REPET. Seventy Putative Effector Sequences, 61 of Which Have Not Been Reported Before, Were Predicted for *S. sclerotiorum* -------------------------------------------------------------------------------------------------------------------- To determine whether the new genome contained previously undiscovered effector-like sequences, version two-predicted proteins were filtered based on various criteria, including identification using the EffectorP software ([@evx030-B88]). Based on these analyses, the version two *S. sclerotiorum* genome contains 900 predicted proteins with a predicted secretion signal; 695 of these lack a predicted transmembrane domain, of which 523 lack both a predicted transmembrane domain and a predicted GPI-anchoring domain. Of these 523 amino acid sequences, 70 were determined to be likely effectors based on EffectorP analysis; the mean amino acid length of these sequences was 176.543. These will henceforth be referred to as "*S. sclerotiorum* putative effectors" (SsPEs) ([fig. 1](#evx030-F1){ref-type="fig"}; [table 1](#evx030-T1){ref-type="table"}).Table 1**Predicted Effectors in the Version Two***Sclerotinia****sclerotiorum* Genome (SsPEs) and Their Corresponding Version One Loci**Version One IDVersion Two IDPercent IdentityVersion One LengthVersion Two LengthGuyon et al. Prediction?Version Two Length/Version One LengthDomains Predicted by InterProScan (prediction database\|domain identifier\|description\|)% AA Identity of Best Blast Hit*E*-Value of Best Blast HitSS1G_09693Sscle01g0006601004594591\|\|\|57.5342.34e^−050^SS1G_02068Sscle01g0038501005655651\|\|\|55.6212.7e^−056^SS1G_01974Sscle01g0046201004655581.2\|\|\|86.6075.13e^−052^SS1G_01867Sscle01g0053901004904901\|\|\|90.7413.46e^−059^SS1G_01754Sscle01g0063301002,2665320.234774934\|\|\|65.0359.79e^−055^SS1G_01427Sscle01g0089401001,5121,5121Pfam\|PF00024\|PAN domain\|69.4213.37e^−168^SS1G_01426Sscle01g0089501001,0791,0791\|\|\|82.1582.12e^−119^SS1G_01287Sscle01g0099601006246241\|\|\|29.3533.44e^−011^SS1G_01214Sscle01g0104901004044041Pfam;SUPERFAMILY;Gene3D\|PF06766;SSF101751; G3DSA:3.20.120.10\|Fungal hydrophobin;;\|63.2659.41e^−039^SS1G_12778Sscle02g0129401002822821\|\|\|------SS1G_04766Sscle02g0142801008813480.395005675Pfam;SUPERFAMILY;Gene3D\|PF00190;SSF51182; G3DSA:2.60.120.10\|Cupin;;\|84.4042.38e^−059^SS1G_04618Sscle02g0153901004684681\|\|\|72.6111.13e^−079^SS1G_04155Sscle02g0190001001,1391,1391\|\|\|77.5511.96e^−149^SS1G_13012Sscle02g0217801007558381.109933775ProSiteProfiles;Gene3D;SUPERFAMILY; Pfam\|PS50059;G3DSA:3.10.50.40;SSF54534; PF00254\|FKBP-type peptidyl-prolyl *cis-trans* isomerase domain profile.;;;FKBP-type peptidyl-prolyl *cis-trans* isomerase\|90.9093.64e^−112^SS1G_01032Sscle03g0225501001,1781,1781\|\|\|79.8940SS1G_01003Sscle03g0227901004504501Gene3D;Pfam\|G3DSA:3.20.120.10;PF06766\|;Fungal hydrophobin\|34.6531.63e^−012^SS1G_13371Sscle03g031910100880880x1\|\|\|61.4047.12e^−083^SS1G_02522Sscle04g0351601008978971\|\|\|87.95.62e^−169^SS1G_02700Sscle04g0365501002766102.210144928\|\|\|70.277.9e^−081^SS1G_03057Sscle04g0392101004844841Pfam;Gene3D\|PF03966;G3DSA:2.20.25.10\| Trm112p-like protein;\|96.858.66e^−083^SS1G_03080Sscle04g039420100852852x1Pfam;PIRSF;Coils\|PF05630;PIRSF029958;Coil\|Necrosis-inducing protein (NPP1);;\|82.1145.47e^−147^SS1G_14237Sscle04g0400801001,1861,1861Pfam\|PF11327\|Protein of unknown function (DUF3129)\|82.49e^−144^SS1G_12123Sscle05g0410501001,3656200.454212454\|\|\|69.5654.06e^−075^SS1G_06068Sscle05g0450601003523521ProSiteProfiles\|PS51257\|Prokaryotic membrane lipoprotein lipid attachment site profile.\|67.7422.5e^−035^SS1G_05939Sscle05g0460601002642641ProSiteProfiles\|PS51257\|Prokaryotic membrane lipoprotein lipid attachment site profile.\|74.7134.11e^−041^SS1G_05938Sscle05g046070100639639x1\|\|\|58.7721.7e^−045^SS1G_07491Sscle06g0489201007095260.741889986\|\|\|75.515.86e^−078^SS1G_07320Sscle06g0501001009309301SUPERFAMILY;TIGRFAM;PRINTS;PRINTS;PRINTS;PRINTS; PRINTS;Gene3D;SMART;ProSiteProfiles;Pfam; SMART\|SSF52540;TIGR00231;PR00328;PR00328; PR00328;PR00328;PR00328;G3DSA:3.40.50.300;SM00178; PS51422;PF00025;SM00177\|;small_GTP: small GTP-binding protein domain;GTP-binding SAR1 protein signature;GTP-binding SAR1 protein signature; GTP-binding SAR1 protein signature;GTP-binding SAR1 protein signature;GTP-binding SAR1 protein signature;;Sar1p-like members of the Ras-family of small GTPases;small GTPase SAR1 family profile.;ADP-ribosylation factor family;ARF-like small GTPases; ARF, ADP-ribosylation factor\|99.4718.88e^−135^SS1G_07230Sscle06g0508201002982981\|\|\|36.3642.1SS1G_07027Sscle06g0523601007107101\|\|\|87.9311.17e^−109^SS1G_12482Sscle06g0544001007977971\|\|\|73.821.5e^−114^SS1G_12431Sscle06g0548101009396160.656017039\|\|\|45.373.47e^−022^SS1G_12365Sscle06g055280100843843+1\|\|\|67.6061.84e^−096^SS1G_03381Sscle07g0570001001,8412,1241.153720804\|\|\|61.0291.69e^−043^SS1G_11673Sscle07g0617701005065061SUPERFAMILY;ProSiteProfiles;Gene3D\|SSF57414;PS50948; G3DSA:3.50.4.10\|;PAN/Apple domain profile.;\|65.6253.07e^−057^SS1G_11693Sscle07g061960100850850+1\|\|\|67.6061.84e^−096^SS1G_11706Sscle07g0620601002976142.067340067\|\|\|82.6394.75e^−082^SS1G_05103Sscle08g0641801005205201Coils\|Coil‖66.991.51e^−042^SS1G_05152Sscle08g0645901005295291\|\|\|74.0512.25e^−055^SS1G_05569Sscle08g0677101004984981\|\|\|46.9144e^−035^SS1G_14184Sscle08g0682001001,5601,350x0.865384615ProSitePatterns;SMART;ProSiteProfiles;ProSitePatterns; SUPERFAMILY;SUPERFAMILY;ProSitePatterns;ProSiteProfiles; SUPERFAMILY;ProSiteProfiles;Gene3D;Gene3D;SMART; SMART;SMART;Pfam;Gene3D;Gene3D;ProSitePatterns; ProDom;Pfam;SUPERFAMILY\|PS00026;SM00236; PS51164;PS00026;SSF57016;SSF57016;PS00562;PS50941; SSF57016;PS50941;G3DSA:3.30.60.10;G3DSA:3.30.60.10; SM00270;SM00270;SM00270;PF00187;G3DSA:3.30.60.10; G3DSA:3.30.60.10;PS00026;PD001821;PF00734; SSF57180\|Chitin recognition or binding domain signature.;Fungal-type cellulose-binding domain;CBM1 (carbohydrate binding type-1) domain profile.;Chitin recognition or binding domain signature.;;;CBM1 (carbohydrate binding type-1) domain signature.;Chitin-binding type-1 domain profile.;;Chitin-binding type-1 domain profile.;;;Chitin binding domain;Chitin binding domain;Chitin binding domain;Chitin recognition protein;;;Chitin recognition or binding domain signature.;DEGRADATION HYDROLASE GLYCOSIDASE CELLULOSE METABOLISM CARBOHYDRATE POLYSACCHARIDE PRECURSOR SIGNAL I;Fungal cellulose binding domain;\|34.7594.64e^−014^SS1G_10892Sscle09g0690901005765761Pfam\|PF12296\|Hydrophobic surface binding protein A\|38.151.68e^−028^SS1G_03897Sscle09g0726301004814811\|\|\|64.3849.4e^−025^SS1G_03721Sscle09g074030100869869+1\|\|\|61.4043.39e^−084^SS1G_08128Sscle10g0749201004444441\|\|\|83.813.41e^−047^SS1G_08163Sscle10g07514099.493933951.005089059\|\|\|66.6671.75e^−033^---Sscle10g076600------------Pfam;SUPERFAMILY\|PF07249;SSF50685\|Cerato-platanin;\|87.4173.34e^−085^SS1G_13851Sscle10g0805801004412750.623582766\|\|\|49.2314.2e^−010^SS1G_08088Sscle11g0810201001591591\|\|\|------SS1G_07837Sscle11g0829701001,0121,0121\|\|\|50.9682.64e^−051^SS1G_07613Sscle11g0847201006286281Pfam;Gene3D;SUPERFAMILY;SMART\|PF02221; G3DSA:2.70.220.10;SSF81296;SM00737\|ML domain;;;Domain involved in innate immunity and lipid metabolism.\|86.6283.87e^−107^SS1G_11151Sscle12g0879601004714711\|\|\|33.7585.21e^−017^SS1G_11085Sscle12g0885301007727721\|\|\|47.5731.37e^−055^SS1G_11065Sscle12g0886601006366361\|\|\|40.3853.4SS1G_11928Sscle12g0903801001591591\|\|\|---------Sscle13g094760------------Pfam\|PF10270\|Membrane magnesium transporter\|89.0411.21e^−090^SS1G_06729Sscle13g0949201008028021Pfam;ProSiteProfiles;Coils;SUPERFAMILY\|PF01105;PS50866; Coil;SSF101576\|emp24/gp25L/p24 family/GOLD;GOLD domain profile.;;\|96.5525.94e^−145^SS1G_06763Sscle13g0952301007097091\|\|\|44.6244.16e^−039^SS1G_14379Sscle13g097000100843843+1\|\|\|68.0753.96e^−097^SS1G_08669Sscle14g0987101002131,0554.953051643Gene3D;SUPERFAMILY;Pfam\|G3DSA:1.10.1280.10;SSF48056; PF00264\|;;Common central domain of tyrosinase\|68.7057.45e^−099^SS1G_08706Sscle14g0989201004054051SUPERFAMILY;Gene3D;Pfam\|SSF50685;G3DSA:2.40.40.10; PF03330\|;;Rare lipoprotein A (RlpA)-like double-psi beta-barrel\|88.8066.08e^−081^SS1G_08892Sscle14g1003101009089081\|\|\|65.8458.4e^−123^SS1G_13394Sscle15g1023901005875871\|\|\|58.7572.72e^−059^SS1G_09420Sscle15g1051601005285281\|\|\|77.5861.6e^−095^SS1G_09288Sscle15g1060801005945941\|\|\|33.3332.32e^−019^SS1G_09150Sscle15g1071901006486481\|\|\|36.7820.001SS1G_10104Sscle16g107730100880880+1\|\|\|61.8423.08e^−084^SS1G_10129Sscle16g1078901003333331\|\|\|46.0320.026SS1G_14320Sscle16g1110801004834831Pfam;ProSiteProfiles;Gene3D;SMART;SUPERFAMILY\|PF01817; PS51168;G3DSA:1.20.59.10;SM00830;SSF48600\|Chorismate mutase type II;Chorismate mutase domain profile.;;Chorismate mutase type II;\|58.9044.27e^−048^---Sscle16g111300------------\|\|\|68.0753.96e^−097^[^3] Of the 70 SsPEs, 13 are different in length to their corresponding loci in the version one assembly. The most divergent sequence was 4.95 times longer than the previous model ([table 1](#evx030-T1){ref-type="table"}). An additional three did not correspond to genes predicted in the version one assembly. The remaining 54 SsPEs were identical to their corresponding loci in the version one assembly. Of the 70 SsPE sequences, 61 were not identified as likely effectors by [@evx030-B42]) who used a different effector-prediction pipeline that did not include EffectorP analysis. Of the 70 SsPEs, 63 had homologs (*e*-value \< e^−10^) in other species of fungi, 22 of which contained predicted protein domains based on an InterProScan analysis. A total of 48 SsPEs had no predicted protein domains, seven of which were unique to *S. sclerotiorum* ([table 1](#evx030-T1){ref-type="table"}). RNASeq Analysis Demonstrates Significant Differential Expression of SsPEs during Infection of *B. napus* Relative to during Growth *In Vitro* --------------------------------------------------------------------------------------------------------------------------------------------- To determine whether SsPEs were significantly upregulated *in planta* relative to during growth *in vitro*, an RNASeq time course consisting of samples encompassing *in vitro* growth, 1, 3, 6, 12, 24, and 48 hpi of *B. napus* was analyzed. In total, 66 SsPEs were expressed under at least one condition tested ([figs. 1 and 4*a*](#evx030-F1 evx030-F4){ref-type="fig"}). Of these 66, 19 were significantly differentially expressed at at least one *in planta* time point relative to during growth *in vitro*. Of these 19 gene models, nine were significantly upregulated during infection relative to during growth *in vitro* at at least one time point and exhibited a log fold change of at least 1 ([fig. 3*b* and *c*](#evx030-F3){ref-type="fig"}). Of these gene models, five were significantly upregulated at 1, 3, 6, and/or 12 hpi relative to during growth *in vitro*; three of which were also upregulated at 24 and/or 48 hpi ([fig. 3*b* and *c*](#evx030-F3){ref-type="fig"}). A further four gene models were exclusively upregulated at later time points, 24 and/or 48 hpi ([fig. 3*c*](#evx030-F3){ref-type="fig"}).F[ig]{.smallcaps}. 4.---Expression analysis of SsPEs. (*a*) Heatmap showing expression (log~10~(FPKM + 1)) of all expressed SsPEs *in vitro* (IV) and at each *B. napus* time point tested (X1h, X3h, X6h, X12h, X24h, X48h); from light to dark amber represent low to high expression (0 to 3+). (*b*) Expression of SsPEs significantly upregulated *in planta* at early time points (1, 3, 6, or 12 hpi) relative to during growth *in vitro*; asterisks represent a *P*-adjusted value of \< 0.05 based on cuffdiff analysis. The *y* axis scale is FPKM and the *x* axis is all conditions tested from *in vitro* (left) to 48 hpi (right). (*c*) Same as (*b*) but for SsPEs that were significantly upregulated at later time points (24 or 48 hpi) relative to during growth *in vitro.* There are some overlapping SsPEs between (*b*) and (*c*). Seven Putative *S. sclerotiorum* Effectors Are Paralogous and Associated with Potentially Recent Duplication Events ------------------------------------------------------------------------------------------------------------------- OrthoMCL was used to determine which SsPEs were likely to be recent paralogs (i.e., generated after the divergence of *B. cinerea* and *S. sclerotiorum*). This identified seven SsPEs grouped into a single family, with a single homolog in *B. cinerea*. None of these sequences contained predicted functional domains based on InterProScan analysis, though they were broadly conserved throughout fungi, matching numerous models without functional domain predictions ([supplementary table 6](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Six of these proteins were previously identified by Guyon et al. ([@evx030-B42]), who demonstrated a close proximity (1.1 kb) to a Tad1 retroelement of one of them. To expand on these analyses, and to determine whether these sequences were embedded within regions that might have resulted from relatively recent segmental duplications, 4--5 kb regions spanning SsPEs were extracted and aligned; this created an alignment of 4,869 bp. These regions were homologous and exhibited a higher overall percentage identity across the multiple alignment in the region immediately surrounding and spanning the SsPEs. Four of the SsPE-containing loci (those containing SsPE models Sscle13g097000, Sscle16g111300, Sscle06g055280, and Sscle07g061960) were more than 90% identical across an approximately 3,000-bp region immediately surrounding and spanning the aligned SsPEs. Additionally, similarly sized regions surrounding SsPE models Sscle09g074030 and Sscle16g107730 exhibited a pairwise identity of more than 90% but were divergent from the previously mentioned SsPE regions ([fig. 5*a*](#evx030-F5){ref-type="fig"}).F[ig]{.smallcaps}. 5.---A seven member SsPE family associated with transposition of a retro element. (*a*) Alignment of the seven paralogous SsPEs (six of which were identified by Guyon et al.), including flanking genomic regions. Coloring from white (0%) to blue (100%) represents identity across the multiple alignment. The gray arrow above represents the position of the SsPE loci in relation to the alignment of the flanking regions. (*b*) The genomic context of each of the seven SsPEs. Black gene diagrams represent SsPEs, gray bars represent fragments or the complete *Gypsy* retro element associated with this family, green bars represent other adjacent repeats, and blue gene diagrams represent other gene models. (*c*) Genomic context of two SsPEs flanked by corresponding LTR sequences and target site duplications. Red bars represent LTRs, black gene diagrams represent positions of SsPEs, magenta bars represent Pfam domains of genes in this region, blue gene diagrams represent other genes. (*d*) Expression of the seven member SsPE family *in vitro* and across the *B. napus* time points tested. Expression (FPKM + 1) is plotted on the *y* axis. Bars represent standard deviation. Different colored bars represent different sample conditions for each SsPE. To determine whether homologous SsPE-containing loci may have arisen through activity of TEs, genomic regions containing these seven SsPEs were inspected for the presence of TE sequences predicted by REPET and assessed for the presence of LTRs and transposition-associated Pfam domains. This showed that all SsPEs were embedded in either a complete or fragmented predicted chimeric LTR retroelement ([fig. 5*b*](#evx030-F5){ref-type="fig"}). Blasting of the consensus retroelement to the Dfam database showed that it belongs to the *Gypsy* family of TEs. The model Sscle09g074030 was embedded within the complete retroelement, which contained four upstream open reading frames (ORFs) (within the gene models Sscle09g073990, Sscle09g074000, Sscle09g07410 and Sscle09g07420) that harbored unknown function, zinc knuckle, reverse transcriptase, integrase core, and chromatin organization modifier domains.Additionally, a 5′-LTR was identified 269 bp upstream of the first ORF (Sscle09g073990) and a 3′-LTR was identified 2,170 bp downstream of the SsPE. The SsPE model Sscle07g061960 was situated between 5′- and 3′-LTRs but the region lacked ORFs with predicted transposition-associated domains. The 5′-LTR was situated 5,376 bp from the 5′-end of the SsPE and the 3′-LTR was situated 238 bp from the 3′-end of the SsPE ([fig. 5*c*](#evx030-F5){ref-type="fig"}). The 5′- and 3′-LTRs surrounding these two SsPEs were adjacent to target site duplications. Other SsPE models were not situated between 5′- and 3′-LTRs but were flanked by predicted retroelements and DNA TEs containing LTRs and inverted terminal repeats, respectively (data not shown). To determine whether members of this SsPE family were active during infection, data from the previously mentioned RNASeq analysis were analyzed for evidence of expression. This showed that six of the seven paralogous SsPEs were expressed at at least one time point during infection of *B. napus*. A single paralogous SsPE was not expressed under any conditions, including both *in planta* and *in vitro.* The most highly and consistently (across all conditions tested) expressed SsPE was the model Sscle07g061960. None of the paralogous SsPEs were significantly differentially expressed between conditions. Overall, expression levels of these gene models were low, with mean FPKM values of below 10 ([figs. 1, 4*a*, and 5*d*](#evx030-F1 evx030-F4 evx030-F5){ref-type="fig"}). Identification of RIP in the Five Most Frequently Occurring Repeats in the *S. sclerotiorum* Genome --------------------------------------------------------------------------------------------------- To determine whether *S. sclerotiorum* exhibits bimodal GC content, OcculterCut was run on the version two genome sequence and the genome sequences of *L. maculans, B. graminis*, and *P. infestans*. The only genome with a bimodal GC content was *L. maculans*, which has been shown previously to harbor alternating RIP-affected and non-RIP-affected regions ([@evx030-B79]). The genomes of the other three species tested, including *S. sclerotiorum*, exhibited unimodal GC content ([fig. 6*a*](#evx030-F6){ref-type="fig"}).F[ig]{.smallcaps}. 6.---Analysis of RIP in the version two *S. sclerotiorum* genome. (*a*) Plot from OcculterCut analysis showing GC content for *S. sclerotiorum, P. infestans, L. maculans*, and *B. graminis* f. sp. *Hordei.* The *x* axis scale is GC content (%) and the *y* axis scale is proportion of the genome that shows a GC content of *x* percentage. (*b*) RipCal alignment based analysis of the most abundant repeat in the *S. sclerotiorum* genome "RXX-chim_Blc59_repet-L-B64-Map1_reversed." The colored bars at the top of the figure represent an alignment of the 50 longest copies of this repeat. Red bars represent likely CpA  = \> TpA mutation, blue bars represent likely CpC  = \> TpC mutation, green bars represent likely CpG  = \> TpG mutation, and turquoise bars represent CpT  = \> TpT mutation; gray bars, black bars and white bars represent mismatches, matches, and gaps relative to the consensus sequence, respectively. Graph below shows total number of mutations for each potential type of RIP (i.e., CpA  = \> TpA, CpC  = \> TpC, CpG  = \> TpG, CpT  = \> TpT) at each site in the alignment. (*c*) Comparison of the TpA/ApT index for the five most numerous repeats in the version two *S. sclerotiorum* genome against a random set of equivalent sequences from the same genome (same number and sizes). Only the 50 longest sequences were considered for each repeat family. The *y* axis scale represents the TpA/ApT index for each set of sequences. Black bars represent median values and boxes and whiskers represent second and third quartiles, and interquartile range, respectively. Asterisks represent statistical significance (\*\*\**P* \< 0.001; Student's *t*-test). To determine whether particular repeats in *S. sclerotiorum* were affected by RIP, regardless of whether they were compartmentalized into particular low-GC content regions or not, the five highest copy number repeats in the *S. sclerotiorum* genome were subjected to RIP-based analyses. First, the 50 longest copies of the most abundant repeat, flagged as "RXX-chim_Blc59_repet-L-B64-Map1_reversed" by REPET, were subjected to an alignment-based RipCal version 1.0 ([@evx030-B45]) analysis to identify dominant forms of RIP. This showed that these 50 repeat elements are likely to have undergone both CpA  = \> TpA and CpT  = \> TpT RIPs ([fig. 6*b*](#evx030-F6){ref-type="fig"}). Second, the RIP index TpA/ApT was calculated for the 50 longest sequences of all five repeats and compared against a random sequence set of equivalent size and number from the *S. sclerotiorum* genome. This showed that for each repeat family, the TpA/ApT index was significantly higher than found in a random set of sequences (Student's *t*: *P* \<0.001) ([fig. 6*c*](#evx030-F6){ref-type="fig"}). Comparison of the *S. sclerotiorum* Genome Sequence with Genome Sequences of Three Filamentous Fungal Pathogens Known to Harbor Two-Speed Genome Properties ----------------------------------------------------------------------------------------------------------------------------------------------------------- Based on our observations of apparent TE-mediated expansion of a seven-member effector family in *S. sclerotiorum* and potential RIP activity, we speculated that effectors within *S. sclerotiorum* could be associated with rapidly evolving, repeat-rich and potentially RIP-affected genomic regions. To test whether both secreted proteins and effector-like proteins in *S. sclerotiorum* were significantly associated with RIP-affected sequence, repeats and low gene content, we performed two different analyses. Both of these analyses were also carried out on the two plant pathogenic fungi *B. graminis* and *L. maculans*, and the oomycete *P. infestans.* The first analysis compared positions of secreted proteins and effector-like proteins with random sets of gene sequences. This analysis showed that in *S. sclerotiorum* there was no significant difference between the distance of secreted or effector proteins from nearest repeat elements than random sets. This was different to all three additional genomes tested. In *L. maculans*, secreted proteins were significantly closer to repeats than a random set (Wilcoxon's test: *P* \< 0.05), though there was no significant difference between effector proteins and the random set. However, adjusting α to 0.1 indicated a potential association between effector proteins and repeat regions in *L. maculans* (Wilcoxon's test: *P* \< 0.1). In both *B. graminis* and *P. infestans*, both secreted and effector-like proteins were significantly closer to repeats than random sets of proteins (Wilcoxon's test: *P* \< 0.001 for *B. graminis* and *P* \< 0.01 for *P. infestans*) ([fig. 7*a*](#evx030-F7){ref-type="fig"}). In *S. sclerotiorum* secreted proteins were significantly closer than the random set to regions with a high RIP index (Wilcoxon's test: *P* \< 0.01), whereas effector proteins were not. There was some indication that effector proteins in *S. sclerotiorum* may be further away from repeats if α was set to 0.1 (Wilcoxon's test: *P* \<0.1) ([fig. 7*b*](#evx030-F7){ref-type="fig"}).F[ig]{.smallcaps}. 7.---Association of repeat secreted protein and effector-like protein encoding genes with repeat regions and regions with a high RIP index. (*a*) Violin plots showing distances of secreted and effector-like protein encoding genes compared with random gene sets from each of the genomes tested. Violin plots represent kernel density of the data points. Asterisks and "x"s represent different *P* values---x = *P* \< 0.1; \* = *P* \< 0.05; \*\* = *P* \< 0.01; and \*\*\* = *P* \< 0.001 (Wilcoxon's test). The *y* axis scale is distance in bp from the nearest repeat element. Each type of sequence (from left to right: randomized set equivalent in size to secreted protein encoding gene set; secreted protein encoding genes; randomized set equivalent in size to effector-like gene set; effector-like gene set) is colored differently for clarity. Organism names are displayed above plots. (*b*) Jitter plots showing the same as (*a*) but for regions with a high RIP index (as predicted by RipCal). Points are staggered horizontally for each gene set so that differences, as statistically evaluated, are easier to interpret. In *B. graminis*, both secreted and effector-like proteins were significantly further away from regions with a high RIP index than the random sets (Wilcoxon's test: *P* \<0.001). In *L. maculans*, both secreted proteins and effector-like proteins were closer to regions with a high RIP index than the random sets (Wilcoxon's test: *P* \< 0.05). In *P. infestans*, no significant differences were observed for either secreted or effector proteins ([fig. 7*b*](#evx030-F7){ref-type="fig"}). The second analysis tested correlation between proportion of genes encoding secreted proteins and proportion of CDS sequence and GC content in a 100-kb sliding window, incrementing by 100 kb (end-to-end). This showed that in *S. sclerotiorum* there was no correlation between proportion of secreted proteins and CDS content or GC content. In *L. maculans*, there was a significant negative correlation between proportion of secreted proteins and CDS content (Spearman's test: rho = −0.31; *P* \< 0.001), and a significant negative correlation between proportion of secreted proteins and GC content (Spearman's test: rho = −0.36; *P* \< 0.001). In *B. graminis*, there was a significant negative correlation between proportion of secreted proteins and CDS content (Spearman's test: rho = −0.80; *P* \< 0.001), and no correlation between secreted protein proportion and GC content. In *P. infestans*, there was a significant negative correlation between secreted protein proportion and CDS content (Spearman's test: rho = −0.72; *P* \< 0.001), and no correlation between secreted protein proportion and GC content ([fig. 8](#evx030-F8){ref-type="fig"}).F[ig]{.smallcaps}. 8.---Correlation of GC content and CDS content with secreted protein proportion across a 100-kb end-to-end sliding window for all species tested. (*a*) The *x* axis scale is proportion of secreted protein encoding genes (defined as secreted/nonsecreted) for each sliding window, and the *y* axis scale is GC content. Both axes are plotted on a log scale. Blue lines represent least squares regression. (*b*) Same as for (*a*) but for CDS content instead of GC content . Discussion ========== A Complete Assembly of the *S. sclerotiorum* Genome --------------------------------------------------- In this article, we present a complete and accurately annotated genome of the destructive plant pathogenic fungus *S. sclerotiorum*. The previous genome assembly was based on Sanger sequencing at a depth of 9.1× combined with optical mapping. Though this assembly was relatively good quality, it was estimated to have been missing approximately 1.6 Mb of sequence. The new assembly of *S. sclerotiorum* contains an additional 805,146 bp. Though this is only half the estimated missing sequence, we conclude that the new *S. sclerotiorum* genome is practically complete, because 14 of the 16 chromosomes have been sequenced from telomere to telomere with virtually no gaps ([fig. 1](#evx030-F1){ref-type="fig"}; [supplementary fig. 1](#sup1){ref-type="supplementary-material"}*a* and [table 2](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). The only gap present is a region of the nucleolar organizing complex (containing rDNA repeats). A similar result was found in the recently published PacBio-sequenced genome of the plant-pathogenic fungus *V. dahliae.* In this study, which resulted in what was determined to be a finished genome, the authors noted read-stacking in the rDNA repeat region ([@evx030-B30]). The only other two complete fungal plant pathogen genomes (apart from *V. dahliae*) are those of the wheat head blight fungus *Fusarium graminearum* ([@evx030-B51]), and the broad host-range necrotroph *B. cinerea* ([@evx030-B98]); the first of these was assembled with Illumina mate pair technology and the second was assembled using PacBio sequencing and optical mapping. Thus, the new *S. sclerotiorum* assembly represents an addition to a thus far relatively small pool of complete genomes of fungal plant pathogens, and should be of use in future comparative genomics studies. Most of the additional bases assembled (78.7%) fall within what were predicted as repetitive regions. The total proportion of the genome predicted to be composed of TEs in this study is 12.96%, which is higher than the previously predicted 7.7--9.5% ([@evx030-B2]; Amselem, Lebrun, et al. 2015). However, as TEs were not manually curated following automated detection, several of these sequences could, in the future, be marked as dubious. Though they have not been thoroughly investigated in this study, more complete analysis of TEs in *S. sclerotiorum*, expanding on findings by [@evx030-B82] and Amselem, Lebrun, et al. (2015) may now be possible with the additional repetitive sequence. A More Accurate Set of Gene Annotations in the New *S. sclerotiorum* Genome --------------------------------------------------------------------------- Though moderate improvements were made in the genome assembly of *S. sclerotiorum*, the main improvements were in new gene annotations. In the previous genome assembly, 14,522 genes were predicted using *ab initio* gene finders trained on a manually curated gene set of 542 *S. sclerotiorum* genes. These 14,522 gene predictions were further evaluated based on predictions from additional softwares, EST and BLAST evidence, homology to TEs, and length. This resulted in a total of 11,860 nondubious gene calls ([@evx030-B2]). In the new gene annotation set of *S. sclerotiorum*, even when considering only the 11,860 "nondubious" genes of the previous genome version, there are still 730 fewer, as the new version contains 11,130 predictions. Furthermore, only 10,528 of these sequences had reciprocal best BLAST hits with version one sequences ([supplementary table 4](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online), which suggests that a further 602 sequences were either not present in the previous genome or had changed substantially enough in the version two prediction set for them to be unable to retrieve their corresponding loci as hits through BLAST analysis. Further inspection revealed that 1,301 version two loci resulted from fusion of previous loci. This is in contrast to the 279 version two gene predictions that resulted from splitting of version one predictions. Thus, it can be surmised that the decrease in filtered gene predictions from the version one assembly to the version two is largely a result of the joining of previous separate loci. Further analyses illustrated that the version two gene predictions were more accurate than the version one predictions. Blasting of amino acid sequences against the NCBI nr database indicated that version two sequences on average covered more of and had a higher identity to their best BLAST hits. This implies that the new gene predictions were able to obtain more BLAST hits from accurate protein predictions that represent truly conserved sequences ([fig. 2*a*](#evx030-F2){ref-type="fig"}). A similar analysis was performed to compare the recent *Parastagonospora nodorum* annotation set with a previous version ([@evx030-B91]). Accuracy at the protein level can also be inferred by the increased number of predicted functional domains and GO terms ([fig. 2*b*;](#evx030-F2){ref-type="fig"}[supplementary table 5](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Additionally, version two predictions were more similar to their corresponding Cufflinks transcripts based on BLAST analysis. Though median coverage was only improved from 99% to 100% in the version two predictions relative to version one, a high interquartile range of 32% for the version one sequences, as opposed to 1% for the version two sequences, would indicate a decreased degree of variability in correspondence with Cufflinks transcripts in the version two predictions ([fig. 2*b*](#evx030-F2){ref-type="fig"}). That the version two predictions are more supported by the RNASeq data is also evident in the observation that the median coverage of version two CDSs is higher than those of version one ([fig. 2*c*](#evx030-F2){ref-type="fig"}). Prediction of Effector Sequences Using an Updated Approach and the New, Improved Gene Models -------------------------------------------------------------------------------------------- A major theme of plant pathological research in recent decades has been the identification and analysis of so-called "effector" proteins. These are small, secreted proteins produced by pathogenic fungi whose function is to manipulate host physiology to promote infection ([@evx030-B60]). To date, several effector-like sequences have been functionally characterized in *S. sclerotiorum* ([@evx030-B105]; [@evx030-B106]; [@evx030-B100]; [@evx030-B63]; [@evx030-B21]). Two *in silico* analyses using the previous genome version attempted to define the *S. sclerotiorum* secretome and in one of these 79 effector-candidates were predicted ([@evx030-B47]; [@evx030-B42]). In this study, a new list of 70 effector candidates was identified that was markedly different from the previous list. Only nine genes in the new effector-candidate list were identified by Guyon et al. which highlights the differences not only between the annotation sets but also between the outcomes of effector-prediction pipelines that use different criteria ([fig. 1](#evx030-F1){ref-type="fig"}; [table 1](#evx030-T1){ref-type="table"}). Of the sequences in the new set, only 22 had predicted functional domains. Four of these functional domain predictions have been associated with effector-like activity in other fungi. A cerato-platanin domain was identified in the SsPE Sscle10g076600. Although the exact role of cerato-platanins remains elusive, it has been proposed that they may be involved in causing plant cell-wall instability or could possibly act as pathogen-associated molecular patterns ([@evx030-B7]). Indeed, a cerato-platanin protein deleted in *B. cinerea* was shown to be essential for full virulence ([@evx030-B35]). This protein was also shown to induce a hypersensitive response in tobacco and *A. thaliana* leaves and induce systemic acquired resistance in tobacco ([@evx030-B35]; [@evx030-B34]). The low level of expression of this gene *in vitro* and throughout infection ([fig. 4*a*](#evx030-F4){ref-type="fig"}), however, would indicate that abundance of its protein product during *S. sclerotiorum* infection is not necessary for full virulence on *B. napus*. However, further wet-lab studies are needed to test this hypothesis, perhaps including deletion experiments and heterologous expression or infiltration of the protein in *B. napus*. A necrosis-inducing protein (NPP) domain was identified in the effector-candidate Sscle04g039420. The previous locus corresponding to this gene model was already characterized as being only weakly expressed but nonetheless able to cause necrosis when infiltrated into tobacco leaves ([@evx030-B21]). In our study, expression was not detected *in vitro* or at any time points during infection ([fig. 4*a*](#evx030-F4){ref-type="fig"}), supporting the observation that this gene is only weakly expressed. A chitin-binding domain was identified in the gene model Sscle08g068200. The previous locus corresponding to this model was also identified by Guyon et al. Expanding on this discovery, our RNASeq analysis demonstrated that this gene was weakly expressed *in vitro* and throughout infection ([fig. 4*a*](#evx030-F4){ref-type="fig"}). Finally, a chorismate mutase domain was identified in the effector candidate Sscle16g111080. A chorismate mutase with an effector-like function was first identified in the maize pathogen *Ustilago maydis*, where it was shown to be secreted into host cells where it catalyses conversion of chorismate to prephenate. This reaction diverts chorismate away from the salycilic acid biosynthesis pathway, leading to a dampened immune response in infected plants. This dampening occurs because SA is a key signaling molecule in plant defense ([@evx030-B25]). RNASeq analysis showed that Sscle16g111080 was expressed *in vitro* and throughout infection. This would indicate that the gene is active and may also have an important function in *S. sclerotiorum* infection of *B. napus*. SsPEs that lacked predicted functional domains were generally homologous to sequences from other fungi (mean amino acid identity of best hit = 66.425%) ([supplementary table 6](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online); however, seven were not. These seven sequences were all expressed *in planta*. Intriguingly, one of these genes, Sscle02g012940, was significantly upregulated in *B. napus* from 12 to 48 hpi relative to during growth *in vitro* ([fig. 4*b* and *c*](#evx030-F4){ref-type="fig"}). At these time points during *S. sclerotiorum* infection, the plant becomes increasingly necrotized. It is possible that this gene model encodes a necrosis-inducing effector in *S. sclerotiorum*. The fact that this protein was not conserved in any other sequenced fungus may indicate that it is under positive selection pressure, and has become specialized to hosts of *S. sclerotiorum*. To determine this, further studies involving resequencing of *S. sclerotiorum* isolates from diverse regions and hosts, infiltration of this protein onto various host plants and cultivars, and deletion or knockdown experiments could be performed. An alternative hypothesis is that this gene sequence has no homologs simply because no fungal species with a homologous gene sequence has yet been sequenced. As the repertoire of fungal genomes increases, this may be elucidated. Another putative nonconserved gene model that was highly expressed was Sscle06g050820, though differentially expressed *in planta* relative to during growth *in vitro*, this gene was downregulated. However, as expression levels were high *in vitro* and throughout infection ([fig. 4*a*](#evx030-F4){ref-type="fig"}), it still offers an attractive candidate for further study. The rest of the SsPEs that were not conserved in other fungi were not significantly differentially expressed *in planta* relative to during growth *in vitro*. Overall, they exhibited relatively low levels of expression. It is possible that these sequences are specific to hosts other than *B. napus*, and require differential signals for induction *in planta*. An alternative hypothesis is that low transcript abundance is all that is needed for the activity of these proteins. It is also possible that they are only highly expressed at a very specific time point, which was missed in this infection assay. Such an expression pattern has been shown for effector-like genes in other fungi, for example, *Zymoseptoria tritici* ([@evx030-B81]). A total of nine of SsPEs were significantly upregulated either early (defined as 1, 3, 6, and 12 hpi) or late (defined as 24 and 48 hpi) or both early and late during infection ([fig. 4*b* and *c*](#evx030-F4){ref-type="fig"}). Apart from the already mentioned model Sscle02g012940, all of these SsPEs were conserved in other fungi ([supplementary table 6](#sup1){ref-type="supplementary-material"}, [Supplementary Material](#sup1){ref-type="supplementary-material"} online). Of particular note was the SsPE Sscle01g008950, which was only significantly upregulated at 3 hpi ([fig. 4*b*](#evx030-F4){ref-type="fig"}). This would suggest that this SsPE is required before the onset of host necrosis, perhaps as a suppressor of immune responses as has been demonstrated for the *S. sclerotiorum-*secreted integrin like protein and numerous effector proteins in other fungal plant pathogens ([@evx030-B101]; [@evx030-B106]). Identification of a Seven-Member Effector Gene Family in *S. sclerotiorum* That May Have Arisen through Recent Transposition ---------------------------------------------------------------------------------------------------------------------------- In several plant pathogenic fungal and oomycete species, a clear link between the genomic positions of effector-like genes and TEs has been identified ([@evx030-B79]; [@evx030-B41]; Amselem, Lebrun, et al. 2015; [@evx030-B84]; [@evx030-B6]; [@evx030-B31]). It has been hypothesized that such positioning allows for expansion of effector gene families and subsequent diversification through mutation ([@evx030-B26]). This, in theory, could enhance the capacity of a fungus to rapidly adapt to newly evolved R genes or lost susceptibility loci in host plants. Additionally, RIP, a fungal defense mechanism against TEs, is thought to enhance mutation rates of effectors embedded within TE-rich regions. This process involves mutation of C to T bases in duplicated sequences. Repeat-rich regions containing effector-like proteins are usually gene-poor compared with the rest of the genome. This allows for housekeeping genes to remain unaffected by TE-associated evolutionary processes. In the *B. napus* pathogen *L. maculans*, it has been shown that most TEs are inactive. This is thought to be a result of an ancestral TE invasion followed by extensive RIP. Approximately 20% of effector-like sequences in *L. maculans* are associated with RIP-affected TE containing regions, whereas only 4.2% are associated with gene-rich regions. Though the sequences in repetitive regions generally do not appear to be in families that have expanded through TE activity, it has been shown that they are likely to have undergone RIP-like mutation ([@evx030-B79]). In the powdery mildew fungus, *B. graminis*, several hundred effector-like genes have been identified *in silico.* These genes occur in 72 families of up to 59 members. These sequences are generally associated with TE-rich regions, and their expansion is thought to be a result of transposition ([@evx030-B70]). Interestingly, it has been shown that the EKA (effectors homologous to *Avr* k *1* and *Avr* a *10*) family of effectors is likely to have originated from a truncated TE ORF, which the fungus co-opted as an effector (Amselem, Vigouroux, et al. 2015). In the oomycete pathogen *P. infestans*, approximately 74% of the genome is repetitive. These repetitive regions harbor extensively expanded, rapidly evolving effector proteins ([@evx030-B44]). This is perhaps one of the most extensively studied and clearest examples of a compartmentalized genome in a microbial plant pathogen. Based on these observations in fungal and oomycete species, we hypothesized that a similar dynamic may be present in *S. sclerotiorum*. We carried out several analyses to test this hypothesis. To determine whether SsPEs occurred in families, we conducted OrthoMCL analysis using the gene predictions of *S. sclerotiorum* and *B. cinerea* to detect recent paralogs. Further, to determine whether SsPE families were likely formed through TE activity, we conducted multiple sequence alignment of regions containing effectors and inspected them for the presence of TE sequences as predicted by REPET. By doing this, we identified a single SsPE family, which included seven proteins with more than 90% amino acid identity. The sequences in this family exhibited a single ortholog in *B. cinerea*, indicating that they possibly appeared subsequent to the divergence of *S. sclerotiorum* and *B. cinerea*. Six of the sequences in this family were already identified by [@evx030-B42]), though a detailed investigation of their association with transposition activity has not been carried out. None of these sequences contained any predicted functional domains, underlining their possible specialized functions as effectors. Furthermore, multiple alignment of regions surrounding these sequences showed that they were within a 4--5 kb genomic segment that was repeated across six different chromosomes. Intriguingly, each of these SsPEs was embedded in either a partial or complete *Gypsy* LTR-retroelement ([fig. 5](#evx030-F5){ref-type="fig"}). This would indicate that these duplicated sequences arose as a result of TE activity. Thus, there is some evidence of effector evolution occurring through transposition in *S. sclerotiorum*, the first evidence of this phenomenon in this species thus far. Based on this observation, we hypothesized that *S. sclerotiorum* may harbor TE*-*rich, gene sparse, and potentially RIP*-*affected regions that are enriched for effector proteins. To test this hypothesis, we performed a comparative analysis of the genome of *S. sclerotiorum* with the "two*-*speed" genomes of *B. graminis*, *L. maculans*, and *P. infestans*. We found that out of these four microbial genomes, only that of *L. maculans* showed a bimodal GC content ([fig. 6*a*](#evx030-F6){ref-type="fig"}). This is consistent with results from the same analysis performed by [@evx030-B93]), who used the previous version of the *S. sclerotiorum* genome. This would indicate that GC content of *B. graminis, P. infestans*, and *S. sclerotiorum* is consistent across the whole genome. There are two possible explanations for this: 1) The genomes analyzed do not exhibit or only exhibit a limited level of RIP and 2) the genomes exhibit a consistent level of RIP throughout, and do not harbor RIP-affected regions in distinct, GC-depleted compartments. It has been shown that *S. sclerotiorum* and *L. maculans* exhibit active RIP, whereas *B. graminis* does not; *P. infestans* is not thought to harbor RIP as RIP is a fungus-specific mechanism ([@evx030-B46]). *Sclerotinia sclerotiorum* displays an intermediate frequency of RIP and *L. maculans* displays a high frequency of RIP. *Blumeria graminis* shows an RIP-like signature in some TE copies but it is thought that this is the result of ancestral activity, as this species does not harbor the genes necessary for RIP. Unlike *L. maculans*, *S. sclerotiorum* has been shown to exhibit two kinds of RIP, both CpT = \>TpT and CpA  = \> TpA transitions (Amselem, Lebrun, et al. 2015). Our results are consistent with this analysis as *S. sclerotiorum* was found to exhibit dominance of these two types of RIPs across copies of the most abundant TE ([fig. 6*b*](#evx030-F6){ref-type="fig"}). Additionally, the TpA/ApT RIP indices of the five most numerous TEs were significantly higher than random sets of control sequences ([fig. 6*b*](#evx030-F6){ref-type="fig"}). As *B. graminis* has only an extremely weak signature of ancestral RIP, it is unlikely to exhibit a bimodal GC content at the whole-genome scale. However, as RIP appears to be at a fairly significant level in *S. sclerotiorum*, the lack of bimodality of its genome GC content could indicate a lack of specific RIP-affected genome compartments such as those observed in *L. maculans*. To test whether secreted and effector-like proteins were associated with TEs and/or RIP in the four filamentous plant pathogens, both secreted proteins and effector-like proteins were compared with randomized control sets of proteins. This showed that in both *B. graminis* and *P. infestans* there was a significant association between repeats and both secreted and effector-like proteins. In *L. maculans* there was a significant association between secreted proteins and TE sequences, but no significant association between effector-like proteins and TE sequences. However, adjusting α to 0.1 showed that there was possible evidence of a general trend toward association of effector sequences with repeats in *L. maculans* ([fig. 7*a*](#evx030-F7){ref-type="fig"}). These results are consistent with the previously proposed hypotheses that *B. graminis* and *P. infestans* do not exhibit active RIP but do compartmentalize effector sequences in TE-rich regions ([@evx030-B44]; [@evx030-B70]). It is also consistent with the hypothesis that *L. maculans* repeats have been subjected to extensive RIP, and were hence not as easily discovered by our standardized repeat calling pipeline (RepeatMasker). Despite this, there was still evidence of association of secreted and effector-like proteins with repeat sequences in *L. maculans*. Conversely, in *S. sclerotiorum*, there was no significant association between secreted or effector-like proteins and TEs. This would suggest that TE activity is not a dominant mode of effector or secreted protein evolution in *S. sclerotiorum*. Indeed, the high degree of homology of SsPEs to proteins in other fungi (90% were homologous at e^−10^) would indicate that they may be fulfilling conserved functions that are not under extensive selection pressure exerted by host species. To test whether secreted and effector-like proteins are co-located with RIP-affected sequence in *S. sclerotiorum*, we performed the same analysis as for the repeat sequences for regions with a high RIP index as specified by RipCal. This showed that *L. maculans* exhibited a significant association between both secreted proteins and effector-like proteins and RIP-affected sequences, whereas *B. graminis* exhibited a significant negative association between effector and secreted proteins and RIP-affected sequence. In *P. infestans*, there was no association between secreted or effector proteins and RIP-affected sequence. This is consistent with the previous observation that *L. maculans* exhibits significant RIP activity, which has a particular impact on the evolution of effector proteins compartmentalized into AT-rich regions ([@evx030-B79]). The data would also indicate that *B. graminis* requires active transposons for effector diversification and that effector proteins are significantly further from regions that may have undergone RIP ancestrally. Intriguingly, in *S. sclerotiorum*, there was a significant association between secreted proteins and RIP-affected sequence. However, there was no association between effector-like sequences and RIP-affected sequence. This would indicate that there may be some hot spots of RIP associated with secreted protein diversification in the *S. sclerotiorum* genome. Indeed, several regions can be readily identified in [figure 1](#evx030-F1){ref-type="fig"}. For example, at the distal ends of chromosomes 7, 15, and 16 there appear to be clusters of secreted and effector-like proteins that occur in proximity to regions with a high RIP index and several repeat sequences. However, the impact of these regions on evolution and host specificity remain to be elucidated. Based on these observations, we tested the hypothesis that secreted proteins were associated with AT-rich regions in *S. sclerotiorum*. Furthermore, we tested whether they were associated with gene sparse regions, which would be indicative of a significant level of compartmentalization away from housekeeping genes. We found that in *S. sclerotiorum*, *B. graminis*, and *P. infestans*, there was no correlation between secreted protein proportion and GC content. However, in *L. maculans*, there was a significant negative correlation between GC content and secreted protein proportion ([fig. 8*a*](#evx030-F8){ref-type="fig"}). This would indicate that in *S. sclerotiorum*, there is no significant compartmentalization of secreted proteins into AT-rich (i.e., RIP-affected) sequence regions. Though it is possible that there are hotspots of this kind of activity in the genome, a 100-kb sliding window was too large to detect them. This is a limitation of this study. However, as it was able to detect an association in *L. maculans*, for which RIP-affected genome compartments with effectors have been characterized, it would at least demonstrate that in *S. sclerotiorum* this kind of phenomenon is not as extensive as in *L. maculans.* In addition, we found that the genomes of *B. graminis*, *P. infestans*, and *L. maculans* exhibited a significant negative correlation between CDS content (used to infer gene richness) and secreted protein proportion. This supports previous observations that effector-like and secreted proteins are often positioned in gene sparse regions in these species. Conversely, there was no correlation between gene content and secreted protein proportion in *S. sclerotiorum*. Again, it is possible that there are small hotpots of secreted protein diversification that are gene poor; [figure 1](#evx030-F1){ref-type="fig"} illustrates such candidate regions (e.g., the distal end of chromosome 7). However, at a macroscale, this phenomenon was not identified in *S. sclerotiorum* as it was in organisms previously characterized ([@evx030-B76]). Therefore, we propose that compartmentalization of secreted proteins in *S. sclerotiorum*, if at all present, is far subtler than it is in the host-specific biotrophic and hemibiotrophic organisms included in this study. Conclusion ========== In conclusion, we have produced a complete and accurate genome of the important broad host range necrotrophic fungus *S. sclerotiorum*. We have identified a number of novel effector candidates for future studies and elucidated their expression patterns *in planta*. Further, we show that the genome architecture of *S. sclerotiorum*, in terms of TEs, RIP, and secreted and effector-like proteins, is different to three of the most well-characterized filamentous fungi and oomycetes that conform to the two-speed genome hypothesis. Instead, we demonstrate subtle signatures of enhanced mutation of secreted proteins and effectors in *S. sclerotiorum* through RIP activity and transposition, which is not generally detectable at a whole-genome scale as in the other species tested. This has implications for effector discovery and comparative genomic analyses in the future. Supplementary Material ====================== [Supplementary data](#sup1){ref-type="supplementary-material"} are available at *Genome Biology and Evolution* online. Acknowledgements ================ MFS is supported by a Veni grant of the Research Council for Earth and Life Sciences (ALW) of the Netherlands Organization for Scientific Research (NWO). KH-K is supported by the UK Biotechnology and Biological Sciences Research Council (BBSRC) through the Institute Strategic Programme Grant 20:20 Wheat (BB/J/00426x/1). SH was funded by a BBSRC Industrial Collaborative Award in Science and Engineering (CASE) studentship supported by Syngenta entitled 'Early sensing of plant pathogenic fungi'. JR received support for this project from the USDA National Institute of Food and Agriculture, Hatch project 1005726. MM, ON and SR are funded by the European Research Council (ERC-StG 336808 project VariWhim) and the French Laboratory of Excellence project TULIP (ANR-10-LABX-41; ANR-11-IDEX-0002--02; New Frontiers grant 'ScleRNAi'). DH and SS are funded by SaskCanola and the Government of Canada through the Developing Innovative Agri-Products programme (project \# J-000269: P032). RPO conducted part of this research in Wageningen Agricultural University Laboratory of Phytopathology as a KNAW research fellow. MCD, MD-G and RPO are funded by the Grains Research and Development Corporation (GRDC) as part of a bilateral agreement between Curtin University under the grant CUR00023, within the Centre for Crop and Disease Management (CCDM). This work was in part supported by resources provided by The Pawsey Supercomputing Centre with funding from the Australian Government and the Government of Western Australia. MCD would like to personally thank Robert King and Keywan Hassani-Pak (Rothamsted Research) for advice and instruction regarding genome assembly during the initial stages of the project. [^1]: **Data deposition:** Details= Chr_1 CP017814 Chr_2 CP017815, Chr_3 CP017816, Chr_4 CP017817, Chr_5 CP017818, Chr_6 CP017819, Chr_7 CP017820, Chr_8 CP017821, Chr_9 CP017822, Chr_10 CP017823, Chr_11 CP017824, Chr_12 CP017825, Chr_13 CP017826, Chr_14 CP017827, Chr_15 CP017828, Chr_16 CP017829. [^2]: **Associate editor:** Rebecca Zufall [^3]: N[ote]{.smallcaps}.---In column six, a "x" indicates an "effector-like" prediction in [@evx030-B41], and a "+" indicates a sequence in the six-gene family paralogous to Sscle03g031910 (previously SS1G_13371) identified in Guyon et al.

1. Introduction {#sec0005} =============== One of the leading causes of water pollution is the unchecked release of wastewater from various industries into water bodies and many other environments \[[@bib0005],[@bib0010]\]. The generation of wastewater is mostly due to rapidly growing industrial sector \[[@bib0015]\] for the development and expansion of the nation's economy. Amongst the innumerable industries, the pesticide industry is counted as one of the key contributors of water contamination. Organochlorine (OC) and organophosphorus (OP) pesticides are most important contaminants released by pesticide industry around the world as well as in India \[[@bib0020],[@bib0025]\]. The existence of pesticide residues in water and soils impact on the vegetables as well as fruits and thus poses grave danger to human health. Many findings displayed that even very low level of pesticides cause natal defects \[[@bib0030]\]. Numerous scientific endeavours in the area of genotoxicity of wastewater suggested direct association with mutagenicity of pollutants into water bodies \[[@bib0035],[@bib0040]\]. Several industrial wastewater effluents and sludges has shown high mutagenic potential \[[@bib0045],[@bib0050]\]. With the fast pace in the development and era of modern mechanization the problem of pollution, specifically water pollution has been increased alarmingly in numerous developing countries including India \[[@bib0055], [@bib0060], [@bib0065]\]. A lot of toxicants in the environment act by damaging of DNA and therefore causing mutations \[[@bib0070], [@bib0075], [@bib0080]\]. Genotoxicity evaluation of industrial effluents on surface water indicates the presence of mixtures comprise of various toxic substances that may stance risk of hazard and carcinogenicity \[[@bib0085],[@bib0090]\]. Biological assays with prokaryotic system detect mutagenic agents that persuade the gene level mutation and primarily damages the DNA. In contrast, eukaryotic based bioassay revealed exposure of a more degree of injury/impairment, variable from gene mutations to chromosomal aberrations and aneuploidies \[[@bib0095]\]. Applying both the prokaryotic and eukaryotic based detection systems reinforce and relate the observations to make certain if the substances actually hold any adverse effects on the genetic materials. Ames *Salmonella*/microsomal test is extensively applied in examining the mutagenic potential of toxic chemicals \[[@bib0100],[@bib0105]\]. *A. cepa* plant model is also extensively used for the evaluation of genotoxicity due to high sensitivity towards the xenobiotic compounds \[[@bib0110]\]. Mung bean (*Vigna radiata*) seed is another important short-term assay for genotoxicity evaluation using different parameters such as seed germination, seedling vigour index \[[@bib0035]\] to reflects the impacts of contaminants on the growth of plant. Many other *in-vitro* tests for the evaluation DNA damages are also routinely applied in the studies of wastewater monitoring. Among these tests, plasmid-nicking assay has been usually employed and delivers an effective indicator of genotoxicity \[[@bib0115]\]. By the means of both the prokaryotic and eukaryotic evaluation systems, it supports and corelate the observations and confirmed that the xenobiotic compounds/toxic chemicals severely affect the genes of both the systems. Present study focused on the cytotoxicity, genotoxicity and phytotoxicity of pesticide industry wastewater collected from in the vicinity of Ghaziabad city, India, using different prokaryotic and eukaryotic assays. Additionally, plasmid nicking assay has also been used to evaluate the direct impact of wastewater on DNA integrity. 2. Materials and methods {#sec0010} ======================== 2.1. Sample collection {#sec0015} ---------------------- Wastewater samples were collected from the open channels, receiving sewage from industrial area of Ghaziabad, situated about 10 km North of the Hindon River at latitude 28°40´ North and longitude 77°25´ East. A total 12 samples of wastewater were taken from January 2015 to June 2017 and transferred to the laboratory as described in standard methods \[[@bib0120]\]. Two litres of samples taken from five diverse points and make 10 L by composite mixing. 2.2. Physico-chemical characteristics of the wastewater {#sec0020} ------------------------------------------------------- Physico-chemical properties of the wastewater such as total dissolve solid (TDS), pH, carbonate, bicarbonate, sulphate and chloride were carried out according to method adopted by Gupta \[[@bib0125]\]. 2.3. Preparation of concentrated wastewater extracts {#sec0025} ---------------------------------------------------- ### 2.3.1. XAD-concentration of wastewater {#sec0030} One litre of wastewater was used to concentrate organic constituents. Whatman filter paper no. 1 with pore size 11 μm and 0.45 μm pore size (Axiva, India) were used to filter the wastewater. The adsorbent columns were prepared by intermixing equal quantity of XAD-4 and XAD-8 \[[@bib0130]\]. Organic compounds present in the wastewater were adsorbed on the resins using methods described by Wilcox and Williamson \[[@bib0135]\]. The adsorbed organic compounds were eluted with 20 ml of acetone (HPLC grade) and then eluate was evaporated to dryness at room temperature (25 °C) under decreased pressure and further dissolved in Dimethyl sulfoxide (DMSO) (HPLC grade, SRL, India). The samples were filtered through membrane filter (0.22 μm) to sterilized and stored for further use at −20 °C. ### 2.3.2. Liquid-liquid extraction {#sec0035} Industrial wastewater samples were sequentially extracted by means of different organic solvents such as Dichloromethane (DCM) and n-Hexane (HPLC grade, SRL, India) as defined in standard procedures \[[@bib0120]\]. Wastewater was vigorously shaken in a separating funnel with the solvent and was set aside to hold up till the aqueous (water) and organic solvent layers were separated. The solvent layer was collected in a 100 ml beaker then evaporate at 25 °C to concentrate up to 5 ml. The obtained extracts of wastewater were filter (pore size 0.22 μm) sterilized and stored at −20 °C for further use of genotoxicity testing \[[@bib0140]\]. 2.4. Gas chromatographic (GC) analysis of pesticides in wastewater {#sec0040} ------------------------------------------------------------------ Gas chromatography analysis of wastewater extracts was performed via GC-2010 gas chromatograph (Shimadzu, Japan). The parameter of the instrument and operating conditions are as follows: {column: Rtx-5MS, temperature: (injector: 290 °C, detector: 300 °C, oven: initial temperature 100 ◦C/min then increase 300 °C then 5 °C/ram, hold time: 1--9 min, carrier flow rate of gas helium: 21 ml/min, flow rate of carrier gas helium: 1.21 ml/min afterward makeup 30 ml/min)}. By comparing retention time of standards obtained from Sigma-Aldrich, peaks of samples were identified. Multi-standard of 20 organochlorine (CRM-47426) and nine organophosphorus (48,391) pesticide mixtures purchased from Sigma-Aldrich company containing Aldrin, α-BHC, β-BHC, σ-BHC, Endrin Aldehyde, Endrin, Endrin ketone, Chlordane, γ-Chlordane, Dieldrin, 4-4″ DDT, 4-4″ DDE, Lindane, Heptachlor, Heptachlor Epoxide, Endosulfan I, Endosulfan II, Endosulfan sulfate, Methoxychlor, (Organochlorine), Azinphos-methyl, Chlorpyriphos, Dichlorvos, Ethoprophos, Disulfoton, Parathion-methyl, Fenchlorphos, Prothiofos, Malathion (organophosphorus) and stored at 4◦C. 2.5. Ames/*Salmonella* mutagenicity test {#sec0045} ---------------------------------------- The *Salmonella* mutagenicity test was performed as described by Maron and Ames \[[@bib0100]\] with minor changes as adopted earlier \[[@bib0145]\]. Five different doses of individual wastewater extract i.e., 2.5, 5, 10, 20 and 40 μL per plate (add 0.1 ml of the overnight grown bacterial culture) were incubated for 30 min at 37 °C in triplicate. Two ml of top agar with trace amount of biotin and histidine were added and poured onto plates of minimal glucose agar and incubated at 37 °C for 2--3 days. Positive control comprising bacterial culture and methyl methane sulfonate (MMS) and negative control includes bacteria and double distilled water. Parallel tests were also performed in the presence of S9 microsomal fraction to detect incidence of pro-mutagens in the samples containing 20 μL of S9 liver homogenate per plate. 2.6. *Allium cepa* anaphase-telophase test {#sec0050} ------------------------------------------ To determine the toxic effect of wastewater, samples were tested using root tip cells of *A. cepa* as described by Fiskesjo \[[@bib0150]\]. The small bulbs of *A. cepa* (2n = 16) with 1.5--2.0 cm in diameter were purchased from local market. Before starting the assay, outer dead scales and dry bottom of *A. cepa* bulbs were detached without disturbing root primordia. The bulbs were put in beakers, comprising DD water. The basal portions of bulbs dipped into the water and allow to evolve for 2--3 days at room temperature (25 °C). The freshly grown roots upto 2 cm were used in this assay. The newly grown root tips were treated with several different concentration of wastewater i.e., 5, 10, 25, 50, and 100% for 3 days. Simultaneously, positive control with MMS and negative control with double distilled water were also performed in each test. After exposure to wastewater samples upto three days, root tips were selected randomly, fixed in the ratio of 3:1 ethanol and glacial acetic acid (v/v) and incubated for overnight at 4 °C. The fixed root tips were washed with tap water followed by heating for 2--3min in 1 N HCl then with DD water and stained with acetocarmine and observed under light microscope (Olympus, BX60). By observing approximately 6000 dividing cells (2000 cells per slide), mitotic index (MI) was calculated as follow:$$\text{M}\text{i}\text{t}\text{o}\text{t}\text{i}\text{c}\,\text{i}\text{n}\text{d}\text{e}\text{x}\,(\text{\%}) = \frac{\text{T}\text{o}\text{t}\text{a}\text{l}\,\text{N}\text{u}\text{m}\text{b}\text{e}\text{r}\,\text{o}\text{f}\,\text{D}\text{i}\text{v}\text{i}\text{d}\text{i}\text{n}\text{g}\,\text{C}\text{e}\text{l}\text{l}\text{s}}{\text{T}\text{o}\text{t}\text{a}\text{l}\,\text{N}\text{u}\text{m}\text{b}\text{e}\text{r}\,\text{o}\text{f}\,\text{C}\text{e}\text{l}\text{l}\text{s}\,\text{E}\text{x}\text{a}\text{m}\text{i}\text{n}\text{e}\text{d}} \times 100\,$$ Chromosomal aberrations were evaluated by observing approximately 300 dividing cells (preferably 100 cells per slide). 2.7. Phytotoxicity testing of wastewater {#sec0055} ---------------------------------------- ### 2.7.1. Effect of wastewater on germination and growth of *Vigna radiata* (mung bean) under *in vitro* condition {#sec0060} Phytotoxicity of wastewater on seed germination was done according to the method of Kalyani et al \[[@bib0155]\]. Briefly, mung bean seeds (*Vigna radiata* L. Wilczek) were surface sterilized using 70% ethanol followed by 3% sodium hypochlorite solution for 3 min. The sterilized seeds were repeatedly washed through DD water. The sterilized seeds of mung bean were soaked in different concentrations of filtered wastewater for overnight and then placed on 0.7% agar plates. The agar plates were also prepared with sterile DD water (control). All the plates were incubated at room temperature (25 °C) with humidity upto 75%. After 5--7 days of incubation the emergence of seeds, length of plumule and radicle, and dry biomass were recorded.$$\%\text{Germination} = \frac{\text{number\ of\ seeds\ germinated}}{\text{total\ number\ of\ seeds}} \times 100$$ The seedling vigor index (SVI) were calculated from percent germination of seeds \[[@bib0160]\]. ### 2.7.2. Oxidative damage induced by wastewater under confocal microscopy {#sec0065} In order to assess the oxidative damage induced by wastewater; confocal microscopy was used to observe the dead cells. Briefly, roots of *V. radiata* plants were grown on 0.7% agar plates amended with filtered wastewater of different concentration i.e., 10%. 25%, 50% and 100% for seven days \[[@bib0165]\]. Negative (DD water) and positive control (MMS) were also performed in each assay. After three time washing with phosphate buffer saline (PBS), propidium iodide (PI) were used to stain the root samples and fixed on a glass slide and observed under LSM-780 Confocal Microscope (Zeiss, Germany). 2.8. Plasmid nicking assay {#sec0070} -------------------------- The plasmid-nicking assay was performed as described by Siddiqui et al \[[@bib0115]\] with minor changes, the covalently closed circular pBR322 plasmid DNA (0.5 μg) was treated with different concentration of wastewater in a total volume of 25 μl for 3 h at room temperature. After incubation at room temperature, 5 μl of 5x tracking dye (40 mM EDTA, 0.05% bromophenol blue with glycerol 50% v/v) was mixed into the reaction tube and were run with one percent agarose on gel electrophoresis at 50 mA for 90 min followed by staining with ethidium bromide (0.5 μg/l). The DNA bands in the agarose gel were visualized on a BIO-RAD Chemi Doc XRS imaging system and photographed. 2.9. Statistical analysis {#sec0075} ------------------------- Mutagenic Index, induction factor (Mi) and mutagenic potential (*m*) were calculated as defined by Ansari and Malik \[[@bib0055]\].$$\text{Mutagenic\ index} = \frac{\text{Number\ of\ his\ +\ revertants\ induced\ in\ the\ sample}}{\text{Number\ of\ his\ +\ revertants\ induced\ in\ the\ negative\ control}}$$$$\text{Induction\ factor}\left( \text{Mi} \right) = \frac{\text{ln}\left( \text{n-c} \right)}{\text{c}}$$Where "n" is no. of revertant bacterial colonies in the samples while "c" is the number of revertant colonies in control. The mutagenic potential of the wastewater samples was calculated as described by Khan et al \[[@bib0040]\]. The total number of *his^+^* revertant bacterial colonies in comparison to control was recognized by one-way analysis of variance (ANOVA) at p ≤ 0.05. Data were represented in terms of percent mitotic index and percentage of abnormal cells. In case of seed germination assay, percent germination of seed and plumule-radicle growth were represented as Mean ± Standard Deviation (SD) and analysed with Duncan Multiple Range Test (DMRT) were applied to analyse significance in the treatment sets as well as in contradiction of positive and negative control data. 3. Results {#sec0080} ========== 3.1. Physico-chemical and heavy metal analysis {#sec0085} ---------------------------------------------- The physico-chemical characteristics of wastewater is presented in the [Table 1](#tbl0005){ref-type="table"}. Test samples displayed pH in the range of 7.0--7.3. The concentration of total dissolve solids, carbonate, bicarbonate, chloride and sulphate were recorded to be 767 mg L^−1^, 164.5 mg L^−1^, 70.47 mg L^−1^, 45.38 mg L^−1^ and 0.05 mg L^−1^, respectively. Atomic absorption spectrophotometric (AAS) analysis revealed the presence of numerous heavy metals i.e. Ni (0.45 mg L^−1^), Cu (0.13 mg L^−1^), Cr (1.91 mg L^−1^), Pb (1.17 mg L^−1^), Cd (0.02 mg L^−1^) and Zn (0.13 mg L^−1^), with concentration of Cd, Cr and Pb being pointedly higher than permissible limits as given by Unites States Environmental Protection Agency (US EPA).Table 1Physico-chemical and heavy metal analysis of industrial wastewater.Table 1ParametersWastewaterpH7.17 ± 0.1TDS767 ± 9.46Carbonate164.5 ± 6.41Bicarbonate70.47 ± 2.52Chloride45.38 ± 2.39Sulphate0.05 ± 0.001Dissolve oxygen2.27 ± 0.02Free CO~2~21.83 ± 0.62Total CO~2~26.24 ± 1.03Nickle0.45 ± 0.12Cadmium0.2 ± 0.01Lead1.17 ± 0.1Copper0.13 ± 0.08Chromium1.91 ± 0.3Zinc0.13 ± 0.01[^1] 3.2. Determination of organochlorine and organophosphate pesticides {#sec0090} ------------------------------------------------------------------- The gas chromatographic (GC) analysis revealed that the industrial wastewater contains several organochlorine pesticides such as α-BHC, β-BHC, σ-BHC, Aldrin, γ-Chlordane, Endosulfan I, Endosulfan II, Endosulfan sulfate, Dieldrin, Endrin Aldehyde, Endrin ketone at the concentration of 82.9, 38.01, 7.52, 108.6, 12.4, 4.37, 311.7, 22.49, 6.43, 125.8, 4.21 and 66.61, ng mL^−1^ respectively ([Table 2](#tbl0010){ref-type="table"}), while organophosphorus pesticides Ethoprophos, Disulfoton, Parathion-methyl, Chlorpyrifos, Prothiofos were detected at the levels of 3.54, 22.4, 20.6, 8.93 and 321.5 ([Table 2](#tbl0010){ref-type="table"}) ng mL^−1^, respectively. Many other unidentified peaks were also detected in the gas chromatograms of test samples showing the occurrence of more organic pollutants apart from pesticide.Table 2Concentration of pesticides in industrial wastewater as determined by gas chromatography.Table 2Organochlorine (OC)Concentration(ng/ml)Organophosphate (OP)Concentration(ng/ml)α-BHC82.9 ± 6.72DichlorvosNDβ-BHC38.01 ± 2.51Ethoprophos3.54 ± 0.67LindaneNDDisulfoton22.4 ± 1.83σ-BHC7.52 ± 0.73Parathion-methyl20.6 ± 2.01HeptachlorNDFenchlorphosNDAldrin108.6 ± 6.2Chlorpyrifos8.93 ± 1.55Heptachlor epoxideNDProthiofos321.5 ± 33γ-Chlordane12.4 ± 2.1Azinphos-methylNDα-ChlordaneNDMalathionNDEndosulfan I4.37 ± 0.454-4" DDENDDieldrin311.7 ± 49.7EndrinNDEndosulfan II22.49 ± 4.74-4" DDDNDEndrin aldehyde6.43 ± 1.2Endosulfan sulfate125.8 ± 22.34-4" DDT4.21 ± 0.57Endrin Ketone66.61 ± 6.23MethoxychlorND[^2] 3.3. Reversion of *Salmonella* tester strains {#sec0095} --------------------------------------------- The wastewater samples were evaluated for their mutagenicity using *S. typhimurium* strains. Liquid-liquid extracted (DCM and n-hexane) and XAD concentrated samples were tested in the presence and absence of S9 fraction. In XAD concentrated samples, the number of revertant colonies increased upto 20 μL/plate with all the tester strains. The maximum number of revertants were observed in TA98 and showed mutagenic index of 13.11 (without S9 fraction) and 13.19 (with S9 fraction) among all the tester strains. The strain TA98 revealed highest response in terms of induction factor (2.50 without and with S9 fraction both) and mutagenic potential/slope (m) were observed 7.7 with S9 fraction and 6.8 without S9 fraction ([Table 3](#tbl0015){ref-type="table"}). On the basis of induction factor and mutagenic index, the order of responsiveness both in the presence and absence of S9 fraction for XAD concentrated wastewater sample was as: TA98\>TA97a\>TA100\>TA102\>TA104. TA98 showed maximum responsiveness followed by TA100, TA97a, TA104 and TA102.Table 3Reversion of *Salmonella* tester strains in the presence of XAD concentrated wastewater.Table 3Wastewater extract (μL/plate)F valueStrainS9Control2.55102040MimLSD (*p*≤0.05)TA97a--88 ± 18225 ± 27 (2.53)297 ± 27 (3.35)373 ± 18 (4.21)456 ± 22 (5.14)381 ± 12 (4.31)1.435.8729.8281.6+94 ± 11241 ± 17 (2.56)317 ± 17 (3.37)405 ± 21 (4.31)484 ± 24 (5.15)407 ± 23 (4.33)1.436.2114.9145.1TA98--35 ± 6221 ± 22 (6.26)268 ± 25 (7.57)367 ± 18 (10.37)463 ± 14 (13.11)382 ± 19 (10.41)2.506.7924.1174.7+39 ± 11250 ± 15 (6.40)299 ± 26 (7.65)405 ± 22 (10.37)515 ± 16 (13.19)433 ± 20 (11.01)2.507.7129.6121.5TA100--127 ± 14232 ± 18 (1.82)301 ± 11 (2.37)411 ± 19 (3.23)501 ± 18 (3.94)399 ± 27 (3.13)1.085.9227.585.6+138 ± 16253 ± 20 (1.83)329 ± 22 (2.37)451 ± 17 (3.26)556 ± 14 (4.01)453 ± 16 (3.27)1.116.9517.2111.2TA102--226 ± 12324 ± 15 (1.43)410 ± 21 (1.81)497 ± 25 (2.20)560 ± 19 (2.48)432 ± 14 (1.91)0.394.1313.5231.5+241 ± 8354 ± 17 (1.46)441 ± 19 (1.83)510 ± 24 (2.12)598 ± 17 (2.48)465 ± 17 (1.93)0.394.4710.8118.9TA104--301 ± 17379 ± 19 (1.25)464 ± 24 (1.54)539 ± 21 (1.77)625 ± 17 (2.07)515 ± 22 (1.71)0.074.6819.5165.4+318 ± 16408 ± 22 (1.28)493 ± 21 (1.55)568 ± 16 (1.78)668 ± 24 (2.09)569 ± 16 (1.78)0.104.4916.7217.9[^3] The revertant colonies in the hexane extracted wastewater sample increased with increasing dose from 2.5 to 20 μL/plate, while at 40 μL/plate dose, revertant colonies decreased ([Table 4](#tbl0020){ref-type="table"}). Mutagenic index in strain TA98 showed maximum response (12.09 without S9 and 12.15 with S9 fraction); induction factor (2.41 without S9 and 2.42 with S9 fraction); and mutagenic potential is 8.0 with S9 and 7.4 without S9 fraction ([Table 4](#tbl0020){ref-type="table"}). The mutagenic index and induction factor, the response of tester strains showed dissimilar trends as observed in the XAD concentrated wastewater samples was as follows: TA98\>TA97a\>TA100\>TA102\>TA104.Table 4Reversion of *Salmonella* tester strains in the presence of hexane extracted wastewater.Table 4Wastewater extract (μL/plate)F valueStrainS9Control2.55102040MimLSD (*p*≤0.05)TA97a--88 ± 13167 ± 19 (1.90)204 ± 14 (2.32)276 ± 13 (3.15)347 ± 25 (3.96)281 ± 12 (3.21)1.074.236.2214.1+93 ± 11191 ± 12 (2.06)221 ± 15 (2.38)293 ± 16 (3.16)378 ± 22 (4.07)323 ± 21 (3.48)1.114.99.52156.9TA98--35 ± 7144 ± 15 (4.08)260 ± 23 (7.36)344 ± 19 (9.73)427 ± 15 (12.09)384 ± 14 (10.85)2.417.48.65321.2+38 ± 11161 ± 9 (4.20)286 ± 18 (7.47)375 ± 13 (9.79)466 ± 24 (12.15)418 ± 20 (10.91)2.428.012.8286.5TA100--130 ± 8219 ± 10 (1.68)302 ± 21 (2.31)376 ± 18 (2.88)443 ± 24 (3.40)406 ± 23 (3.11)0.885.97.69123.2+147 ± 12253 ± 13 (1.73)343 ± 21 (2.33)424 ± 24 (2.89)501 ± 17 (3.41)458 ± 13 (3.12)0.886.68.9254.1TA102--234 ± 9291 ± 22 (1.24)361 ± 18 (1.54)431 ± 22 (1.83)493 ± 17 (2.10)433 ± 13 (1.84)0.104.410.5165.3+246 ± 10325 ± 15 (1.31)396 ± 12 (1.60)458 ± 22 (1.86)532 ± 26 (2.15)466 ± 19 (1.89)0.154.711.9154.8TA104--311 ± 15362 ± 18 (1.16)413 ± 12 (1.32)467 ± 21 (1.49)530 ± 19 (1.70)476 ± 19 (1.53)−0.353.713.0218.9+332 ± 11392 ± 20 (1.17)456 ± 25 (1.37)503 ± 8 (1.51)574 ± 16 (1.72)509 ± 17 (1.53)−0.323.97.99224.8[^4] The reversion of *Salmonella* strains with acidic and basic fraction of DCM extracts are presented in [Table 5](#tbl0025){ref-type="table"}, [Table 6](#tbl0030){ref-type="table"}. DCM extract of basic fraction showed highest response of 12.5 in absence of S9 and 12.72 in presence of S9 fraction in terms of mutagenic index; 2.44 without S9 fraction whereas 2.46 along with S9 fraction in terms of induction factor; and 6.3 with and without S9 fraction in terms of mutagenic potential in strain TA98 ([Table 5](#tbl0025){ref-type="table"}). TA98 strain displayed highest response in the mutagenic index (11.95 with and 11.97 without S9), induction factor (2.38 with and 2.39 without S9 fraction) and mutagenic potential (6.7 with and 6.8 without S9 fraction) while treated with DCM extract of acidic fraction ([Table 6](#tbl0030){ref-type="table"}). On the basis of mutagenic index and induction factor, the order of responsiveness in the presence as well as absence of S9 fraction for DCM (acidic and basic) fractions was found to be TA98\>TA97a\>TA100\>TA102\>TA104. The responsiveness order of tester strains of mutagenic potential/slope was different with mutagenic index and induction factor in DCM extracts (acidic and basic fractions). TA98 showed maximum number of revertants followed by TA100, TA97a, TA104 and TA102.Table 5Reversion of *Salmonella* tester strains in the presence of basic fraction of dichloromethane extracted wastewater.Table 5Wastewater extract (μL/plate)F valueStrainS9Control2.55102040MimLSD (*p*≤0.05)TA97a--88 ± 13200 ± 20 (2.26)267 ± 18 (3.02)320 ± 15 (3.61)397 ± 16 (4.49)302 ± 24 (3.42)1.254.226.3147.2+98 ± 14238 ± 12 (2.42)297 ± 19 (3.03)355 ± 18 (3.62)442 ± 24 (4.51)359 ± 13 (3.65)1.255.121.3213.1TA98--35 ± 10135 ± 17 (3.85)238 ± 19 (6.80)336 ± 21 (9.60)438 ± 19 (12.5)323 ± 19 (9.22)2.446.328.7184.6+39 ± 8151 ± 16 (3.88)268 ± 13 (6.88)376 ± 15 (9.65)496 ± 18 (12.72)373 ± 19 (9.56)2.467.325.9165.2TA100--156 ± 15257 ± 9 (1.65)301 ± 13 (1.93)386 ± 19 (2.47)463 ± 20 (2.97)360 ± 20 (2.31)0.674.331.4129.8+164 ± 10275 ± 17 (1.68)322 ± 21 (1.96)411 ± 13 (2.51)489 ± 25 (2.99)396 ± 10 (2.42)0.684.818.6204.3TA102--235 ± 12323 ± 15 (1.37)402 ± 11 (1.71)480 ± 17 (2.04)536 ± 15 (2.27)451 ± 26 (1.91)0.244.427.6178.1+247 ± 12351 ± 20 (1.41)431 ± 16 (1.74)507 ± 17 (2.05)573 ± 8 (2.31)475 ± 23 (1.92)0.274.619.6149.8TA104--305 ± 12366 ± 14 (1.19)428 ± 20 (1.40)493 ± 19 (1.61)555 ± 17 (1.81)470 ± 21 (1.54)−0.193.614.5167.0+314 ± 19381 ± 13 (1.21)452 ± 17 (1.44)526 ± 20 (1.67)579 ± 25 (1.84)496 ± 14 (1.57)−0.163.914.6200.5[^5]Table 6Reversion of *Salmonella* tester strains in the presence of acidic fraction of dichloromethane extracted wastewater.Table 6Wastewater extract (μL/plate)F valueStrainS9Control2.55102040MimLSD (*p*≤0.05)TA97a--84 ± 13202 ± 14 (2.40)278 ± 18 (3.31)340 ± 20 (4.04)422 ± 14 (5.02)304 ± 14 (3.62)1.393.512.7168.7+90 ± 8226 ± 19 (2.51)301 ± 23 (3.34)366 ± 16 (4.06)453 ± 17 (5.03)344 ± 17 (4.81)1.404.913.9163.1TA98--35 ± 8119 ± 12 (3.42)231 ± 10 (6.66)321 ± 15 (9.26)414 ± 14 (11.95)335 ± 19 (9.67)2.386.711.1331.2+41 ± 10145 ± 12 (3.55)280 ± 15 (6.87)379 ± 22 (9.31)487 ± 15 (11.97)394 ± 15 (9.68)2.396.812.6353.9TA100--135 ± 12219 ± 13 (1.62)305 ± 15 (2.25)417 ± 19 (3.09)519 ± 16 (3.84)391 ± 16 (2.90)1.045.912.6247.2+143 ± 11254 ± 21 (1.77)337 ± 18 (2.35)446 ± 22 (3.12)557 ± 21 (3.89)416 ± 23 (2.91)1.066.016.1165.8TA102--227 ± 10333 ± 8 (1.46)454 ± 22 (2.00)514 ± 18 (2.26)561 ± 12 (2.47)444 ± 14 (1.94)0.384.012.07206.7+241 ± 10355 ± 19 (1.47)487 ± 11 (2.02)547 ± 1 (2.27)597 ± 23 (2.48)478 ± 21 (1.98)0.394.314.5163.02TA104--275 ± 7356 ± 18 (1.29)404 ± 18 (1.47)481 ± 14 (1.75)553 ± 14 (2.0)459 ± 21 (1.67)0.014.013.02113.8+289 ± 12382 ± 15 (1.32)432 ± 19 (1.49)509 ± 17 (1.76)582 ± 17 (2.01)498 ± 22 (1.72)0.014.414.6100.9[^6] The XAD concentrated wastewater sample was observed most mutagenic compared to other extracts as evident from mutagenic index, mutagenic potential and induction factor, the values were observed at the dose of 20 μL/plate. XAD extracted sample exhibited maximum toxicity followed by Hexane and DCM extracts respectively. All the values were pointedly higher with respect to control in all strains and signifying strong mutagenicity. 3.4. *Allium cepa* chromosomal aberration assay {#sec0100} ----------------------------------------------- The genotoxic effects of wastewater on the MI and the incidence of mitotic phases of *A. cepa* on root meristematic cells are shown in [Table 7](#tbl0035){ref-type="table"}. The value of MI was significantly decreased as the concentration of wastewater increased (26.7% at 5% and 11.03% at 100% wastewater concentrations). Highest MI (30.3%) was found in negative control (distilled water) while the cells treated with MMS shows the lowest MI (8.23%). Additionally, it was also observed that occurrence of mitotic phase was affected by the treatment, as the prophase cells percentage increased and metaphase cells decreased progressively with increase in wastewater concentrations upto 100%, while no uniform pattern was observed in anaphase-telophase stage. Moreover, the meristematic cells of root by treating with wastewater also revealed distinct forms of chromosomal aberrations such as Stickiness in telophase, Multipolar anaphase, Vagrant chromosome, Anaphase with chromosome break, Disturbed metaphase, Laggard chromosome, C-mitosis, Metaphase with chromosome break ([Table 7](#tbl0035){ref-type="table"}; [Fig. 1](#fig0005){ref-type="fig"}). The aberrant cells percentage was increased on increasing the wastewater concentration. Cells treated with MMS (positive control) showed highest number of aberrations however distilled water treated cells show very rare aberrations. The statistical studies of the sample showed that MI and percent abnormal cells triggered by treating along with wastewater were significant (P \< 0.05) and relatively distinct from the positive and negative control samples by DMRT.Table 7Effect of different concentrations of wastewater on mitotic index and mitotic phase of *Allium cepa* root meristematic cells.Table 7SamplesConcentration (% v/v)Mitotic Phases (%)Mitotic index (%±SD)ProphaseMetaphaseAnaphase-TelophaseWastewater546.9722.8330.226.70 ± 1.0^b^1046.5223.0830.419.17 ± 3.7^ab^2554.5416.9628.516.67 ± 0.6^ab^5055.219.825.013.46 ± 1.4^ab^10058.8526.214.9511.03 ± 1.7^a^Positive control63.6322.7213.658.23 ± 1.6^a^Negative control49.531.7518.7530.3 ± 2.4^c^[^7]Fig. 1Normal phases and different types of chromosomal aberrations induced by the wastewater in *Allium cepa* root-tips A) Normal prophase, B) Normal metaphase, C) Normal anaphase, D) Normal telophase, E) Stickiness in telophase, F) Multipolar anaphase, G) Vagrant chromosome, H) Anaphase with chromosome break, I) Disturbed metaphase, J) Laggard chromosome, K) C-mitosis, L) Anaphase with chromosome break.Fig. 1 3.5. *In vitro* toxicity assessment of wastewater to *V. Rradiata* {#sec0105} ------------------------------------------------------------------ We also assess the toxic behaviour of wastewater on another plant i.e. *V. radiata* (mung bean), to confirm the toxicity. Under untreated condition, percent germination of seed, seedling vigor index (SVI), radicle length (RL), plumule length (PL), dry biomass of radicle (DBR) and dry biomass of plumule (DBP) was found to be 97%, 2610 SVI, 12, 15 cm, 0.25 and 0.37 gm, respectively ([Fig. 3](#fig0015){ref-type="fig"}). However, all these plant parameters were reduced as the concentration of wastewater increased. Percent germination, SVI, RL, PL, DBR and DBP were significantly decreased by 52%, 76%, 56%, 47%, 64% and 57%, respectively ([Fig. 3](#fig0015){ref-type="fig"}), when grown on soft agar plates amended with highest concentration of wastewater ([Fig. 2](#fig0010){ref-type="fig"}). The damage in the cells of root tips due to wastewater were observed and clearly visible under fluorescent microscope with red fluorescence produced by propidium iodide. The intensity of fluorescence continuously enhanced while increasing wastewater concentration ([Fig. 4](#fig0020){ref-type="fig"}).Fig. 2Dose-dependent reduction in the radicle and plumule length of mung bean seeds germinated on 0.7% agar plates amended with 10%, 25%, 50% and 100% wastewater concentrations.Fig. 2Fig. 3Plant parameters of *Vigna radiata* (mungbean) seeds germinated on soft agar plates treated with different (10, 25, 50 and 100%) concentrations of wastewater; % germination and seedling vigor index (SVI) (a), radicle and plumule length in (cm) (b) and dry biomass (c). Each value is a mean of five independent replicates (n = 5) where each replicate constituted five seeds/plates. Mean values followed by different letters are significantly different at *p≤*0.05 according to Tukey's-b test. Vertical bars represent means ± SD (n = 5).Fig. 3Fig. 4Confocal laser scanning microscopic (CLSM) images of *Vigna radiata* roots stained with propidium iodide (PI) and treated with different concentrations of wastewater; control (a), treated with 10% (b), 25% (c) 50% (d) 100% (e) and MMS (positive control) (e). As the treatments of wastewater increased, the uptake of dye (PI) also increased.Fig. 4 3.6. Plasmid nicking assay {#sec0110} -------------------------- [Fig. 5](#fig0025){ref-type="fig"} presents the DNA band profiles obtained after the pBR322 plasmid nicking assay with a series of wastewater concentration. The test volume of 5 to 20 μl of the wastewater sample in a 25 μl reaction mixture resulted in the conversion of pBR322 DNA from supercoiled form into the open circular ([Fig. 5](#fig0025){ref-type="fig"}; lane a). The intensity of open circular form of plasmid was increased on increasing the wastewater concentration and the band intensity of supercoiled form was decreased to complete loss of supercoiled form ([Fig.5](#fig0025){ref-type="fig"}; lane b--e). The maximum conversion of supercoiled to open circular was observed in pBR322 plasmid treated with MMS ([Fig. 5](#fig0025){ref-type="fig"}; lane f).Fig. 5Plasmid-nicking assay conducted on the wastewater-samples. Lane m: 1 kb ladder Lane a pBR322 DNA alone. Lane b-e: pBR322 DNA + 5 μl, 10 μl, 15 μl, 20 μl of wastewater respectively. Lane f: pBR322 DNA + MMS.Fig. 5 4. Discussion {#sec0115} ============= Ghaziabad district is one of the key industrial hubs in Northern region of India, the location of wastewater sampling site is 28°44′N and 77°17′E, and a number of industries including the pesticide industries, that produce vast quantity of wastewater that are steadily discharged into the river. Physico-chemical parameters employed to observe quality of water were TDS, carbonate, bicarbonate, chloride and sulphate were above the permissible limits as defined by USEPA ([Table 1](#tbl0005){ref-type="table"}). Increase in the TDS is also indicating the pollution level of water that marks the self-purification process of the wastewater and also harmful for aquatic animals due to osmotic stress \[[@bib0170]\]. In addition to physico-chemical analysis, toxicity evaluation of wastewater is of applied significance as it would support in expecting the collective effects of diverse compounds into the water. Various heavy metals existing in the wastewater demonstrated deleterious influence on the environment and health of humans \[[@bib0040]\]. In our previous studies, we have reported the existence of high concentrations of several metals in wastewater and contaminated soil \[[@bib0140],[@bib0175]\]. The detection of specific organic substances along with mutagenic activity in untreated wastewater or even in effluents of industries is quite problematic, because only a very limited compounds are found in the detectable limit. Environmental pollution of wastewaters by residues of pesticide is of great concern. Insecticides are a group of organic compounds that shows broad range of toxic effects and ultimately cause a potential threat to the environment \[[@bib0180],[@bib0185]\]. EI-Gawad \[[@bib0190]\] detected several pesticides of organochlorine group such as Alpha-BHC, Gamma-BHC, Aldrin, Heptachlor, Heptachlor epoxide, in water samples at high concentrations. Toxicity of Cr and Ni are reported in lipid peroxidation, generation of reactive oxygen species (ROS), oxidative stress and DNA damage \[[@bib0195]\]. High level of organochlorine and organophosphate pesticides ([Table 2](#tbl0010){ref-type="table"}) are reported in past \[[@bib0055],[@bib0180]\]. Bedi et al \[[@bib0200]\] reported persistent organic pollutants containing lindane, DDE, DDD, endosulfan sulfate as well as polychlorinated biphenyl in the fish sample. In Nigeria, the surface water of fifteen different sites of two river were evaluated for the quantification of twenty organochlorine pesticides \[[@bib0205]\]. They also detected the organochlorine pesticides in brackish fish (*Drepane africana* and *Mochokus niloticus*) samples of the Niger River with concentration range of 2237--6368 μg/kg of fresh weight. The present study showed the genotoxic, cytotoxic and mutagenic potential of the wastewater. The combined effect of cytotoxicity and genotoxicity including bacterial (prokaryotic) and plant (eukaryotic) entity, were performed to obtain a thorough impact of wastewater on the environment. These analyses are significant for the assessment of harmful waste and threat calculation correlated with contaminants \[[@bib0210], [@bib0215], [@bib0220]\]. A huge number of mutagens were extracted in different organic solvents (dichloromethane, n-hexane, ethyl acetate, acetone, acetonitrile etc.) and were identified including aromatic amines, polycyclic aromatic hydrocarbons, polychlorinated compounds \[[@bib0105]\]. The industrial effluents comprising diverse range of chemicals that have been found to be genotoxic and responsible for various stages of DNA damages in the organisms of aquatic system \[[@bib0085]\]. Due to complex substances present in wastewater samples, a single test cannot assess all the mode of toxicity in the samples that are mixtures of contamination \[[@bib0225],[@bib0230]\] In the Ames *Salmonella*/microsome assay the tester strains contain a certain alteration in 'histidine operon' (i.e., TA97a / TA98 frameshift mutations, strain TA100 base pair substitution / missense mutations and TA102 / TA104 transitions / transversions) and hence distinguish a particular form of mutagen \[[@bib0100],[@bib0235]\]. Rehana et al \[[@bib0240]\] also used five different Ames *Salmonella* bacterial strains to test genotoxicity of Ganges river water at different sites and observed that TA98 and TA100 displayed high mutagenicity with and without S9 fraction. Numerous workers observed that in XAD concentrated extracts, TA98 strain was more responsive compare to TA100 in both with and without S9 fraction, moreover extracted concentration of XAD was also more mutagenic than the samples of liquid-liquid extraction, as reported by several workers \[[@bib0040],[@bib0245]\]. Wastewater discharge from industrial area of Lucknow (pesticide industry) is used for irrigation purposes and the soil irrigated with wastewater showed strong mutagenic activity in comparison to soil irrigated with ground water \[[@bib0250]\]. The life cycle of *A*. *cepa* root meristematic cells is short (20 h) and contains smaller number of chromosomes (2n = 16) compare to other plants. That's why it is preferable eukaryotic plant system for the evaluation of damages in chromosomes \[[@bib0150]\]. Cytotoxicity and genotoxicity of wastewater were assessed by detecting different cytological parameters for example mitotic index and aberrations in chromosomes including breaks in chromosomes, laggard chromosome, C-mitosis, anaphase bridges and stickiness. The two important parameters were used i.e. reduction in mitotic index and increase in chromosomal aberration for assessing genotoxicity and cytotoxicity of numerous compounds present in the test samples \[[@bib0255]\]. Many earlier reports have confirmed that *A. cepa* and mammalian test systems has good correlation \[[@bib0255],[@bib0260]\]. In this study we observed a significant reduction in mitotic index as concentration of wastewater increased in comparison of the control ([Table 7](#tbl0035){ref-type="table"}), this is due to high level of trace elements in single or in combination of other metals have inhibitory effect on cell division \[[@bib0265], [@bib0270], [@bib0275]\]. The wastewater is responsible for decline in mitotic index in roots of *A. cepa* due to toxicity of pesticides, heavy metals and many other pollutants, ultimately cell death occurred \[[@bib0280], [@bib0285], [@bib0290]\]. The chromosomal aberrations are direct indicative of DNA damage that could not be easily repaired \[[@bib0295]\]. In the present study, several types of aberrations in the chromosomes were observed i.e. C-mitosis, disturbed metaphase and vagrant chromosomes being most distinguished ([Table 8](#tbl0040){ref-type="table"}; [Fig. 1](#fig0005){ref-type="fig"}). C-mitosis is occurred due to the spindle disturbance in mitotic phase \[[@bib0300]\]. The vagrant chromosomes observed because of failure of chromosomal separations in the stage of metaphase \[[@bib0305]\] and risk of aneuploidy increased \[[@bib0255]\]. Thus, the occurrence of several type of aberrations in chromosomes (C-mitosis, disturbed metaphase, stickiness, vagrant chromosomes etc.) in the meristematic cells of *A. cepa* root could be attributed to collective effect of clastogenic as well as aneugenic actions of several compounds in the wastewater \[[@bib0310]\].Table 8Chromosomal aberrations in the root meristematic cells of *Allium cepa* exposed to different concentrations of wastewater for 72 h.Table 8SampleConcentration (% v/v)Types of aberrationsTotal aberrant cells (%±SD)VCCMLCMADMSCABDATWastewater5212----11--5.53 ± 0.72^e^10223--22--110.10 ± 1.850^d^25464--335320.15 ± 1.56^c^505287615428.98 ± 4.77^c^100107448110734.26 ± 2.98^b^Positive control1312103512111440.97 ± 2.66^a^Negative control3--1--2--125.82 ± 0.32^e^[^8] The cell membrane is one of the important, selectively permeable organelles that controls the exchange of ions and molecules and permit the cells to communicate with the neighbouring environment. CLSM has proved to be the most delicate and reliable technique for obtaining a 3D image of basic tissue. Seed germination is dynamic phenomenon during the life cycle of plants therefore, SG and SVI is considered as the most significant physical characteristics of seeds that are used for cultivation. In this context, delayed germination following the application of wastewater has been associated with disturbed germinative metabolism which is a complex process. The toxic impact of pesticide on the germination efficiency of *Dimorphandra wilsonii*, belongs to Fabaceae family has been reported \[[@bib0315]\]. Similarly, the reduced length of radicle and plumule in germinated seeds of pea due to the toxic influence of another environmental stressor molecule (pesticide) has recently been reported \[[@bib0165]\]. The breakage in the length of DNA strand is a significant aspect to evaluate mutagenic impact of several chemical substances on integrity of DNA. The damage or break in DNA is caused by an exogenous agent or it may be formed in the repair processes of DNA, or physiologic responses into the cell \[[@bib0320]\]. In this study, the plasmid nicking assay also revealed genotoxic and mutagenic potential of wastewater. ([Fig. 5](#fig0025){ref-type="fig"}). 5. Conclusion {#sec0120} ============= The physico-chemical, GC and AAS analysis of wastewater revealed numerous genotoxic substances in the form of organic and inorganic pollutants which are directly or indirectly harmful for ecosystem and human health. A set of bacterial and plant-based tests demonstrated that the wastewater showed mutagenicity and genotoxicity by reverting the *Salmonella* tester strains (TA97a, TA98, TA100, TA102 and TA104). Strain TA98 showed highest response in the terms of induction factor, mutagenic index as well as mutagenic potential. The industrial wastewater also comprised of phytotoxic and cytotoxic substances, that's why decrease in mitotic index occurred and caused different forms of chromosomal abnormalities in meristematic cells of *A. cepa*. The effect of wastewater on *V. radiata* showed decreased percent seed germination, reduced length (radicle and plumule) and uptake of propidium iodide as observed under CLSM. Furthermore, the wastewater also induced damage in the naked DNA in plasmid nicking assay. As evident by an array of cytotoxic and genotoxic assays, it is recommended that the effluents from the industries should be treated appropriately to minimize the presence of the genotoxic and cytotoxic compounds before entering into the river system. Declaration of Competing Interest ================================= The authors declare that they have no conflict of interest. The authors would like to thanks Advanced Instrumental Research Facilities (AIRF), JNU, New Delhi for GC analysis and University Grant Commission (UGC), Government of India, New Delhi, India for providing UGC Non-NET Fellowship. [^1]: All parameters except pH are in mg/L. [^2]: ND = not detected. [^3]: Values in parentheses are mutagenic index; Mi = induction factor; m = mutagenic potential; LSD = least significant difference. [^4]: Values in parentheses are mutagenic index; Mi = induction factor; m = mutagenic potential; LSD = least significant difference. [^5]: Values in parentheses are mutagenic index; Mi = induction factor; m = mutagenic potential; LSD = least significant difference. [^6]: Values in parentheses are mutagenic index; Mi = induction factor; m = mutagenic potential; LSD = least significant difference. [^7]: Means with the same letters do not significantly differ at 0.05 level (Duncan multiple range test); **±**: Standard deviation. [^8]: CM: C-mitosis, AB: anaphase bridge, LC: laggard chromosome, BN: binucleated cell, S: stickiness, DM: disturbed metaphase DAT: disturbed anaphase-telophase, VC: vagrant chromosome; MA: multipolar anaphase; Means with the same letters do not significantly differ at 0.05 level (Duncan multiple range test); **±**: Standard deviation.

INTRODUCTION ============ Diabetic neuropathies are neuroendocrine disorders related to chronic hyperglycemia, affecting nerve fibers of the body part in approximately half of all diabetes patients ([@B1][@B2]). The condition of insulin secretory function, insulin sensitivity, insulin resistance (IR), and subsequent hyperinsulinism play a vital role in the pathogenesis of the diabetic complication and the development of type 2 diabetes mellitus (T2DM) ([@B3][@B4]). Multifactorial interventions, dyslipidemia management including strict glycemic control are known to diminish the progression but cannot remove the risk of cardiac autonomic neuropathy (CAN) in patients with T2DM ([@B5][@B6][@B7]). High levels of plasma triglycerides (TGs) are a risk marker for cardiovascular disease (CVD), and it has often been related to impaired fasting glucose, T2DM, and diabetic complication ([@B8][@B9][@B10]). Thus, much stress had been placed on controlling hyperglycemia along with hypertension and hyperlipidemia to prevent and delay the onset of diabetic complications ([@B11]). Recently, the triglyceride glucose index (TyG index), has been suggested in several studies as a simple assessment method for metabolic abnormality ([@B12][@B13][@B14]). This index is the mathematical product of the fasting blood glucose and TG levels, which correlates with the degree of IR ([@B15]). Recently a report revealed that dyslipidemia and hyperglycemia contribute to the development of CAN ([@B16]). However, data on the direct relationship between indices and peripheral neuropathy in T2DM are lacking, and there is no comparative study regarding IR indices among patients with CAN and T2DM. The majority of studies on the prevalence and associated risk factors of CAN are executed in Western countries in the past. To the best of our knowledge, none of the previously published studies investigated the TyG index on the presence of CAN with regard to T2DM in Indian population. In this study, we aimed to assess the role of IR indices like TyG index in the development of CAN along with the prevalence of CAN in Indian T2DM patients. MATERIALS AND METHODS ===================== Study design ------------ It was a prevalence study conducted in a tertiary care setting hospital of Jamia Hamdard. This investigation included 202 patients treated in May 2015 to November 2016. Based on Ewing autonomic function test and symptoms, the patients were divided into 2 groups. The first group of 62 subjects consisted of T2DM patients with CAN and was considered as CAN patients group while the other group consisted of 140 T2DM patients without CAN and was considered as non-CAN patients group and 9 patients withdrew from the study due to medical expenses. The study was conducted in the department of medicine, with both outdoor and indoor patients and patients attending the Hakeem Abdul Hameed Centenary Hospital (HAHC) associated with Hamdard Institute of Medical Sciences and Research, Jamia Hamdard (HIMSR). Included subjects were of either sex, above 18 years of age, and had the diagnosis of T2DM. Exclusion criteria were heart failure, urinary tract infection, fever, cirrhosis of liver, and prostatitis. Patients were asked to avoid smoking, consumption of coffee, and tea before assessment. All patients had a complete history of neurological symptoms using a semi-structured questionnaire which was documented. Symptoms like numbness, asleep feeling, burning sensation, deep aching, unsteadiness in walking, unexplained resting tachycardia, postural fainting, orthostatic hypotension, sweating, ejaculation problems, and other symptoms documented in previous literature were considered for neuropathy symptoms. Manometer, handgrip dynamometer, mouth piece, and electrocardiography (ECG) machine were used during the autonomic function test. Clinical feature measurement ---------------------------- Blood pressure (BP) of the participants was measured at the time of recruitment in the sitting position in the right arm by sphygmomanometer (Diamond Deluxe BP apparatus; BP Instruments, Pune, India). Body weight was measured on the same scales in light clothing and no shoes before breakfast, and upright height was measured on the same wall-mounted stadiometer. Individual body mass index (BMI) was then calculated as weight (kg)/height (m^2^). The right-arm BP of each seated subject was obtained after 10 minutes of rest using a sphygmomanometer. A brief prescription detail, ongoing medication, and medical history were recorded for all subjects. A thorough general and physical examination including the vital data was done for all the subjects. Blood was collected in a fasting state for plasma glucose, glycated haemoglobin (HbA1c), lipid profile, and another clinical variable. Quantitative estimation of HbA1c was done by using high performance liquid chromatography (HPLC). Retinal conditions were assessed by ophthalmologists using a combination of clinical examination. Neuropathy assessment --------------------- Neuropathy testing was performed by the experienced physician and technician according to standardized procedures. Neuropathy was assessed using monofilament, pinprick sensations, ankle reflexes, and vibration perception threshold (VPT) test. Quantification of peripheral neuropathy was assessed by VPT using a Biothesiometer (Proactive Health Inc, Ahmedabad, Gujarat, India) by a single observer for all the participants. VPT was then measured at 5 different locations on the both legs. Monofilament of 10 g is a simple instrument that is used frequently in assessing the loss of protective sensation. It is adjusted in such a way that it takes 10 g of force to buckle. During this test the monofilament was placed perpendicular to the skin site of each foot in such a way that it would not slip and pressure was applied to the filament just buckled with a contact time of 2 seconds. Inability to perceive the sensation felt at any one site of each leg was considered loss of protective sensation. In addition, ankle reflexes were also assessed with a percussion hammer, and recorded as either present or absent. All tests were conducted on a quiet workbench. Autonomic function test ----------------------- Ewings autonomic function test was performed to confirm the diagnosis of CAN. Out of 5 autonomic function tests, 3 of these measurements mainly assessed parasympathetic function heart rate responses to deep breathing (beat-to-beat variation), to standing (30:15 ratio), and to the Valsalva manoeuvre. Sympathetic function was assessed by BP response during posture change and handgrip test ([@B17]). Deep breathing test ------------------- The patient laid quietly and breathed deeply at a rate of 6 breaths (expiration and inspiration) per minute and heart rate along with the time elapsing between 2 consecutive R waves in the electrocardiogram (R-R interval) was recorded by ECG. The ratio of R-R interval during expiration and inspiration was considered as E:I ratio. The result was then expressed as the mean of the difference between maximum and minimum heart rates for the 6 measured cycles in beats/minute and considered as abnormal response if it is \< 10 beats/minute. Heart rate response to standing ------------------------------- The patient laid quietly and stood up unaided; the heart rate was recorded continuously by ECG. The '30:15 ratio,' which is the ratio of the longest R-R interval (around the 30th beat after starting to stand up) to the shortest R-R interval (around the 15th beat), was then calculated and considered as normal if 30:15 ratio was above 1.04. Valsalva manoeuvre ------------------ The patient forcibly exhales into the mouthpiece of a manometer, exerting a pressure of 40 mmHg for 15 seconds. ECG was recorded continuously in supine position in during the whole process. The longest and shortest R-R intervals were measured. The mean of 3 Valsalva ratios taken as the final value and considered as normal if the Valsalva ratio above than 1.20 Sympathetic function test ------------------------- ### BP response to standing BP was measured while the patients were active in standing position and again when in a supine position. The postural fall in BP was taken as the difference between systolic BP lying and just after standing. ### BP response to handgrip test It consists of variation in BP during handgrip test. After instructions in using handgrip dynamometer, the subject gripped maximally with his dominant arm for 5 minutes, this was repeated thrice. The highest of the 3 readings is called maximum voluntary contraction (MVC) and considered as abnormal if diastolic blood pressure (DBP) \< 10 mmHg. Patients were classified as CAN positive if any of the above-mentioned tests found abnormalities. Further, CAN patients were classified as the early CAN if 1 of 3 heart rate tests were abnormal, definite CAN if 2 heart rate tests were abnormal and severe CAN if 2 heart rate tests were abnormal + 1 or both BP tests were abnormal ([@B17][@B18]). Later during analysis process non-CAN were specified as stage 0, early CAN as stage 1, definite CAN as stage 2, and severe CAN as stage 3. ### Calculation of TyG index TyG index was measured with fasting blood glucose and fasting TG level. TyG index were calculated: TyG index = Ln \[fasting TGs (mg/dL) × fasting glucose (mg/dL)/2\] Statistical analysis -------------------- The study variables contained both categorical and continuous variables. Frequencies with proportions were represented the categorical variables and scale variables as mean ± standard deviation (SD). All continous variables were tested for normality using the Shapiro-Wilks test. Categorical variables and continuous variable were compared using the χ^2^ test and Student\'s t-test. Kaplan-Meier survival curves were plotted to assess the time taken for CAN development and compared using the log-rank test. The 2-sided critical region with *P* ≤ 0.05 was considered as statistical significance. All analysis was performed using SPSS ver. 21.0 statistical software (IBM Corp., Armonk, NY, USA). Ethics statement ---------------- The study was conducted in compliance with the declaration of Helsinki and approved by Institutional Review Board (IRB-Approval number-JHIEC-2015(05/15), inform consent: present) of Jamia Hamdard, Hamdard University, New Delhi, India. Written informed consent was obtained from each patient prior to inclusion in the study and for using their health-related data. RESULTS ======= Characteristics of participants ------------------------------- In total, 362 subjects diagnosed with T2DM visited the study centre. A total of 202 patients aged 18--80 (range) years who fulfilled the inclusion criteria were enrolled. The flow diagram of participant inclusion is shown in [Fig 1](#F1){ref-type="fig"}. Comparison of baseline data of the non-CAN and the CAN patients group are shown in [Table 1](#T1){ref-type="table"}. Demographic and clinical data were collected as follows: age, gender, systolic blood pressure (SBP), DBP, weight, height, and BMI. Weight and height were measured by balance beam scale and tape measure. The mean BMI, SBP, DBP, TG, cholesterol, and fasting blood glucose levels were significantly higher in the CAN patients group (*P* \< 0.001). According to autonomic function test results 140 patients (69.30%) out of 202 T2DM patients had 2 scores and categorized as T2DM without CAN, 7 patients (11.29%) out of 62 had 4 scores, were considering as early CAN while 24 (38.70%) out of 62 were considered definite and 31 (50%) patients out of 62 had more than 5 scores which were categorized as the severe CAN. Over all prevalence of CAN was 30.7% (62 cases), however, it was less than earlier prevalence report conducted in India ([@B19]). We also assessed the neuropathy condition in the patients by various tests including monofilament and VPT. Based on cut-off range of Biothesiometer (Proactive Health Inc), score of monofilament test and other symptoms, there were 29 patients in the CAN who had diabetic peripheral neuropathy (DPN) simultaneously. Out of these DPN patients in CAN, one was diagnosed with diabetic foot. Further more, we found that prevalence of microvascular complication (retinopathy, nephropathy, and DPN) in CAN was significantly higher with respect to non-CAN patients group. In addition, TyG index levels did not differ significantly depending on the presence or absence of DPN (TyG index in DPN it was 10.12 vs. 9.79, *P* = 0.06, in CAN and non-CAN, respectively). Variation in heart rate during deep breathing (HRD), heart rate variation during standing (HRS), BP variation during posture change and BP variation in the hand grip test were statistically significant when compared to non-CAN with CAN patients group as shown in [Table 2](#T2){ref-type="table"}. There was a significant difference in medication modalities (insulin, oral hypoglycemic agent, and angiotensin-converting enzyme inhibitor \[ACEI\]/angiotensin receptor blocker \[ARB\] intake) among CAN patients and T2DM patients group. Smoking habit was obtained by questionnaire and patients were divided into smokers (present or former) and non-smokers. Out of 202 patients, 67 patients were smokers. Smoking was significantly associated with CAN (*P* \< 0.001). The SBP and DBP was significantly (*P* \< 0.001) associated with CAN. {#F1} ###### Clinical and laboratory profiles of all subjects  Variables T2DM without CAN (n = 140) T2DM with CAN (n = 62) *P* value ---------------------- ---------------------------- ------------------------ -------------- Sex (male:female), % 61:39 33:67 0.317 Age, yr 51.3 ± 10.8 55.7 ± 10.0 0.005 Duration, yr 7.0 ± 3.1 8.6 ± 4.1 0.002 Smoker 36 (26.4) 31 (51.6) \< 0.001^\*^ Nephropathy 13 (9.1) 21 (33.9) \< 0.001^\*^ Retinopathy 7 (5.00) 19 (30.64) \< 0.001^\*^ DPN 12 (8.5) 29 (46.8) \< 0.001^\*^ FG, mmol/L 10.8 ± 2.8 12.9 ± 3.5 \< 0.001 HbA1c, % 8.5 ± 1.0 10.4 ± 1.3 \< 0.001 BMI, kg/m^2^ 25.2 ± 1.8 25.8 ± 1.6 \< 0.001 SBP, mmHg 137.0 ± 8.2 133.5 ± 8.7 \< 0.001 DBP, mmHg 83.2 ± 5.6 79.3 ± 7.9 \< 0.001 TC, mmol/L 5.60 ± 0.60 5.50 ± 0.57 0.090 HDL, mmol/L 1.30 ± 0.30 1.30 ± 0.16 NS TG, mmol/L 1.70 ± 0.31 2.90 ± 0.46 \< 0.001 TyG index 9.5 ± 0.2 10.3 ± 0.2 \< 0.001 Medication modality  OHA 76 (54.3) 56 (90.4) \< 0.001^\*^  PPI 34 (24.3) 32 (51.2) 0.071^\*^  Insulin 25 (17.8) 20 (32.2) \< 0.001^\*^  Statin 34 (24.3) 18 (29.0) 0.477  Beta-blocker 4 (2.8) 3 (4.8) 0.478  ACEI/ARB 47 (33.5) 42 (67.7) \< 0.001^\*^  TCA 13 (9.2) 7 (11.3) 0.660 Values are mean ± SD or number (%). Data was analyzed using student\'s t-test. T2DM = type 2 diabetes mellitus, CAN = cardiac autonomic neuropathy, DPN = diabetic peripheral neuropathy, FG = fasting glucose, HbA1c = glycated haemoglobin, BMI = body mass index, SBP = systolic blood pressure, DBP = diastolic blood pressure, TC = total cholesterol, HDL = high-density lipoprotein, TG = triglyceride, TyG = triglyceride glucose, OHA = oral hypoglycemic agent, PPI = proton pump inhibitor, ACEI = angiotensin-converting enzyme inhibitor, ARB = angiotensin receptor blocker, TCA = tricyclic antidepressant, NS = not significant, SD = standard deviation. ^\*^Intergroup *P* value were analyzed by χ^2^ test. ###### Autonomic function tests in patients with CAN and non-CAN  Variables T2DM without CAN (n = 140) T2DM with CAN (n = 62) *P* value ---------------------------------------- ---------------------------- ------------------------ ----------- HR response to deep breath 16.10 ± 0.80 9.98 ± 0.70 \< 0.001 HR response to standing (R-R ratio) 1.10 ± 0.00 0.96 ± 0.00 \< 0.001 HR response to Valsalva 1.3 ± 0.0 1.2 ± 0.0 NS BP response to standing (SBP decrease) 4.7 ± 0.8 7.5 ± 0.8 \< 0.001 BP response to handgrip (DBP increase) 16.0 ± 2.7 9.3 ± 0.9 \< 0.001 Significance level, *P* \< 0.05. CAN = cardiac autonomic neuropathy, T2DM = type 2 diabetes mellitus, HR = heart rate, BP = blood pressure, SBP = systolic blood pressure, DBP = diastolic blood pressure, NS = not significant. Association of BMI, age, duration, and gender --------------------------------------------- Mean BMI in our study was found 25.8 ± 1.6 kg/m^2^, 25.2 ± 1.8 kg/m^2^ in CAN and non-CAN patients group, respectively. The mean age of subjects was 55.7 ± 10.0 and 51.3 ± 10.2 years in CAN and non-CAN patients group, respectively. Further, we found 3.2% (2 of 62) CAN patients were below 40 years, 22.58% (14 of 62) CAN patients were between 40--49 years of age group, 27.40% (17 of 62) CAN patients were between 50--59 years of age group, 38.70% (24 of 62) CAN patients were between 60--69 years of age group and 8.06% (2 of 62) CAN patients were ≥ 70 years of age group. Diabetes duration appeared to be longer in those with CAN group. Forty-three point five four percent (27 of 62) CAN patients were ≤ 5 years of duration of T2DM, 35.48% (22 of 62) CAN patients were in the range of 6--10 years of duration of T2DM, 16.12% (10 of 62) CAN patients were in the range of 11--15 years of duration of T2DM and 4.8% (3 out of 62) CAN patients were ≥ 15 years of duration of T2DM. Comparison of mean age at diagnosis of diabetes and duration took for the CAN development made. No significant differences were observed either of sex. Kaplan-Meier analysis ([Fig. 2](#F2){ref-type="fig"}) were plotted to support the previous result as shown in [Table 3](#T3){ref-type="table"}. Prevalence of CAN in the female patients was found approximately 2 fold than male. {#F2} ###### Sex differences in the age at diagnosis of T2DM (non-CAN) and time gap for CAN development  Variables Female Male *P* value ----------------------------------------- ------------- ------------- ----------- Age at the time of diagnosis of non-CAN 43.3 ± 10.7 46.0 ± 9.0 0.145 Age at the time of diagnosis of CAN 55.6 ± 9.6 56.0 ± 11.1 0.622 Time gap between non-CAN and CAN 12.3 ± 0.1 10.0 ± 2.1 0.065 The duration taken for the development of CAN (years). It was calculated by subtracting the age at diagnosis of T2DM (non-CAN) from the age at diagnosis of CAN. Data presented as mean ± SD. T2DM = type 2 diabetes mellitus, CAN = cardiac autonomic neuropathy, SD = standard deviation. Association of TyG Index with autonomic neuropathy -------------------------------------------------- As demonstrated in [Table 1](#T1){ref-type="table"}, the TyG index was significantly correlated with CAN (*P* \< 0.001). The mean ± SD for TyG was 10.3 ± 0.2 and 9.5 ± 0.2 in CAN and non-CAN patients, respectively. The difference of TyG index, in CAN and non-CAN patients group was highly significant (*P* \< 0.001). Furthermore, we also find any correlation between TyG index and types of CAN. A significant difference was observed when TyG index of early, definite and severe CAN was compared with T2DM group separately ([Table 4](#T4){ref-type="table"}). ###### The relation of TyG index to category of CAN  Stage Category No. of patients TyG index *P* value ------- ------------------ ----------------- ------------ -------------- 0 T2DM without CAN 140 9.5 ± 0.2 \- 1 Early CAN 7 10.2 ± 0.4 \< 0.031^\*^ 2 Definite CAN 24 10.3 ± 0.1 \< 0.049^†^ 3 Severe CAN 31 10.3 ± 0.3 \< 0.034^‡^ T2DM with CAN specified as stage 0, T2DM with early CAN as stage 1, T2DM with definite CAN were specified as stage 2, T2DM with severe CAN as stage 3. TyG = triglyceride glucose, CAN = cardiac autonomic neuropathy, T2DM = type 2 diabetes mellitus. ^\*^*P*, stage 1 vs. stage 0; ^†^*P*, stage 2 vs. stage 0; ^‡^*P*, stage 3 vs. stage 0. Pearson\'s correlation analysis ------------------------------- A significant association was found between autonomic function parameter and clinical variables ([Table 5](#T5){ref-type="table"}). The TyG index shows significant correlation with autonomic function test (*P* \< 0.001). Similarly, age factor found to be positively associated with BP variation during the hand grip test. Further, we found that deep breathing also significantly associated with BP variation during posture change. ###### Univariate correlation between cardiac autonomic function tests and TyG index in patients  Variables HRD HRS Valsalva BPS BPH ----------- -------- ---------- ---------- ---------- -------- ---------- -------- ---------- ------- ---------- TyG index 0.524 \< 0.010 0.504 \< 0.010 −0.547 \< 0.010 0.586 \< 0.010 0.431 \< 0.010 Age −0.201 \< 0.034 −0.150 \< 0.022 −0.161 \< 0.022 −0.176 0.012 0.275 0.001 Duration −0.126 0.075 −0.103 0.144 −0.192 −0.006 0.159 0.006 0.123 0.080 Correlation result with autonomic function test. TyG = triglyceride glucose, HRD = heart rate during deep breathing, HRS = heart rate variation during standing, BPS = blood pressure standing, BPH = blood pressure handgrip. Multivariate regression analysis -------------------------------- Results of multivariate linear regression are shown in [Supplementary Table 1](#S1){ref-type="supplementary-material"} and expressed after adjustment with other variable as standardized coefficient β and *P* value. The TyG index (continuous variable) of the participants as independent risk factor was significantly associated with CAN (β = 0.389; *P* = 0.005). The autonomic function test like heart rate response to standing (continuous variable) of the participants showed a slightly significant association with CAN (β = −0.236; *P* = 0.061). Furthermore, retinopathy (categorical variable) of the participants showed a significant association with CAN (β = −0.310; *P* = 0.025). DISCUSSION ========== We investigated whether TyG index is associated with CAN in patients with T2DM. The novel insight of the present study shows that TyG index is potentially elevated in CAN with respect to non-CAN patients group. Our study report shows that there is a significant difference in the TyG index in CAN and non-CAN patients group. In this study, we also found that TyG index significantly increases in early, definite and severe CAN with respect to non-CAN patients group. The results from this study suggest that product of fasting glucose and TG has a significant association with the CAN patients group. Although the pathophysiology regarding the role of IR in diabetic neuropathy is complex, mitochondrial dysfunction has been stated to play a pivotal role in the pathology of IR ([@B20]). In addition to that, one of the investigations reported that diabetic parasympathetic neuropathy affects the IR in type 2 diabetic patients ([@B21]). It has the potential to be a simple clinical marker of the metabolic syndrome. The finding of the present study could be interesting in search for a sensitive biomarker for diabetic neuropathy and a low-cost IR index (TyG index) may play a vital role in predicting the disease course. Further, we found that there was a significant difference in medication modalities (insulin, oral hypoglycemic agent, and ACEI/ARB intake) among CAN and non-CAN patients group. Hyperlipidemia develops in the course of diabetes mellitus, and the late development of an abnormal lipid profile coincides with the delayed encounter and progression of diabetic neuropathy ([@B22][@B23]). IR has been observed as a pathogenic component in diabetic neuropathy for a decade. Longitudinal Rochester Study also supports the role of IR in diabetic neuropathy ([@B24]). One of the experimental studies also revealed that insulin deficiency is a major component in the diabetic neuropathy due to neurotrophic effects of insulin ([@B25]). Earlier investigational reports suggest that age, duration of diabetes, and metabolic control has a significant association with CAN progression ([@B26][@B27]). Our study also suggests that age and duration of diabetes was strongly associated with CAN. The results showed that the prevalence of CAN was 30.7% however this was lesser as reported in previous study conducted in India ([@B19]). Variation in prevalence widely seen in such type of study, which may be due to population studied, types of diabetes and methodology followed for assessment. The levels of HbA1c, high blood glucose were relatively high among CAN and non-CAN patients group in our study because of patients with long standing diabetes. The incidence of development of retinopathy, nephropathy, and peripheral neuropathy (Triopathy) was relatively high among the CAN patients group. Findings of earlier investigations suggest the high percentage of nephropathy, peripheral neuropathy, retinopathy were found in the study population ([@B27][@B28]). In our study, there was no significant difference in the incidence of the development of insulin resistant among the different patterns of CAN. In addition to that, no significant differences were observed among the CAN with or without presence of other micro vascular complication. Furthermore, we performed the Pearson\'s correlation analysis and multivariate linear regression analysis which suggest the positive correlation of TyG index with autonomic function parameter. In addition to that, in our study, besides the CAN evaluation parameter, age, duration was also independently and significantly a risk factor for cardiovascular risk in autonomic neuropathy. TyG index, a measure of IR was found to be independently and significantly associated with autonomic neuropathy. Previous reports also revealed the high IR indices, age, and other clinical risk factor levels among autonomic neuropathy and diabetes mellitus ([@B22][@B28][@B29]). Screening parameters specific to nerve related dysfunction in T2DM are not studied much. Limited literature data to autonomic neuropathy are available where the emphasis on an alternative tool for CAN diagnosis was investigated. The predicting ability of TyG index has been studied in various metabolic disorders. This association between atherogenic lipoprotein abnormalities and the development of diabetes was described earlier, which reported that changes in the TyG index over time altered the incidence and risk of diabetes ([@B30]). Due to variation in ethnicity and other factors for determining cut-off range of TyG index ([@B31]), we have set a cut-off point for TyG index for diabetes risk (TyG index 9) after observing the variation of TyG index in non-CAN and CAN patient groups ([@B32]). However, the TyG index was more than the cut point in all T2DM and CAN group. TyG index in the various category of CAN group was significantly high in comparison to T2DM group. However, no significant difference in TyG index was observed when compared to either sex in the patients cohort. Various prospective studies have reported that IR predicts incident cardiac disease in both the general and diabetic subjects ([@B33][@B34][@B35][@B36]). The TyG index has been proposed as a surrogate of IR, in the hyperinsulinemic-euglycemic clamp test and with the homeostasis model assessment of insulin resistance (HOMA-IR) ([@B37]). The utility of the TyG index to early identify individuals at risk of diabetes was well described in the past ([@B38]). Hence in the present study, we have explored the association of TyG index in CAN. The ominous impact of long-term poor glycemic control on the development and progression of CAN is now generally accepted. One of the studies has revealed that fasting TGs is significantly associated with diabetic neuropathy ([@B39]). The prevalence of hypertriglyceridemia has increased and become a major concern with a growing population with obesity, metabolic syndrome, and T2DM. The previous study has been done with limited sample size focussed on the IR and metabolic disorder including CAN ([@B32][@B40]). In these circumstances TyG index, is a low-cost alternative that may be a more suitable clinical marker for the diabetes-related metabolic disorder. Apart from the enthusiastic result, there are some limitations of the present study. It is very difficult to establish the causative relation between CAN and TyG index due to insufficient data as we could not collect the insulin secretion, food habit, and energy consumption. We could not go through plasma catecholamine or its metabolite for assessing sympathetic autonomic activity. Despite limitations, the present study demonstrates that TyG index is significantly associated with CAN patients. Our study result revealed that TyG index significantly increased in CAN patients. TyG index has potential impact in neuropathy condition of T2DM patients. Further investigation of TyG index in a larger T2DM population may have a great impact in developing the clinical biomarker for diagnosis and disease progression of CAN. We would like to thank Dr. Bhawani Singh, Professor, Department of Medicine, Hamdard Institute of Medical Sciences and Research, Jamia Hamdard (HIMSR) and Hakeem Abdul Hameed Centenary Hospital (HAHC), Jamia Hamdard, Hamdard University, as well as Mr. Irshad (Assistant Staff at HAHC Hospital, Jamia Hamdard) for their enormous help accessing the patients. We would also like to thank all the patients that took part in the present study. **Funding:** Authors are thankful to Sun Pharma, India, for providing partial assistantship under the joint collaboration for PhD program with Hamdard University, India. **DISCLOSURE:** The authors have no potential conflicts of interest to disclose. **AUTHOR CONTRIBUTION:** Conceptualization: Habib A, Ahmad R. Data curation: Akbar M. Investigation: Akbar M, Bhandari U, Habib A. Writing - original draft: Akbar M, Bhandari U. Writing - review & editing: Akbar M, Bhandari U, Habib A, Ahmad R. Supplementary Material ====================== ###### Supplementary Table 1 Multiple linear regression analysis for TyG index in CAN

Introduction {#Sec1} ============ Coronary artery disease (CAD) is a progressive atherosclerotic condition and is, together with thrombus formation, the most important underlying mechanism of an acute myocardial infarction (AMI) \[[@CR1]--[@CR4]\]. In addition to platelet activation, thrombin generation and fibrin formation play an important role in the development of an intracoronary thrombus, which may lead to an acute coronary occlusion \[[@CR2], [@CR5], [@CR6]\]. The generation of thrombin through the tissue factor pathway is essential in the haemostatic process \[[@CR1], [@CR2], [@CR7]\]. It is crucial in normal physiology, whereas an inappropriate generation of thrombin may contribute to vascular occlusions such as in myocardial infarction. Increased thrombin generation, as an expression of activation of the coagulation system, was previously shown in patients with acute coronary syndrome and unstable angina pectoris \[[@CR8], [@CR9]\]. When prothrombin is converted to activated thrombin, prothrombin fragment 1 + 2 (F1 + 2) is formed, thus indicating thrombin generation *in vivo*, with subsequent fibrin formation. From the fibrinolytic system, plasmin converted from plasminogen degrades fibrin, resulting in degredation products like D-dimer. Elevated D-dimer levels therefore indicate both ongoing coagulation and fibrinolytic activation. Both markers have been shown to be persistantly elevated for months after the acute myocardial infarction \[[@CR5]\], whereas an early decrease in D-dimer levels has been shown to be associated with improved prognosis \[[@CR10]\]. The endogenous thrombin potential (ETP) has been proposed as an informative method to determine the degree of hypercoagulability, measuring the potential to generate thrombin *ex vivo* \[[@CR11]\]. Several studies have shown increased levels of prothrombotic markers in patients with myocardial infarction \[[@CR10], [@CR12]\] and also association to clinical outcome \[[@CR10], [@CR13], [@CR14]\]. There is, however, limited knowledge about activation of the coagulation cascade in the acute phase of ST-elevation myocardial infarction (STEMI) and also limited data on the degree of hypercoagulability in relation to the degree of myocardial injury and severity of the disease in these patients. The aim of the present substudy was therefore to investigate whether circulating levels of selected prothrombotic markers were associated with the degree of myocardial necrosis assessed by peak troponin T (TnT) and with left ventricular impairment assessed by left ventricular ejection fraction (LVEF) and N-terminal pro b-type natriuretic peptide (NT-proBNP) in STEMI patients. Furthermore, the degree of hypercoagulability was studied in relation to traditional risk factors and baseline characteristics of the STEMI population. Material and methods {#Sec2} ==================== A total of 987 percutaneous coronary intervention (PCI)-treated STEMI patients from a cross sectional cohort study were included, all admitted to Oslo University Hospital Ullevål, Oslo[,]{.ul} Norway in the period from June 2007 to August 2011. STEMI was defined as ST segment elevation of \>2 mm in two or more contiguous chest leads or \> 1 mm in two or more limb leads or left bundle branch block, together with typical chest pain and elevated troponin levels above the recommended diagnostic threshold. Patients on warfarin treatment, below 18 years of age and patients unable or unwilling to give written informed consent were excluded. Blood samples were collected at median time of 24 h after symptoms and 18 h after the PCI procedure, between 8 and 10 a.m. the following morning. In order to standardize blood sampling, and also to avoid any influence of diurnal variations and food intake, all samples were taken after an overnight fast. Routine blood samples were drawn at hospital admission, and samples for TnT were measured after standardized time intervals. Citrated blood (0.129 M trisodium citrate in dilution 1:10) was centrifuged within 30 min at 2500 × g at 4 °C and kept frozen at ÷80 °C until analyzed. D-dimer and F1 + 2 were determined by ELISA (Asserachrom D-dimer, Stago Diagnostica, Ansiere, France and Enzygnost F1 + 2, Siemens, Marburg, Germany, respectively). Coefficients of variation (CV) were for D-dimer 6.5 % and F1+ 2 5.4 %. ETP was determined by the Calibrated Automated Thrombogram (CAT) assay according to the manufacturer's instructions (Thrombinoscope BV, Maastricht, The Netherlands) and thrombin generation was measured on the Fluoroscan Ascent fluorometer (Thermo Fisher Scientific OY, Vantaa, Finland). A reagent mixture of rTF and phospholipids in addition to a thrombin-specific fluorogenic substrate in Hepes buffer containing CaCl~2~ was added to the plasma to obtain a final concentration of 5 pM, 4 μM and 416.7 μM, respectively. In order to calculate the final results, plasma was measured along with a thrombin calibrator. The software (version 3.0.0.29; Thrombinoscope BV) enabled the calculation of the lag time (LT), peak thrombin generation (pTG), ETP and time to peak (TTP). Further, V~T~ (Velocity Index) = TP/(TTP-LT), indicating the average net rate of prothrombin activation during the propagation phase, was calculated. All experiments were run in duplicates and the interassay coefficients of variation for the different CAT parameters were 14.2, 4.6, 5.0 and 8.0 %, respectively. CRP was measured with kits from DRG Instruments (Marburg/Lahn, Germany), CV \<5 %. Electrochemiluminescence technology for quantitative measurement was used for repeated measures of TnT (3rd generation cTroponinT, Elecsys 2010, Roche, Mannheim, Germany). The lower detection limit of the assay is 10 ng/L with a recommended diagnostics threshold of 30 ng/L. The inter-assay coefficient of variation was 7 %. NT-ProBNP was measured in serum using Elecsys proBNP sandwich immunoassay on Elecsys 2010 (Roche Diagnostics, Indianapolis, USA). The inter-assay coefficient of variation was 7 %. Left ventricular ejection fraction (LVEF) was measured by echocardiography before hospital discharge or at a clinical follow-up within 3 months after the AMI (*n* = 767). Diabetes was defined according to the American Diabetes Association criteria \[[@CR15]\] and hypertension (HT) was defined as previously diagnosed and treated hypertension. Smokers were defined as current smokers or quit within the last month. Clinical information was collected from hospital records and questionnaires acquired at the time of inclusion. Patients on warfarin were not included in this patient cohort. The study was approved by The Regional Ethics Commitee and all patients gave written informed consent. Statistical analysis {#Sec3} -------------------- Continous variables are presented as median values with 25,75 percentiles and categorial variables as number or proportions. As most of the variables were skewed, correlation analyses were performed using Spearman's method. Differences between groups were tested by Mann--Whitney U test for continuous variables. Associations between prothrombotic markers and peak TnT and left ventricular impairment were tested in multivariate regression models, adjusting for relevant covariates. As the markers are strongly inter-related they were analyzed in separate models. Skewed data were log-transfomed before entered in the model. P values \< 0.05 were considered statistically significant. The statistical analyses were performed with SPSS software version 18.0 (SPSS Inc, Chicago, USA). Results {#Sec4} ======= Baseline characteristics of the total population are given in Table [1](#Tab1){ref-type="table"}. The cohort was a typical STEMI population of relatively young, predominantly male patients (81 %) with medium size infarction (measured by peak TnT). Only 23 % with previous CVD, 12 % with known diabetes and half of the patients were smokers.Table 1Characteristics of the study population (*n* = 987)Age (years) (range)61 (24--94)Male sex800 (81)Current smokers474 (48)Previous CVD229 (23)Treated hypertension334 (34)Treated diabetes mellitus124 (12)BMI (kg/m^2^)26.6 (24.3,29.2)Prehospital thrombolysis119 (12)Aspirin231 (23)Statins233 (23)Total cholesterol (mmol/L)4.9 (4.1,5.6)HDL (mmol/L)1.06 (0.88,1.30)Triglycerides (mmol/L)1.25 (0.89,1.78)CRP (mg/L)13.4 (7.0,31.3)Admission glucose (mmol/L)7.4 (6.3,9.0)Fasting glucose (mmol/L)5.8 (5.2,6.6)HbA1c (%)5.9 (5.6,6.3)Peak Troponin T (ng/L)3850 (1710,7250)NT-ProBNP (pmol/L)31 (10,118)LV Ejection fraction (%)50 (44,55)D-dimer (ng/mL)456 (287,796)F1 + 2 (pmol/L)246 (178,356)ETP (nM⋅min)1564 (1366,1743)Time from onset of symptoms to blood sampling (hours) (range)24 (5--118)Number (proportions) or median (25,75 percentiles) are given*BMI* Body mass index, *CVD* Cardiovascular Disease, *HDL* High Density lipoprotein cholesterol, *CRP* C-reactive Protein, *ETP* endogenous thrombin potential Levels of the haemostatic variables in the total population are shown in Table [1](#Tab1){ref-type="table"}. There were strong inter-correlations between D-dimer and F1 + 2 (*r* = 0.504, *p* \< 0.001), and a weaker, inverse correlation between D-dimer and ETP (*r* = −0-.102, *p* \< 0.001). As visualized in Table [2](#Tab2){ref-type="table"}, age was significantly correlated with F1 + 2 and D-dimer, inversely to ETP (all *p* \< 0.001) and further weakly correlated to lag time, time to peak, peak hight and velocity index (all *p* \< 0.05).Table 2Correlations between prothrombotic markers and selected variablesD-dimerF1 + 2ETPLTTTPpTGV~T~Ager0.4120.277−0.229−0.075−0.155−0.0860.088p\<0.001\<0.001\<0.0010.021\<0.0010.0080.006BMIr−0.184−0.2030.2180.1150.1030.1830.070p\<0.001\<0.001\<0.001\<0.0010.001\<0.0010.032NT-ProBNPr0.2430.120−0.1180.0720.003−0.0330.066p\<0.001\<0.0010.0020.0260.920.300.044HbA1cr0.063−0.0150.0330.0250.0060.0550.048p0.1410.7650.1610.440.850.090.146Fasting glucoser0.0060.062−0.0540.042−0.0310.0610.104p0.8440.0740.1150.190.340.060.001Peak TnTr0.2600.364−0.072−0.012−0.0590.0010.065p\<0.001\<0.0010.0150.700.0620.970.044LVEFr−0.160−0.0900.022−0.0400.024−0.065−0.107p\<0.0010.0130.5530.270.510.0770.003*TnT* Troponin T, *LVEF* left ventricular ejection fraction, *BMI* Body mass index, *LT* lag time, *TTP* Time to peak, *pTG* peak thrombin generation, *V* ~*T*~ Velocity index*r*-values refer to Spearman's rank correlation coefficient Prothrombotic markers and association with myocardial injury {#Sec5} ------------------------------------------------------------ Statistically significant correlations were found between peak TnT and D-dimer and F1 + 2 (both *p* \< 0.001) (Table [2](#Tab2){ref-type="table"}). Linear trend analysis across quartiles of peak TnT revealed increased levels of both markers with increasing quartiles (p for trend \< 0.001). When adjusting for relevant covariates as visualized in Table [4a](#Tab4){ref-type="table"} both D-dimer and F1 + 2 remained significantly associated with peak TnT (both *p* \< 0.001) (Table [4a](#Tab4){ref-type="table"} and Fig. [1a](#Fig1){ref-type="fig"}). Weak, but statistically significant correlations were observed between TnT and ETP and velocity index (Table [2](#Tab2){ref-type="table"}).Fig. 1D-dimer and F1 + 2 (medians) in quartiles of peak TnT (**a**) and NT-ProBNP (**b**). **a** \* = adjusted for age, sex, BMI, hypertension, time from symptoms to blood sampling, CRP and NT-ProBNP; **b** \* = adjusted for age, sex, BMI, hypertension, time from symptoms to blood sampling and CRP Prothrombotic markers and association with myocardial function {#Sec6} -------------------------------------------------------------- Significant correlations were found between D-dimer and F1 + 2 and NT-ProBNP (both *p* = 0.001). Weak, but statistically significant correlations were also observed for the CAT-variables ETP, lagtime and velocity index (Table [2](#Tab2){ref-type="table"}). When dividing NT-ProBNP levels into quartiles there were significant trends for increased levels of D-dimer and F1 + 2 across quartiles (both *p* \< 0.001). D-dimer remained significantly associated with NT-proBNP after adjusting for covariates as visualized in Table [4b](#Tab4){ref-type="table"} (*p* = 0.001), whereas the association between NT-proBNP and F1 + 2 was no longer statistically significant (*p* = 0.446) (Table [4b](#Tab4){ref-type="table"}, Fig. [1b](#Fig1){ref-type="fig"}). A weak, but significant inverse correlation was found between LVEF and D-dimer (*p* \< 0.001), F1 + 2 (*p* = 0.013) and velocity index (*p* = 0.003) (Table [2](#Tab2){ref-type="table"}). When dichotomizing LVEF levels at 40 % we observed significantly higher levels of all variables (*p* \< 0.001, *p* = 0.016 and *p* = 0.004, respectively) in patients with LVEF below (*n* = 147), compared to above 40 % (Table [3](#Tab3){ref-type="table"}). After adjustments for the covariates visualized in Table [4b](#Tab4){ref-type="table"} the difference in D-dimer levels remained statistically significant (*p* = 0.003) whereas the association between LVEF and F1 + 2 and velocity index was no longer significant (*p* = 0.552 and *p* = 0.084, respectively).Table 3Levels of the prothrombotic markers according to group characteristics of the populationnD-dimer (ng/mL)F1 + 2 (pmol/L)ETP (nM⋅min)SexMale800424 (275,755)238 (175,350)1576 (1401,1748)Female187609 (399,1008)287 (203,395)1494 (1284,1702)p0.0010.0010.001Smoking+474448 (275,796)245 (179,347)1573 (1364,1762)-513467 (295,795)250 (178,382)1557 (1368,1726)p0.6750.5940.269Previous CVD+229484 (286,798)244 (177,332)1569 (1348,1733)-758452 (286,796)247 (180,369)1563 (1370,1745)p0.5530.3860.503HT+553553 (355,952)275 (197,398)1563 (1366,1730)-418418 (273,773)235 (174,346)1565 (1366,1745)p0.0010.0010.381Diabetes+870454 (246,848)232 (169,324)1490 (1261,1754)-117456 (292,795)249 (180,365)1572 (1373,1754)p0.2850.1420.004LVEF (%)\>40147440 (285,755)242 (174,363)1573 (1371,1735)≤40620679 (408,1156)297 (189,397)1488 (1293,1691)p\<0.0010.0160.017Median (25,75 percentiles) values are given*HT* hypertension, *CVD* cardiovascular disease, *LVEF* left ventricular ejection fractionTable 4Determinants of peak Troponin T (a) and NT-proBNP (b)DeterminantsStandardized beta95 % CI^c^*p*-valueDeterminantsStandardized beta95 % CI*p*-valuea)D-dimer0.138\<0.001-- \< 0.001\<0.001F1 + 20.2160.001--0.001\<0.001Age−0.054−0.012--0.0020.156Age−0.052−0.012--0.0020.157Sex0.025−0.117--0.2610.457Sex0.020−0.128--0.2440.542BMI−0.026−0.025--0.0120.466BMI−0.013−0.22--0.0150.706HT−0.006−0.179--0.1520.870HT−0.013−0.194--0.1320.710Time frame^a^−0.076−0.018-- − 0.0010.032Time frame^a^−0.071−0.017-- \< −0.0010.042CRP^b^0.0810.031--0.3330.018CRP^b^0.0830.039--0.3350.013NT-ProBNP0.1870.183--0.423\<0.001NT-ProBNP0.1960.199--0.435\<0.001b)D-dimer0.063\<−0.001-- \< 0.001\<0.001F1 + 20.024\<−0.001-- \< 0.0010.446Age0.1800.011--0.0230.050Age0.1900.012--0.024\<0.001Sex−0.083−0.405-- − 0.061\<0.001Sex−0.085−0.412-- − 0.0670.007BMI−0.086−0.039-- − 0.0060.008BMI−0.087−0.040-- − 0.0060.008HT0.1250.145--0.445\<0.001HT0.1260.148--0.449\<0.001Time frame^a^0.3070.031--0.046\<0.001Time frame^a^0.3070.031--0.046\<0.001CRP^b^0.035−0.059--0.2170.261CRP^b^0.043−0.041--0.2330.169Multivariable regression analysis adjusted for age, sex, BMI, HT, Time frame, CRP and NT-proBNPFor abbreviations, see text^a^Time from symptoms to blood sampling ^b^Logtransformed ^c^ Confidence Intervals Prothrombotic markers and traditional risk factors {#Sec7} -------------------------------------------------- Levels of D-dimer and F1 + 2 were significantly higher in women (*p* \< 0.001, both), while ETP was higher in men (*p* = 0.001) (Table [3](#Tab3){ref-type="table"}). There were no significant sex differences in other CAT variables (data not shown). No difference in any of the prothrombotic markers between smokers and non-smokers or patients with or without previous CVD was observed. In patients with hypertension, D-dimer, F1 + 2 and velocity index levels were significantly higher compared to the group without hypertension (all *p* \< 0.001), however the association weakened after adjustments for covariates (*p* = 0.018, *p* = 0.015 and *p* = 0.026, respectively). Diabetic patients had significantly lower ETP levels compared to non diabetics (*p* = 0.004) without any differences in other CAT variables (data not shown) or D-dimer and F1 + 2. There were also limited correlations between the haemostatic markers and HbA1c and fasting glucose, except for velocity index which correlated weakly to fasting glucose (Table [2](#Tab2){ref-type="table"}). Significant inverse correlations were observed between BMI and D-dimer and F1 + 2 (both *p* \< 0.001), whereas all CAT variables were positively correlated with BMI (all *p* \< 0.05) (Table [2](#Tab2){ref-type="table"}). There were significant trends for decreased levels of D-dimer and F1 + 2 and increased ETP across quartiles of BMI (adjusted *p* = 0.011, *p* = \<0.001, *p* \< 0.001, respectively). Discussion {#Sec8} ========== In this large cohort of STEMI patients we found that levels of D-dimer and F1 + 2 were significantly associated with the extent of myocardial injury as measured by peak TnT. Significant associations between these coagulation markers and myocardial function, assessed by LVEF and NT-ProBNP, were further demonstrated. We observed an inverse pattern for the *in vivo* thrombin generation and *ex vivo* potential to generate thrombin, which confirm previous findings in patients with stable CAD \[[@CR16]\]. It might be speculated that this is due to an increased *in vivo* production of thrombin in the acute phase, resulting in reduced potential to generate thrombin *ex vivo*, as an exhaustion phenomenon. Patients with STEMI admitted to primary PCI, receive heparin before or during the procedure. Heparin could potentially influence the results. However, as heparin was given only during the procedure, any effect on the measured variables was most likely not present when the blood samples were drawn 18 h (median time) after the procedure. There was a clear association between the variables and myocardial necrosis measured by peak TnT. This association was also present after adjustments for potential covariates including CRP. Thus the prothrombotic state, to some degree also reflected in CAT parameters, was probably not a result of inflammation in the acute phase. We have previously reported similar results in another AMI population \[[@CR17]\]. In that particular study the prothrombotic markers were measured 3--4 days after the acute event, probably reflecting a more stable situation. Nevertheless, the results clearly indicate that patients with larger infarctions are in an increased hypercoagulable state. It might therefore be discussed if patients with large infarctions are sufficiently protected by use of double antiplatelet therapy \[[@CR18]\]. Use of warfarin has been shown to reduce both D-dimer and F1 + 2 after AMI \[[@CR19]\], and randomized, clinical studies have shown beneficial effects on clinical outcome by use of warfarin as anticoagulation after acute MI \[[@CR20], [@CR21]\]. The findings of a significant association between procoagulant activity, shown especially by D-dimer and F1 + 2, but also by CAT variables, and impaired myocardial function in the acute phase of a STEMI, has to our knowledge, not been reported before. Elevated prothrombotic markers in the early phase of AMI are known to identify patients with incrased risk of subsequent cardiac death, but such associations have so far been reported to appear independent of LVEF \[[@CR22]\]. Elevated levels of D-dimer and F1 + 2 were shown along with impaired myocardial function in another population not suffering from CAD \[[@CR23]\]. An association between elevated D-dimer and heart failure has also been demonstrated \[[@CR24]\]. Although there is no convincing evidence that oral anticoagulant therapy reduces mortality and vascular events in patients with heart failure and sinus rhythm \[[@CR25], [@CR26]\], prolonged anticoagulant treatment of such patients may be discussed after an AMI. Diabetes is generally associated with elevated levels of prothrombotic markers \[[@CR27]\], also in diabetic patients without coronary heart disease \[[@CR28]\]. In our population, diabetes and glucometabolic disturbances were limited associated with a prothrombotic state, other than lower levels of ETP and a significant correlation between fasting glucose and velocity index. The latter may indicate glucose per se to play a role for the propagation phase of thrombin generation. Our results differ from some other studies showing enhanced thrombin generation in diabetics \[[@CR28]--[@CR30]\]. However, in the study by Tripodi et al. ETP levels were higher in diabetics versus controls only in the presence of added thrombomodulin \[[@CR29]\]. Difference in the populations investigated may also be of importance. The limited findings in our study may be explained by the elevated levels of prothrombotic markers in the acute situation of an AMI, thus masking any difference. In addition, the levels of fasting glucose and HbA1C indicate adequate treatment of diabetes in the present population. Similar results have also been shown in another study on stable patients with CAD \[[@CR16]\]. The inverse correlations between BMI and D-dimer and F1 + 2 indicating a less hypercoagulable state in overweight individuals, are in accordance with previous findings in a population of stable CAD patients \[[@CR16]\] and is not easily explained. In contrast, all CAT variables were positively associated with BMI, indicative of an increased potential to thrombin generation. Other studies have shown positive correlation between BMI and D-dimer, however, only in patients not diagnosed with CVD \[[@CR31]\]. In one study on healthy, obese individuals D-dimer values were found not to be correlated to BMI \[[@CR32]\]. Increased levels of prothrombotic markers in patients with hypertension is well known \[[@CR33]--[@CR35]\]. This was also present in our population of STEMI patients when evaluated in the acute phase, showing elevated levels of D-dimer, F1 + 2 and velocity index in the group of hypertensive patients, although highly dependent of related factors. Limitations {#Sec9} ----------- Single bloodsampling prevented us from studying the time-course of the measured markers. We are not sure to have measured peak values of the variables or transient changes due to the variability in the time frame from onset of symptoms to blood sampling. However, the results did not change when taken this into account in the multivariate models. The blood samples were centrifuged at 4° C, thus any contact activation cannot be ruled out. We have also not included an extra centrifugation step before the analysis. The measure of LVEF by echo cardiography was performed at different time points from hospital discharge until 3 months after the index infarct, and our cohort of STEMI patients was a low risk population with few complications and just slightly reduced LVEF, and this fact may have influenced the results. As we do not have follow-up information of this cohort, any impact of the results on future clinical endpoints cannot be explored. Conclusion {#Sec10} ========== In our cohort of STEMI patients we could demonstrate a significant association between levels of D-dimer and F1 + 2 and the extent of myocardial necrosis as assessed by TnT. The high levels of these markers in patients with low LVEF and high NT-ProBNP may indicate a hypercoagulable state in patients with impaired myocardial function. The inverse relation between BMI and procoagulant activity is not easily explainable, and has to be further explored. **Competing interests** The authors declare that they have no competing interests. **Authors' contributions** CHH wrote the manuscript and participated in the coordination of the study, performed analysis and interpretation of data, contributed to acqusition of data , VR contributed to acqusition of data, SH participated in the design of the study and helped to draft the manuscript, GØA participated in the design of the study and helped to draft the manuscript, RB participated in the design of the study, JE participated in the design of the study, HA participated in the design of the study and helped to draft the manuscript, IS helped perform the statistical analysis, participated in the design and coordination of the study and helped to draft the manuscript. All authors read and approved the final manuscript. This work was supported by the Stein Erik Hagen Foundation for Clinical Heart Research, Oslo, Norway. We thank the study nurses and the staff at the Coronary Intensive Care Unit and Center for Clinical Heart Research for excellent assistance and medical technologist Beate Vestad for laboratory analysis. The study was a part of the Biobanking in myocardial infarction (BAMI) project at Oslo University Hospital, Ullevål, which is lead by a steering committee including Mangschau, A and the following authors: Seljeflot, Arnesen (Chair), Eritsland, Halvorsen, Bjørnerheim and Andersen.

INTRODUCTION ============ The primary goal in the management of urinary incontinence and erectile dysfunction related to the treatment of prostate cancer is the improvement of long-term quality of life of our patients. Urinary incontinence following radical prostatectomy affects 3%-60% patients and may significantly impact quality of life ([@B1]). Sanda et al. found that, at 12 months following prostatectomy, 24% of patients were using pads and 8% found it as a significant problem ([@B2]). The artificial urinary sphincter (AUS), first introduced by Scott et al. in 1974, remains a mainstay in the management of post-prostatectomy stress urinary incontinence (SUI) ([@B3]). Yet, there has been continued refinement in the technique for implantation in an effort to decrease patient morbidity and discomfort. Traditionally, the technique for the implantation of the AUS device utilizes two incisions: a perineal and an inguinal incision. This allows for bulbar urethral placement of the cuff and a retropubic location for the pressure regulating balloon (PRB) into the space of Retzius by piercing transversalis fascia ([@B4]). In 2003, Wilson et al. introduced a single incision technique for placement of AUS through a transverse scrotal incision ([@B5]). Subsequently, Wilson and Delk described 'ectopic' placement of the PRB between transversalis fascia and rectus muscles, which would avoid potential problems placing the PRB in the space of Retzius ([@B6]). Wherein they describe ectopic placement of the PRB from both transverse scrotal and perineal approaches ([@B6]). Ectopic placement avoids the potential hazards of placement into a previously operated field or radiated retropubic space, including injury or obstruction to surrounding vasculature or organs (intestines, bladder, ureter). Anecdotally, placement of the control pump via a perineal incision can be complicated by pump migration within the scrotum and perineum if a true sub-dartos pouch is not created. The failure to replicate the same placement obtained either with sub-scarpal placement from an inguinal incision or sharp dissection from a scrotal incision can result in instability of the pump location, which may require revision to correct. Previously, our preferred surgical approach for AUS implantation utilizes two incisions, a perineal incision and a counter lower abdominal incision. We now demonstrate the feasibility of a single perineal incision placement of an AUS. MATERIALS AND METHODS ===================== We performed a retrospective review of six patients with stress urinary incontinence undergoing AUS placement through a single perineal incision performed by a single surgeon (OLW) between June 2014 and December 2014 at MD Anderson Cancer Center. Institutional review board approval was obtained for the study and informed consent was obtained from all patients included in this study. All patients underwent routine pre-operative evaluation including 24-hour pad test, urodynamics and office cystoscopy. All patients were virgin AUS placements; however, half ([@B3]) underwent simultaneous placement of inflatable penile prosthesis (IPP). Patients who had untreated inguinal hernia or prior inguinal herniorrhaphy were excluded, as well as those patients whose external inguinal rings were inaccessible due to anatomic distance. We utilized the 61-70cm H~2~O PRB filled with 23cc of normal saline in all patients and cuff sizes varied between 3.5-5cm. Additionally, we utilized a cadaveric model to demonstrate the reproducibility of creating the sub-dartos pouch for the AUS pump through the perineal approach, which was identical to the location with traditional placement through the two-incision technique ([Figure-1](#f1){ref-type="fig"}). {#f1} SURGICAL TECHNIQUE ================== All components of the AMS 800™ AUS (Minnetonka, Minnesota) device were placed through a single perineal incision. After placement of a 12 or 14F foley catheter, a standard midline perineal incision is made. Dissection is performed down to the level of the bulbospongiosus muscle followed by exposure of the corpus spongiosum and bulbar urethra. A Lone Star^®^ retractor with blunt hooks exposes the surgical field. A combination of sharp and blunt dissection is used to mobilize approximately a 2cm segment of the proximal bulbar urethra circumferentially with subsequent measurement of the urethra with a cuff sizer. After the urethra is sufficiently mobilized, our attention turns to palpating the external inguinal ring through the perineal incision. Once a finger is placed within the external inguinal ring, a pediatric deaver retractor retracts the anterior wall of the canal, and a ring forceps is advanced cephalad and medial to the spermatic cord spreading to create a potential space between transversalis fascia and rectus muscle ([Figures 2a](#f2){ref-type="fig"} and [b](#f2){ref-type="fig"}). Once this space has been created, a coated ring forceps is utilized to advance the PRB into the prepared ectopic space followed by inflation with 23cc of injectable saline ([Figure-2c](#f2){ref-type="fig"}). The tubing is occluded with hemostat. Interrupted 3-0vicryl sutures are placed around the entrance of the PRB into the ectopic space to prevent migration. {#f2} Attention is turned to creating a sub-dartos pouch for the AUS pump. Blunt dissection is utilized to deviate the tunical sac medially. The right hemi-scrotal skin is inverted through the perineal incision ([Figure-3a](#f3){ref-type="fig"}). The internal spermatic fascia is incised until the dartos fibers are visualized ([Figure-3b](#f3){ref-type="fig"}). A finger is placed within this incision and used to bluntly create a space for the pump by reverting the scrotal skin ([Figure-3c](#f3){ref-type="fig"}). The skin is inverted through the incision and the pump is placed within this space with the pump positioned within the pouch. Per standard, the deactivation button is positioned laterally ([Figure-3d](#f3){ref-type="fig"}). The position is stabilized with a Babcock clamp. {#f3} The cuff is then placed around the urethra using a right angle clamp and clipped into position. The tubing from the cuff is then passed with a needle passer through the bulbospongiosus muscle and through the Colle\'s fascia into the same plane as the pump tubing. The tubing length is planned to assure that the connections will reside in an inguinal location. The Quick-Connect system is used to seal the connections. The tubing is tucked superiorly to an inguinal location. A suture is placed to assure that the tubing does not prolapse into the perineum. The bulbospongiosus is then closed in a running fashion, followed by Colle\'s fascia and then the perineal skin. The Babcock is then removed and the device is cycled and subsequently deactivated. RESULTS ======= A total of 6 AUS devices were placed using the single perineal incision technique. The etiology of the urinary incontinence was prostate cancer treatment related in all cases. All six patients underwent radical prostatectomy (4 robotic assisted and 2 radical retropubic), of which 5 were treated with radiation either as their initial treatment or in the post-operative salvage setting. The average patient age was 61 (SD 7.5 years) and average BMI 31 (SD 5.9). The average pre-operative pad usage was 7.7 pads daily (SD 1.63), 24-hour pad weight was available for 4 of the patients with an associated mean 24 hour pad weight of 517.6g (SD, 605g). Four patients underwent urethral procedures prior to AUS implantation: 3 patients underwent direct visualized internal urethrotomy for bladder neck contracture and one patient required buccal mucosal graft urethroplasty for urethral diverticulum. Pre-operative cystoscopy to document the stability of urethra prior to AUS implantation was performed in all cases. The average operative time was 101 minutes (SD, 27 minutes) for the entire cohort including cases with simultaneous implantation of AUS and IPP; however, the 3 patients that had AUS alone, the mean operative time was 81 minutes (SD, 17 minutes). The mean follow-up for the cohort was 13.9 months (SD 9.45). Four patients met criteria for "socially dry" (1 pad or less per day), of which two were wearing zero pads at last follow-up. There was one patient who continued to have significant incontinence, wearing 10 pads daily after initial AUS placement. Of note, his pre-operative 24-hour pad weight was 1700grams. Initially, he had a 5cm cuff placed and underwent downsizing of his cuff to 3.5cm and was recently activated. One patient was lost to follow-up after activation of his device. There were no reported perioperative complications by any patient over the follow-up period, including infections, erosions, PRB herniation or pump migration. Further no device related morbidity occurred in patients with simultaneous IPP placement, and all had satisfactorily functioning IPP and AUS devices. DISCUSSION ========== The AUS continues to be relevant and the standard management for moderate-to-severe post-prostatectomy stress urinary incontinence ([@B3]). Despite few modifications in the AMS 800 Urinary Control System itself over the last decades, there has been advancement in our understanding of the function of the device as well as novel implantation considerations, with the goal of minimizing complications. Wilson and Delk initially described utilizing a single transverse scrotal incision to implant the AUS, with ventral retraction rather than division of the bulbocavernosus muscle. They found that 66% of patients were completely dry with mean follow-up of 12 months, and compared this to the traditional two-incision approach and found similar continence rates ([@B5]). The trans-scrotal approach was initially utilized for revision and reimplantation cases to avoid a scarred perineum; however, it was adopted as a more efficient approach in the primary setting. Despite this, others have been critical of the outcomes with this technique. Henry et al. suggested that the penoscrotal placement of the AUS had inferior functional outcomes compared to the originally described perineal approach, which they evaluated in both primary and revision settings ([@B7], [@B8]). In a retrospective series of virgin implantations, they found that that 7 of 25 patients (28%) with scrotal incision compared to 17 of 30 patients (56.7%) with a perineal incision were completely dry without pad usage (p=0.03) ([@B7]). However, in regards to social continence, defined as wearing one pad or less daily, there was not a significant difference between the groups ([@B7]). Overall, including both initial placements and revisions, there was a significant difference in completely dry rate between the perineal and scrotal approaches (p=0.01) ([@B7]). In a multicenter study including 158 patients, the perineal incision group was more likely to be completely dry than the scrotal incision group (44.1% versus 27.4%, p=0.04) ([@B8]). The scrotal incision group was also more likely to require tandem cuff placement for continued incontinence after initial implantation (10% versus 1.4%, p=0.04) ([@B8]). There was no difference in AUS device durability nor rates of complications, between the two techniques ([@B8]). Similarly, our preferred approach for virgin AUS placement is perineal to facilitate access to the proximal bulbar urethra. Wilson and Delk initially described the ectopic placement of the PRB to avoid the risks associated with blind puncture of the transversalis fascia and placement into the retropubic space ([@B6]). The ectopic location is a potential space developed between transversalis fascia and posterior rectus muscle using blunt finger dissection through the external inguinal ring. They described ectopic placement both via perineal and transverse scrotal incisions without need for a second counter incision ([@B6]). Morey et al., also described ectopic submuscular placement using a Foerster clamp to develop the potential space ([@B9]). However, in their series, a scrotal counter incision was used for ectopic placement of the PRB, whereas we utilize a single perineal incision. Further, Singla et al., compared the outcomes of the AUS with an ectopic PRB versus a PRB within the retropubic space ([@B10]). There were no significant differences in continence outcomes (88% versus 81% p=0.11), erosion rates (8% versus 9% p=0.66) and need for revisions (8% versus 13% p=0.16) in this series and similar rates of explantation ([@B10]). There is a paucity of high quality evidence regarding AUS implantation; there are non-uniform outcome measures (both objective and subjective) and definitions of continence, making it difficult to interpret the evidence as a whole ([@B11]-[@B14]). A systematic review by Van der Aa et al., 79% patients included were socially continent and 43.5% completely dry without any pads and a 26% reintervention rate ([@B11]). Overall, the early results of our technique are generally consistent with expected outcomes. Our primary concerns were related to the comparability of the pump placement to standard techniques. By the time of activation, the sub-dartos location of the pump should be stabilized, especially with more consistent manipulation by the patient. Thus, the likelihood that dislocation problem would develop decreases with time. Assuming that the continence and long-term complications are similar to other techniques, there are several benefits of a single perineal incision. There is expected improvement in patient discomfort and bother with a single incision, while saving the surgeon the need to open and close a second site and thus shortening the time of the procedure. The limitations of this study include the small sample size, the retrospective nature and limited follow-up regarding single perineal incision AUS implantation. Despite this, we believe this technique is reproducible, safe, and effective in an appropriately selected patient. This study is also limited by the lack of patient satisfaction outcomes and comparison with standard implantation techniques. The endpoint of interest is stability and functionality of the pump due to the mechanism of sub-Dartos pouch formation considering that the cuff and reservoir placement are consistent with our standard technique. Thus, the follow-up is sufficient to determine whether any substantial problems are encountered on the basis of this variation in technique. We will continue to monitor these patients to confirm that the long-term outcomes (e.g. continence, pump migration, PRB migration) are consistent with their two incision counterparts. CONCLUSIONS =========== This study demonstrates the feasibility of a single perineal incision for AUS placement in the properly selected patient. Utilization of a single perineal incision for AUS placement is safe and effective. Longer follow-up will be necessary to confirm no pump related mechanical problems specifically related to this technique. Published as Ahead of Print: November 10, 2017 [^1]: **CONFLICT OF INTEREST** None declared.

INTRODUCTION {#s1} ============ Epidemiological evidence clearly indicates increased association between breast carcinoma and metabolic disorders (i.e., obesity; type 2 diabetes, metabolic syndrome) \[[@R1]--[@R5]\]. In fact, numerous studies have confirmed a higher risk of bearing breast tumors as well as unfavorable breast cancer prognosis in obese, diabetic and hyperinsulinemic patients (reviewed in \[[@R6]\]). While the relative contribution of metabolic alterations versus the direct effects of increased adipose tissue (e.g., adipokine/cytokine secretion, augmented production of estrogens) on breast tumor promotion have been greatly debated \[[@R3]\], extensive research in the field clearly points to the crucial role of hyperinsulinemia and insulin signaling \[[@R2], [@R3], [@R6]--[@R8]\]. Indeed, the aforementioned metabolic conditions are manifested by an increase in circulating insulin \[[@R3]\], since under state of insulin resistance (a central feature in the aetiology of type 2 diabetes, most frequently caused by obesity) the pancreatic β cells induce a physiological compensatory response that results in hyperinsulinaemia. Tumor promoting activity of insulin can be mediated by insulin receptor (INSR) as well as cross-activation of the insulin-like growth factor 1 receptor (IGF1R) family, resulting in activation of downstream signaling cascades \[[@R3], [@R7]\]. In particular, the phosphatidylinositol-3 kinase (PI3K)/AKT pathway is a critical component of the insulin responses, promoting breast tumor cell growth, survival and aggressive behaviour. Its importance has recently been highlighted by the fact that 50% of breast carcinomas have an activated PI3K/AKT pathway due to mutations in one of its components \[[@R9]\]. Of note, hyperinsulinemia increases mammary tumor growth *in vivo*, and blockade of the INSR/IGF1R reduces tumor growth in hyperinsulinemic mice \[[@R7]\]. Moreover, most recent clinical observations revealed that breast carcinoma patients with low IR expression levels had significantly longer progression-free survival and overall survival \[[@R10]\], while presence of phosphorylated INSR/IGF1R is related to poor survival in all breast cancer subtypes \[[@R11]\]. Extracellular matrix (ECM) components, such as heparan sulfate (HS) proteoglycans, and their enzymatic remodeling may also affect insulin-INSR signaling axis \[[@R12]--[@R15]\], however direct contribution of these factors to accelerated breast tumor progression/aggressiveness in patients with metabolic disorders remained underappreciated. Here we present the evidence that activity of heparanase enzyme (the only known mammalian endoglycosidase that cleaves HS) leads to enhanced insulin signaling and activation of downstream tumor-promoting pathways in breast carcinoma cells. HS proteoglycans are ubiquitously found in the ECM and plasma membrane of cells: HS polysaccharide chains bind to and assemble ECM proteins, playing important roles in the integrity of the ECM \[[@R16], [@R17]\]. In addition, HS chains regulate the activity of a wide variety of bioactive molecules (i.e., cytokines, growth factors) at the cell surface and in the ECM \[[@R16], [@R18]\]. Enzymatic degradation of HS by heparanase profoundly affects numerous pathophysiological processes, including cell growth, differentiation and motility \[[@R14], [@R19]--[@R21]\]. Since uncontrolled cleavage of HS could result in significant tissue damage, under normal conditions heparanase is kept tightly regulated: with a few exceptions (placenta, activated immune cells), in non-cancerous cells/tissues heparanase promoter is inhibited and the gene is not expressed. Notably, upregulation of heparanase was reported in various cancer types \[[@R14], [@R19]--[@R21]\]. During tumor progression, the enzyme was reported to contribute, among other aspects, to the breakdown of extracellular barriers, bioavailability of HS-binding growth factors and generation bioactive HS fragments potentiating growth factor-receptor binding/signaling \[[@R18], [@R20]--[@R22]\]. Preferential expression and involvement of heparanase in breast cancer are particularly well documented, both in experimental \[[@R19], [@R23]--[@R26]\] and clinical settings \[[@R27], [@R28]\]. Importantly, increased heparanase levels are associated with reduced patients' survival post operation and increased tumor aggressiveness \[[@R14], [@R21]\]. Moreover, heparanase expression in several tumor types (including breast) was linked to therapy resistance \[[@R28], [@R29]\]. Altogether, these data, along with the recently reported ability of the enzyme to enhance INSR signaling in myeloma \[[@R15]\], prompted us to hypothesize that heparanase may play an important role in diabetes-associated breast cancer, facilitating tumor aggressiveness under hyperinsulinemic state. In the present study we report that in breast carcinoma patients, diabetic state conferred more aggressive phenotype (manifested by increased lymph node involvement) to heparanase-positive, as compared to heparanase-negative tumors. In agreement with clinical observations, *in vitro* heparanase enzyme augmented INSR signaling / downstream AKT activation in breast carcinoma cell lines, and enhanced insulin-induced growth of breast tumor cells. Taken together, our findings highlight the emerging role of heparanase in modulating pro-tumorous effect of hyperinsulinemic state on breast tumorigenesis and imply that heparanase-targeting therapeutic approaches could be particularly beneficial in breast carcinoma patients suffering from metabolic disorders. RESULTS {#s2} ======= Correlation between heparanase expression and lymph node involvement in diabetic breast carcinoma patients {#s2_1} ---------------------------------------------------------------------------------------------------------- Women with diabetes have higher breast cancer incidence and mortality, exhibiting a higher risk of lymph node metastases \[[@R4], [@R5]\]. Therefore, to explore the hypothesized role of heparanase in diabetes-breast cancer link, we first examined correlation between expression of heparanase in tumor tissue and lymph node involvement, using clinical data from 67 breast carcinoma patients (15 with diabetes mellitus and 52 non diabetic/non-obese controls). Since obesity is the main cause of insulin resistance and in many individuals is the first step in the development of type 2 diabetes and metabolic syndrome \[[@R3], [@R30]\], obese patients were excluded from the non-diabetic control group. Among diabetic patients bearing heparanase-positive tumors, more than 70% (5 out of 7) displayed lymph node involvement. In contrast, in only 12.5% diabetic patients with heparanase-negative tumors (one out of eight) node involvement was noted. Statistical analysis confirmed that under diabetic state heparanase-positive tumors are more likely to spread into lymph nodes (two-sided Fisher\'s exact test; p=0.04; Figure [1](#F1){ref-type="fig"}). Of note, presence of either diabetic state or heparanase alone did not confer statistical significant difference in lymph node involvement (Figure [1](#F1){ref-type="fig"}). ![Heparanase expression and lymph node involvement in diabetic breast carcinoma patients\ Human breast carcinoma tissue samples (biopsies) were processed for immunohistochemistry with anti-heparanase antibody (733) directed against a synthetic peptide (^158^KKFKNSTYRSSSVD^171^) corresponding to the N-terminus of the 50-kDa subunit of the heparanase enzyme, as described in \[[@R28], [@R57], [@R58]\]. Diabetic state, BMI and lymph node status were determined from patient history. Node+: patients with lymph node positive tumors; Node-: patients with lymph node negative tumors. Two-sided Fisher\'s exact test confirmed that in diabetic patients heparanase-positive tumors are more likely to spread into lymph nodes (\*p=0.04). Presence of heparanase in non-diabetic breast tumor samples did not confer statistical significant difference in lymph node involvement.](oncotarget-08-19403-g001){#F1} Heparanase enhances INSR signaling and insulin-induced proliferation in breast carcinoma cells {#s2_2} ---------------------------------------------------------------------------------------------- The above clinical observations (although limited by a small sample size), along with importance of insulin-INSR signaling in diabetes/obesity-related cancer \[[@R2], [@R3], [@R6]--[@R8]\] and the recent report on heparanase-augmented INSR signaling in myeloma \[[@R15]\], lead us to hypothesize that heparanase facilitates breast carcinoma progression under diabetic conditions via augmentation of INSR signaling. To test this hypothesis, we first incubated ER-positive MCF-7 human breast carcinoma cells with insulin in the absence or presence of recombinant active heparanase enzyme. As expected, INSR signaling was induced following treatment with insulin, as evident by increased phospho-INSR (pINSR) and phospho-AKT (pAKT) levels (Figure [2A-2C](#F2){ref-type="fig"}). However, presence of recombinant heparanase significantly increased pINSR and pAKT levels in response to insulin treatment (Figure [2A-2C](#F2){ref-type="fig"}). Protein levels of total AKT and total INSR were not affected by presence of heparanase. Notably, this augmented response to insulin in the presence of heparanase was dependent on heparanase enzymatic activity, since treatment with heat-inactivated heparanase (iHpa, Figure [2](#F2){ref-type="fig"} inset) did not affect pINSR/pAKT levels (Figure [2](#F2){ref-type="fig"}). Similar results were obtained in ER-positive E0771 murine breast carcinoma cells (Figure [2D-2F](#F2){ref-type="fig"}). Of note, presence of recombinant heparanase also augmented INSR signaling in ER-negative breast carcinoma cell line MDA-MB-231 (2-fold increase in pINSR levels, not shown), consistent with the reports that progression of both hormone dependent and independent breast tumors is affected by INSR signaling \[[@R31], [@R32]\]. {#F2} We then utilized MCF-7 cells stably transfected with vector encoding for human heparanase (*MCF7-Hpa*) or mock-transfected with empty vector (*MCF7-mock*) (Figure [3A](#F3){ref-type="fig"}). As shown in Figure [3B](#F3){ref-type="fig"}, treatment with insulin resulted in significantly increased pINSR and pAKT levels in MCF7-Hpa cells as compared to MCF7-mock cells. This effect was abolished in the presence of specific inhibitor of heparanase SST0001 (Figure [3C](#F3){ref-type="fig"}). Protein levels of total AKT and total INSR were not affected by presence of heparanase (Figure [3B](#F3){ref-type="fig"}). Next, to test the biological consequence of heparanase-augmented INSR signaling, we compared proliferation rate of MCF7-Hpa and MCF7- mock cells in response to insulin treatment. As shown in Figure [3D](#F3){ref-type="fig"}, proliferation in response to insulin was markedly enhanced in MCF7-Hpa, as compared to MCF7-mock cells. Altogether these results demonstrate that INSR signaling in breast carcinoma cells is enhanced in the presence of heparanase and results in accelerated cell proliferation. {#F3} DISCUSSION {#s3} ========== It is well recognized that hyperinsulinemia and INSR signaling are critical determinants responsible for accelerated progression and aggressive phenotype of breast cancer in patients with metabolic disorders (i.e., diabetes, obesity) \[[@R2], [@R3], [@R6]--[@R8]\]. Heparanase enzyme has also been implicated in breast tumor progression \[[@R19], [@R23]--[@R28]\]. In the present study we show that the interplay between heparanase and insulin signaling may foster breast tumorigenesis, mechanistically linking the enzyme into the breast cancer-promoting action of metabolic disorders. Underscoring this previously unrecognized role of the enzyme in this phenomenon, we found statistically significant association between lymph node involvement and simultaneous presence of both diabetic state and heparanase expression (two-sided Fisher\'s exact test; p=0.04; Figure [1](#F1){ref-type="fig"}). Of note, diabetic state alone or heparanase overexpression alone were not associated with statistically significant increase in lymph node involvement in the patient cohort analyzed in our study (Figure [1](#F1){ref-type="fig"}). In agreement with our clinical observations is the ability of heparanase to augment insulin-induced proliferation in breast carcinoma cells *in vitro* (Figure [3D](#F3){ref-type="fig"}), particularly relevant in light of the previous reports showing association between high proliferative rate of breast carcinoma cells and lymph node metastasis \[[@R33], [@R34]\]. While our findings reveal augmented insulin receptor signaling in the presence of heparanase in breast carcinoma, a limitation of our study is that precise molecular mechanism underlying this phenomenon has not been determined. One intriguing possibility is that soluble heparan sulfate, liberated by heparanase, may facilitate formation and stabilization of insulin-INSR signaling complexes. Indeed, it was previously demonstrated that heparin or heparan sulfate proteoglycans (including HS fragments generated by heparanase enzyme \[[@R22]\]) function as growth factor tyrosine kinase receptor accessory molecules, ligands or co-ligands \[[@R35]--[@R41]\], potentiate ligand-receptor binding, dimerization, and signaling. Of note, the cell surface HS proteoglycan syndecan-1 coupled ternary receptor complex (prevalent on tumor cells and activated endothelial cells) has been described, whereby syndecan-1 clusters IGF-1R and integrins, leading to integrin activation \[[@R42], [@R43]\]. In a similar manner, heparanase-released bioactive HS fragments could stimulate the clustering with INSR leading to its dimerization and autophosphorylation. Additionally, a recent study demonstrated that cell surface heparan sulfate proteoglycans (i.e., glypicans), are capable of interacting with and enhancing INSR signaling \[[@R12], [@R13]\]. The effect on INSR signaling appears to involve release of glypican from the cell surface of adipocytes and possibly other cell types by an enzymatically regulated process and its direct interaction with INSR \[[@R13]\], likely through heparan sulfate chains. This mode of action could be particularly relevant in light of a previously demonstrated role of heparanase enzymatic activity in shedding of cell surface heparan sulfate proteoglycans \[[@R14]\], and release of cell surface-derived bioactive heparan sulfate that potentiates growth factor-receptor signaling \[[@R18], [@R22]\]. Our *in vitro* studies demonstrated ability of heparanase to enhance INSR and perhaps IGF1R signaling in both ER -positive (MCF-7, E0771) and -negative (MDA-MB-231) cell lines. Yet, ER status could still be an important factor in inducing heparanase expression in breast tumors, therefore contributing to the occurrence of INSR-heparanase interplay under diabetic state. Estrogen signaling has been previously shown to induce heparanase in breast carcinoma \[[@R28], [@R44]\]. A frequent accompanying feature of type 2 diabetes is obesity, and obese state is characterizes by elevated estrogen levels in women \[due to increased aromatase expression \[[@R1], [@R45], [@R46]\]. Thus, estrogen is expected to upregulate heparanase in breast tumor cells *per se* \[[@R44]\]. At the same time, several diabetic milieu constituents, known to increase heparanase expression/secretion in endothelial cells and immunocytes \[[@R47]--[@R51]\], may augment the expression of the enzyme in stromal elements of the tumor, suggesting that a self-sustaining circuit may exist in metabolic disorder-related ER positive breast cancer, where estrogen, and components of the diabetic milieu induce heparanase, while heparanase acts in tandem with elevated insulin to promote cell growth via enhanced INSR signaling. Noteworthy, reports by Parish and co-workers also suggested a role for heparanase in the pathogenesis of diabetes *per se* \[[@R52], [@R53]\]. Although further studies (including analysis of larger patient groups) are warranted to fully dissect the complex molecular events underlying action/regulation of heparanase in modulating effects of metabolic disorders on breast cancer progression, our findings point to the importance of the interplay between enzymatic activity of heparanase and insulin signaling in facilitating breast tumorigenesis. Moreover, our observations imply that heparanase-targeting therapeutic approaches, which are now under intensive development/clinical testing \[[@R14]\], alone or in combination with INSR/IGF1R pathway inhibition, may disrupt this interplay and therefore be particularly beneficial in a significant fraction of breast cancer patients. MATERIALS AND METHODS {#s4} ===================== Cell culture and transfection {#s4_1} ----------------------------- MCF-7 and MDA-MB-231 human breast carcinoma cells and E0771 mouse breast carcinoma cells were grown in RPMI 1640 medium supplement with 1 mM glutamine, 50 μg/ml streptomycin, 50 U/ml penicillin and 10% fetal calf serum (FCS) (Biological Industries) at 37°C and 8% CO~2~. For stable transfection, MCF-7 cells were transfected with human heparanase cDNA (H-*hpa* transfectants) or with a control empty pcDNA3 vector (Invitrogen) (mock transfectants), as previously described \[[@R28]\]. Transfected cells were selected with 800 μg/ml G418 and stable populations of heparanase expressing cells were obtained. To rule out the possibility of insertional mutagenesis, all the experiments involving heparanase- and mock-transfected cells have been conducted using a pooled population of clones, which contained over 100 clones mixed together. Expression of heparanase was evaluated by RT-PCR and verified by measurements of enzymatic activity, as described below and in refs. \[[@R19], [@R25], [@R54], [@R55]\]. Prior to insulin treatment 60-80% confluent cells, maintained overnight in serum-free RPMI, remained untreated or were incubated with 100nM insulin (Biological Industries) for 15 or 30 minutes. Cells were then lysed and processed for western blot analysis. Antibodies {#s4_2} ---------- Immunoblot analysis was carried out with the following antibodies: anti-phospho-insulin receptor Tyr1150/1151, anti-phospho-AKT Ser 473, anti-total insulin receptor, anti- total AKT (Cell Signaling) and anti-actin (Abcam). Immunoblotting {#s4_3} -------------- MCF-7 whole cell lysates were homogenized in lysis buffer containing 0.6% SDS, 10 mM Tris-HCl, pH 7.5, supplemented with a mixture of protease inhibitors (Roche) and phosphatase inhibitors (Thermo Scientific). Equal protein aliquots were subjected to SDS-PAGE (8% acrylamide) under reducing conditions and proteins were transferred to a polyvinylidene difluoride membrane (Millipore). Membranes were blocked with 3% BSA for 1 hour at room temperature and probed with the appropriate antibody, followed by horseradish peroxidase-conjugated secondary antibody (KPL) and a chemiluminescent substrate (Biological Industries). Heparanase activity assay {#s4_4} ------------------------- Measurements of heparanase enzymatic activity was carried out as described previously \[[@R19], [@R54], [@R55]\]. Briefly, tested samples were incubated (16-36 h, 37°C, pH 6.2) on dishes coated with sulfate-labeled ECM, prepared as described \[[@R19], [@R54], [@R55]\]. Sulfate-labeled material released into the incubation medium was analyzed by gel filtration on a Sepharose 6B column. Nearly intact heparan sulfate proteoglycans are eluted just after the void volume (peak I, K~av~ \< 0.2, fractions 1-10). This material (peak I) has been previously shown to be generated by a proteolytic activity residing in the ECM itself; Heparan silfate (HS) degradation fragments are eluted later with 0.5\<K~av~\<0.8 (peak II, fractions 15-35) \[[@R19], [@R54]\]. These fragments were shown to be degradation products of HS as they were 5-6 fold smaller than intact HS side chains, resistant to further digestion with papain and chondroitinase ABC, and susceptible to deamination by nitrous acid \[[@R19], [@R54]\]. Each experiment was performed at least three times and the variation in elution positions (*K*~av~ values) did not exceed ±15%. MTS assay {#s4_5} --------- MCF-7 cells were seeded in 96-well culture plates in serum free RPMI. MTS assay (Promega) was performed according to manufacturer instructions and proliferation was measured 72 hours after insulin (100 nM) was added. Each experiment was performed at least 3 times. Each data point shows the mean of pentaplicate cultures. Clinical data and immunostaining {#s4_6} -------------------------------- Data from 67 breast carcinoma female patients (15 diabetic and 52 non-diabetic with BMI\<30) were available from the Sharett Oncology Institute, Hadassah Medical Center, Jerusalem. The use of these data and formalin-fixed, paraffin-embedded breast carcinoma tissues in research was approved by the Human Subjects Research Ethics Committee of the Hadassah Medical Center. The collective median age of the patient cohort was 50 years (range 20--82). Diabetes history was obtained from the clinical charts forms and determined using the medical history coded WHO-DDE - World Health Organization medical dictionaries. Patients for whom diabetic history/diabetic treatment drugs were found upon review of the clinical charts were included in the diabetic group, otherwise the patients were included in the control group. Determination of heparanase expression status: five-micron sections of tumor tissue were deparaffinized and rehydrated. Tissue was then incubated in 3% H~2~O~2~, denatured by boiling (3 min) in a microwave oven in citrate buffer (0.01 M, pH 6.0), and blocked with 10% goat serum in PBS. Sections were incubated with polyclonal rabbit anti-heparanase antibody (733) directed against a synthetic peptide (^158^KKFKNSTYRSSSVD^171^) corresponding to the N-terminus of the 50-kDa subunit of the HPSE enzyme \[[@R56]\]. The antibody was diluted 1:100 in 10% goat serum in PBS. Control slides were incubated with 10% goat serum alone. Color was developed as described in \[[@R28]\] and slides were visualized with a Zeiss axioscope microscope and manually read by an expert pathologist. To define tumor as heparanase-positive, a cutoff point of 25% immunostained tumor cells was chosen on the basis of an initial overview of the cases, in order to improve signal-to-noise ratios. Cutoff was chosen before any attempt at correlating heparanase expression with lymph node involvement. Statistical analysis {#s4_7} -------------------- The results are presented as the mean ±SD. P values ≤0.05 were considered statistically significant. Statistical analysis was performed by unpaired Student\'s t-test unless otherwise stated. Fisher\'s exact test was performed to study the relationship between heparanase immunohistochemical results and clinical parameters. All statistical tests were two-sided. We are grateful to Sigma-Tau Research Switzerland S.A. (Mendrisio, CH) for kindly providing compound SST0001, and to Dr. Israel Vlodavsky (Rappoport Faculty of Medicine, Technion, Israel), for his continuous help and active collaboration. **CONFLICTS OF INTEREST** The authors declare that they have no conflict of interest. **GRANT SUPPORT** This work was supported by grants from the Israel Science Foundation (grant 806/14) and the Mizutani Foundation for Glycoscience.

All relevant data are within the manuscript and its Supporting Information files. Introduction {#sec001} ============ In the United States, there are over 40,000 reported cases of *Salmonella* infection in humans, and approximately 400 deaths are reported annually \[[@pone.0231998.ref001]\]. Consumption of contaminated poultry meat and egg products causes salmonellosis in humans \[[@pone.0231998.ref002]\]. Salmonellosis symptoms include stomach irritation accompanied by vomiting, diarrhea, high fever, and even death in humans, especially in immunocompromised individuals \[[@pone.0231998.ref003]\]. *Salmonella* enterica serovars Enteritidis (*S*. Enteritidis) and Heidelberg (*S*. Heidelberg) are among the most frequent serotypes recovered from humans each year \[[@pone.0231998.ref004]\]. *S*. Heidelberg has been isolated from poultry products in Brazil since 1962, while *S*. Enteritidis is a severe problem in poultry and public health since 1993 \[[@pone.0231998.ref005]\]. Upon ingestion, *Salmonella* will survive passage through the low-pH conditions of the stomach \[[@pone.0231998.ref006],[@pone.0231998.ref007]\], stimulate macrophages, and evade killing by the host immune system \[[@pone.0231998.ref008]\]. Within a few hours, *Salmonella* can invade the intestinal tract and reach the liver and spleen, and *Salmonella* can colonize the ceca in chickens \[[@pone.0231998.ref009]\]. Vaccination is one of the most promising control strategies for the reduction of *Salmonella* in chickens \[[@pone.0231998.ref010]\]. Cross-protection can enhance the clearance of pathogens through the acquired immune response \[[@pone.0231998.ref010],[@pone.0231998.ref011]\]. *Salmonella* serovars have conserved antigens and including conserved antigens in vaccines might induce cross-protection against multiple serovars \[[@pone.0231998.ref012]\]. Flagella \[[@pone.0231998.ref013]\] and outer membrane proteins (OMPs) \[[@pone.0231998.ref014]\] are conserved antigens among several *Salmonella* serovars that can induce an immune response in poultry. Killed vaccines are preferred over live vaccines to control *Salmonella* infections of poultry because live vaccines can regain its virulence \[[@pone.0231998.ref015],[@pone.0231998.ref016]\]. But killed vaccines have a disadvantage that killed vaccines need to be injected, which is time consuming and decreases the value of breast meat. Hence, oral administration of *Salmonella* vaccine is the preferred route \[[@pone.0231998.ref004]\] because oral administration mimics the natural infection and stimulates the mucosal and systemic immune responses \[[@pone.0231998.ref017]\]. However, oral killed *Salmonella* vaccines are not commercially available for broilers currently \[[@pone.0231998.ref018],[@pone.0231998.ref019]\]. This article studies a nanoparticle based oral vaccine for *Salmonella* control in broilers. Nanoparticle vaccines consist of a polymer coating that surrounds the vaccine antigen and protects the vaccine against chemical, enzymatic, and immunological degradation \[[@pone.0231998.ref020]--[@pone.0231998.ref022]\]. The prolonged survivability of the nanoparticles within the gastro intestinal tract (GIT) results in reducing the dosing frequency and the need for adjuvants \[[@pone.0231998.ref020]\], and facilitate the presentation of the vaccine antigens to mucosal immune cells \[[@pone.0231998.ref023]\]. Biodegradable chitosan nanoparticle (CNP) vaccines are ideal for delivering vaccine antigens through the oral route \[[@pone.0231998.ref023]--[@pone.0231998.ref025]\]. A *Salmonella* CNP vaccine was synthesized with a crude enriched OMPs and flagellin extracts from *S*. Enteritidis and surface-tagged with flagellin proteins. This research evaluated the protective effects of the synthesized CNP vaccine delivered through oral route in poultry by (1) Identifying the nanoparticle vaccine dose that can provide optimal immune response to *S*. Enteritidis infection, (2) Characterizing the CNP vaccine-induced anti-*Salmonella* OMPs and flagellar IgG and IgA specific antibodies in serum, cloacal swabs, and bile, (3) Identifying the effect of CNP vaccine on broilers performance parameters, pro- and anti-inflammatory cytokines, and (4) Evaluating the efficiency of the CNP vaccine on *S*. Enteritidis and *S*. Heidelberg loads in broiler birds challenged with *S*. Enteritidis and *S*. Heidelberg. Materials and methods {#sec002} ===================== Two experiments were conducted to characterize the CNP vaccine-induced immune responses in broilers (Cobb-Vantress hatchery, Inc. Cleveland, GA, USA). CNP vaccine was synthesized at the Food Animal Health Research Program, The Ohio State University, USA as described earlier \[[@pone.0231998.ref026]\]. Experiment I identified the optimal dose of CNP vaccine that would induce a protective response against *S*. Enteritidis, and the Experiment II identified the cross protection of CNP vaccine against *S*. Heidelberg. All animal protocols were approved by the Institutional Animal Care and Use Committee at the University of Georgia. Birds were monitored at least once a day for bloody feces, lethargy, refusal to eat food, loss of body weight., diarrhea, and dehydration during experiment I and II. Preparation of *S*. Enteritidis CNP vaccines {#sec003} -------------------------------------------- OMPs and flagellin proteins were isolated from *S*. Enteritidis as described earlier \[[@pone.0231998.ref026]\]. CNP vaccine were prepared using the ionic gelation method where nanoparticles form by intramolecular and intermolecular crosslinking between the positively charged chitosan and the negatively charged sodium tripolyphosphate (TPP) \[[@pone.0231998.ref026]\]. Briefly, 1% (w/v) low molecular weight chitosan (Sigma, MO) solution was prepared by slowly dissolving chitosan in an aqueous solution of 4% acetic acid. The solution was sonicated, and the pH was adjusted to 4.3. The solution was filtered through a 0.44 μm syringe filter. Five milliliters of 1% chitosan solution was added to 5 mL of deionized water and incubated with 2.5 mg OMPs and flagellar proteins. Subsequently, 2.5 mL of 1% (w/v) TPP (Sigma Aldrich, St. Louis, MO) in 2.5 mL deionized water was added to the above solution under magnetic stirring at room temperature to form the nanoparticles. Afterwards, 2.5 mg of flagellin protein in PBS was added to the nanoparticles to surface conjugate the nanoparticles with flagellin proteins. The CNP vaccines were collected by centrifugation at 10,000 X g for 30 minutes, lyophilized and stored at -80°C until further use. Experiment I {#sec004} ------------ Chicks in the experiment were confirmed to be *Salmonella* negative by streaking cloacal swabs on Xylose Lactose Tergitol^™^ 4 (XLT4) agar plates. A total of 14 one-day-old Cobb-500 chicks were distributed to one of the four treatment groups. The four treatment groups were control, 500μg, 1000μg, or 2000μg CNP vaccine groups. The control, 500μg, 1000μg, or 2000μg CNP vaccine groups had 5, 3, 3, and 3 birds on d1. At 1d of age, chickens were orally vaccinated with either 0.1 mL of PBS (Phosphate Buffered Saline; Control) or 500μg, 1000μg, or 2000μg CNP vaccine. A booster dose with the same amount of primary dose was given at d7. Birds were raised in individual battery cages and had access to *ad libitum* feed and water. At 14d of age, birds were challenged with 1 X 10^5^ CFU/bird of live *S*. Enteritidis orally. At d1, 7, 14, 20, and 25 body weight and feed consumption were recorded, and body weight gain and feed consumption ratio was calculated. At d1, 7, 14 (8h post-challenge), 17, 20, 23, and 25 (11d-post-challenge) of age, serum was collected to analyze anti-OMPs and anti-flagellin IgG and IgA. At d17, 20, 23 and 25 (11d-post-challenge) of age, cloacal swabs were collected to analyze anti-OMPs and anti-flagellin IgA. At d14 (8h-post-challenge) and 25 (11d-post-challenge) of age, bile was collected to analyze anti-OMPs and anti-flagellin IgA. At d25 (11d-post-challenge) of age, cecal tonsils, liver, and spleen samples were analyzed for IL-1β, IL-4, IFNγ and IL-10 cytokine mRNA amounts, macrophage nitrite production and *Salmonella* loads in the cecal content. Macrophage-nitrite production in CNP vaccine administered birds {#sec005} --------------------------------------------------------------- At d25 (11d-post-challenge) post-challenge, macrophages were collected from the spleen of one bird from each bird/treatment (n = 3) as described previously \[[@pone.0231998.ref027]\]. Single cell suspension of splenocytes were isolated using Histopaque (1.077 g/ml; Sigma Aldrich, St. Louis, MO). Briefly, 1 X 10^5^ splenocytes/well were plated in flat bottom 96 well plate (Greiner bio-one, NC) in triplicates. Cells were stimulated with 1 μg/mL *S*. Enteritidis lipopolysaccharide and incubated for 72 h at 37 °C in the presence of 5% CO~2~. Samples were centrifuged at 630 X g for 10 min at 4°C and 100 μL of the supernatants were removed. The nitrite content of the supernatant was determined using a sulfanilamide/N-(1-naphthyl) ethylenediamine dihydrochloride solution (Ricca Chemical Co., Arlington, TX) following the manufacturer's instructions. Nitrite concentrations were determined from a standard curve with sodium nitrite standards \[[@pone.0231998.ref028]\]. Antibody response in CNP-vaccine administered birds {#sec006} --------------------------------------------------- At d25 (11d-post-challenge), serum, cloacal swabs, and bile samples were collected one bird/pen (n = 3) and analyzed for anti-*Salmonella* OMPs and -flagellin IgG and IgA antibody response using an enzyme-linked immunosorbent assay (ELISA) as described previously \[[@pone.0231998.ref029]\]. High-binding-flat bottom 96-well plates (Greiner Bio-one, NC) were coated with either OMPs or flagellin (2 μg/mL for IgG and 7.5 μg/mL for IgA) diluted in 0.05 M sodium -bicarbonate coating buffer (9.6 pH), and incubated overnight at 4˚C. Plates were washed three times with PBS- Tween 20 (PBST) (0.05% Tween 20 in PBS, pH 7.4) and blocked with 5% non-fat dry milk powder in PBST for 1h at room temperature. Plates were washed three times with PBST. For analysis, 50 μl of serum and bile samples were diluted in 2.5% non-fat dry milk, and 50 μl of cloacal supernatants were added to the wells in triplicates. Samples were incubated for 2h at room temperature. Plates were washed three times with PBST. HRP-conjugated goat anti-chicken IgG (Southern Biotech, AL) (1: 10,000 in 2.5% non-fat dry milk powder in PBST) or HRP-conjugated goat anti-chicken IgA (Gallus immunotech, NC) (1: 3000 in 2.5% skim milk powder in PBST) secondary antibodies (50 μL/well) were added and incubated for 2h at room temperature. Plates were washed three times, and 50 μl/well of TMB peroxidase substrate was added. The reaction was stopped after 6 min by adding 2M sulfuric acid. The OD was measured at 450 nm using Gen5TM software (BioTek,VT,USA). IgG and IgA values were reported as the mean optical density. The corrected OD was obtained by subtracting the treatment group OD from blank control OD. Cytokine gene expression in the cecal tonsils, liver, and spleen of CNP-vaccine administered birds {#sec007} -------------------------------------------------------------------------------------------------- Total RNA was extracted using TRI reagent (Molecular Research Center, Cincinnati, OH) following the manufacturer's instructions. The RNA was reverse-transcribed into cDNA and analyzed for IL-1β, IFNγ, IL-10, and IL-4, mRNA by real-time PCR (CFX96 Touch Real Time System, BioRad) using SyBr green after normalizing for RPS13 and GAPDH. Fold change from the reference was calculated, as explained previously \[[@pone.0231998.ref030]\]. Primers sequences are described in [Table 1](#pone.0231998.t001){ref-type="table"}. 10.1371/journal.pone.0231998.t001 ###### Real-time PCR primers for cytokine analysis. {#pone.0231998.t001g} Primer Sequence (5'-- 3') T~a~ ------- -------- ------------------------------ -------- GAPDH F `TCCTGTGACTTCAATGGTGA` 55.0°C R `CACAACACGGTTGCTGTATC` RPS13 F `CAAGAAGGCTGTTGCTGTTCG` 55.5°C R `GGCAGAAGCTGTCGATGATT` IL-1β F `TCCTCCAGCCAGAAAGTGA` 57.0°C R `CAGGCGGTAGAAGATGAAGC` IL-4 F `AACATGCGTCAGCTCCTGAAT` 57.5°C R `TCTGCTAGGAACTTCTCCATTGAA` IFNγ F `GTGAAGAAGGTGAAAGATATCATGGA` 57.0°C R `GCTTTGCGCTGGATTCTCA` IL-10 F `GAGGAGCAAAGCCATCAAGC` 57.5°C R `CTCCTCATCAGCAGGTACTCC` Primers were adapted from previous literature. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH), Dube et al. 2014 \[[@pone.0231998.ref031]\]; Ribosomal protein S13 (RPS13), Shanmugasundaram et al. 2018; IL-1β, IFNγ, and IL-10, Shanmugasundaram et al. 2019 \[[@pone.0231998.ref032]\]; IL-4, Renu et al. 2018 \[[@pone.0231998.ref029]\]. Experiment II {#sec008} ------------- Experiment II studied the cross-protective effect of synthesized CNP vaccine against *S*. Heidelberg infection. A total of 216 Cobb-500 one-day-old chicks were randomly assigned to six treatments. Each treatment groups had 36 birds in 6 pens (n = 6) with 6 birds/pen. The six treatments involved in the study were distributed as a 3 (No Vaccine, CNP vaccine and Commercial vaccine) X 2 (*S*. Enteritidis and *S*. Heidelberg challenge) factorial set up of treatments. CNP vaccine groups received 1000μg of CNP vaccine in 100μL of PBS at d1 followed by the same dose as a booster dose on d7 orally. Commercial vaccine group received 100 μl of live *Salmonell*a vaccine (Poulvac ST, Zoetis, NJ) at d1 followed by the same dose as a booster dose on d7 orally. The birds in the control group received 100μl of PBS orally. *S*. Enteritidis or *S*. Heidelberg challenge treatments were conducted in two different rooms to avoid cross-infection. Birds had access to *ad libitum* feed and water. At d1, 7, 14, and 18 body weight and feed consumption were recorded, and body weight gain and feed consumption ratio was calculated. At 14d of age, birds were challenged with an oral gavage with 1 X 10^5^ CFU/bird of live *S*. Enteritidis or *S*. Heidelberg. At d1, 7, 14 (8h-post-challenge), 16, and 18 of age, serum was collected to analyze anti-OMPs and anti-flagellin IgG. At d1, 7, 14 (8h-post-challenge), 16, and 18 of age, cloacal swab samples were collected to analyze anti-OMPs and anti-flagellin IgA. At d18 (4d-post-challenge) of age, bile samples were collected to analyze anti-OMPs and anti-flagellin IgA. At d18 (4 post-challenge) cecal tonsils were collected and analyzed for IL-1β and IL-10 cytokine mRNA amounts. At d16 and 18 (2 and 4 post-challenge), cecal content liver and spleen were collected and analyzed for *Salmonella* loads. *Salmonella* loads in the ceca of CNP-vaccine administered birds {#sec009} ---------------------------------------------------------------- Cecal content was collected from one bird from each pen (n = 6) and analyzed for *S*. Enteritidis loads by real time PCR. Bacterial genomic DNA was isolated as described earlier \[[@pone.0231998.ref030]\]. Cecal samples (200 mg) were washed two times with 1X PBS. The cell pellet was resuspended in EDTA and treated with 20 mg/ml lysozyme for 30 min at 37°C. The samples were treated with lysis buffer containing 20% SDS and 0.1 mg/ml proteinase K (Sigma Aldrich, St Louis, MO) for 5 min at -80°C. The samples were incubated with 5μL of RNase at 37°C for 30 min. The cell lysate was incubated with 6M sodium chloride on ice for 10 min. The supernatant was collected after centrifugation at 400 X g for 10 min. The DNA in the supernatant was precipitated with isopropanol and washed once in ice-cold ethanol. The DNA pellet was resuspended in TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0) and stored at -20°C until further use. The DNA extracted from the different treatment groups was analyzed for *S*. Enteritidis loads by real-time PCR using *Salmonella Enteritidis* primers 5'-`GCAGCGGTTACTATTGCAGC`-3' and 5'-`CTGTGACAGGGACATTTAGCG`-3 \[[@pone.0231998.ref033]\]. The threshold cycle (Ct) values were determined by iQ5 software (Bio-Rad, Hercules, CA) when the fluorescence rises exponentially 2-fold above background. The copy numbers of *S*. Enteritidis was expressed in log units as described earlier \[[@pone.0231998.ref034]\]. *Salmonella* loads in the liver and spleen of CNP-vaccine administered birds {#sec010} ---------------------------------------------------------------------------- *Salmonella* loads in the liver and spleen samples was determined by a 3-tube Most Probable Number (MPN) method. The samples were homogenized in 1: 2 (W/V) of buffered peptone water (BPW). Samples were prepared in 10-fold dilution series, and then 1mL of each dilution was inoculated into triplicate broth culture tubes for incubation at 42°C for 24 h. Following incubation, all tubes are examined for turbidity, and the pattern of growth in the tubes was scored against an MPN table from the U.S. Food and Drug Administration's Bacterial Analytical Manual \[[@pone.0231998.ref035]\]. The bacteria in the broth were confirmed to be *Salmonella* by plating one hundred microliters of each positive culture tube onto XLT4 Agar plates (25 mg/mL novobiocin) and incubated for 24h at 42°C for confirmation of black colonies and MPN index/g. Samples were bacterial growth and had no black colonies on the XLT4 agar plates were discarded as false positives and assigned an MPN value of 0. ### Statistical analyses {#sec011} All data in Experiment I and II were examined by one-way ANOVA (JMP, SAS Institute Inc., Cary, NC), to examine the effect of CNP-*Salmonella* vaccine during *S*. Enteritidis and *S*. Heidelberg. When the treatment effects were significant (*P\<*0.05) differences between means were analyzed using Tukey's test. MPN-non-parametric data was transformed to Log10 before statistical analysis. Results and discussion {#sec012} ====================== Effect of CNP vaccine on bird production performance {#sec013} ---------------------------------------------------- In both experiments, there were no significant differences in body weight gain and feed conversion ratio of birds in different treatment groups indicating that CNP vaccine administration had no adverse effects on broilers' performance parameters ([S1 Table](#pone.0231998.s001){ref-type="supplementary-material"}). Effect of CNP vaccine on macrophage-nitrite production {#sec014} ------------------------------------------------------ In Experiment I, birds vaccinated with 500μg CNP vaccine had higher levels (*P\<*0.05) of nitrite compared to PBS treatment group ([Fig 1](#pone.0231998.g001){ref-type="fig"}). Birds vaccinated with 1000μg, and 2000μg CNP vaccine had non-significant higher levels of nitrite compared to PBS treatment group. Within the CNP vaccine administered group, increasing the CNP vaccine dose from 500 to 2000μg decreased the macrophage-nitrite content. Previous *in vitro* studies have identified that *S*. Enteritidis infection suppresses nitirite production in chicken macrophage HD11-cells \[[@pone.0231998.ref036]\], which can be expected to facilitate the pathogen escapting the host immune response. However, inflammatory responses, such as nitric oxide production, can decrease production performance in broilers \[[@pone.0231998.ref037]\]. Birds adminstered with 1000μg CNP vaccine had the highest amount of nitirte suggestingt that 1000μg dose can provide protection against *Salmonella* challenge. {#pone.0231998.g001} Effect of CNP vaccine administration on anti-OMPs and--flagellin IgG and IgA antibody response {#sec015} ---------------------------------------------------------------------------------------------- In Experiment I, at 8h- post-challenge broilers that were vaccinated with 1000μg CNP dose and orally challenged with *S*. Enteritidis had 39% higher (*P\<*0.05) serum anti-OMP IgG compared to that in the control group ([Fig 2A](#pone.0231998.g002){ref-type="fig"}). At 3d-post-challenge, birds vaccinated with 500μg and 2000μg CNP vaccine dose had higher (*P\<*0.05) serum anti-OMP IgG compared to that in the control group ([Fig 2A](#pone.0231998.g002){ref-type="fig"}). These results demonstrated that the CNP vaccine at doses above 500μg can induce an anti-*Salmonella* antigen-specific immune response in broilers. Similar results were observed in a recent study, where layer chickens that were orally immunized with a *Salmonella* polyanhydride nanoparticle (PNP) vaccine and challenged with *S*. Enteritidis had substantially higher OMPs-specific IgG response in the serum \[[@pone.0231998.ref029]\]. {#pone.0231998.g002} At 8h-post-challenge, birds vaccinated with 1000μg CNP had 185% higher (*P\<*0.05) anti- OMPs serum IgA amounts compared to that in the control-challenge groups ([Fig 2B](#pone.0231998.g002){ref-type="fig"}). At 8h-post-challenge, birds vaccinated with 500μg and 2000μg *Salmonella*-CNP vaccine had similar serum anti-OMPs IgA levels. At 11d-post-challenge, birds vaccinated with 1000μg CNP had 73% higher bile anti-OMPs IgA ([Fig 2B](#pone.0231998.g002){ref-type="fig"}) and 12% higher anti-flagellin IgA ([Fig 2C](#pone.0231998.g002){ref-type="fig"}), compared to that from the control group (*P\<*0.05). At 8h-post-challenge, birds vaccinated with 1000μg CNP had 208% higher cloacal swabs anti- OMPs IgA (*P\<*0.05), compared to that from the control group ([Fig 2B](#pone.0231998.g002){ref-type="fig"}). However, at 8h-post-challenge birds vaccinated with 500μg CNP had higher cloacal swabs anti-*S*. Enteritidis flagellin IgA (*P\<*0.05), compared to that from the control ([Fig 2C](#pone.0231998.g002){ref-type="fig"}). At d25 (11d-post-challenge), birds vaccinated with 2000μg CNP had lower serum anti-OMPs IgG (*P\<*0.05), compared to that from the control group ([Fig 2A](#pone.0231998.g002){ref-type="fig"}). Experiment I results identified that 1000μg CNP vaccine induced *Salmonella* antigen-specific immune response against *S*. Enteritidis in broilers and hence 1000μg CNP vaccine dose was further studied in Experiment II. f In Experiment II, at 8h-post-challenge, birds vaccinated with the commercial vaccine or the CNP had 253% and 173% higher (*P\<*0.05) anti-OMPs serum IgG amounts, respectively, compared to that in the no-vaccine control ([Fig 3A](#pone.0231998.g003){ref-type="fig"}). At 4d-post-challenge, birds vaccinated with the commercial vaccine or the CNP had 72.63% and 72.62% lower (P\<0.05) anti-OMPs serum IgG amounts, respectively, compared to that in the control ([Fig 3A](#pone.0231998.g003){ref-type="fig"}) group. These results identified that vaccinating birds with the CNP vaccine induces a high titer of *Salmonella* specific protective antibodies and in the absence of infection, the antibody titers will eventually decline over time \[[@pone.0231998.ref038]\]. In contrast, a carrier animal will usually maintain persistently high antibody levels in the blood, indicative of a continuing infection \[[@pone.0231998.ref039]\], as previously documented with *S*. Dublin in cattle \[[@pone.0231998.ref039]\]. Similarly, at 2d-post-*S*. Enteritidis challenge, birds vaccinated with the commercial vaccine or CNP vaccine had 51.42% and 66.5% lower (*P\<*0.05) cloacal swabs anti- OMPs IgA amounts, respectively, compared to that in the control group ([Fig 3B](#pone.0231998.g003){ref-type="fig"}). At 4d-post-*S*. Enteritidis challenge, birds vaccinated with the commercial vaccine or CNP vaccine had 65.13% and 59% lower (*P\<*0.05) cloacal swabs anti-OMPs IgA amounts, respectively, compared to that in the control ([Fig 3B](#pone.0231998.g003){ref-type="fig"}). {#pone.0231998.g003} Results for *S*. Heidelberg showed that at d7, birds vaccinated with commercial vaccine or CNP had 51.7% and 46.3% lower (*P\<*0.05) cloacal anti-*S*. Heidelberg OMPs IgA amounts, respectively, compared to that in the control group ([Fig 3B](#pone.0231998.g003){ref-type="fig"}). However, after adminstrating the booseter dose, there was no differences in the cloacal anti-OMPs IgA amounts among all treatment groups ([Fig 3B](#pone.0231998.g003){ref-type="fig"}). A possible explanation for this could be a) the transfer of maternal antibodies \[[@pone.0231998.ref040]\] or b) pre-existing *Salmonella* colonization from exposure to exogenous microbes either in the hatchery or farm environment \[[@pone.0231998.ref041]\]. It has been reported that within broiler hatcheries, 74% of pad samples placed under newly hatched chicks were positive for *Salmonella* \[[@pone.0231998.ref042]\]. At 2d-post-*S*. Heidelberg challenge, birds vaccinated with the commercial vaccine or CNP vaccine had 71.2% and 65.1% lower (P\<0.05) cloacal anti-*S*. Heidelberg OMPs IgA amounts, respectively, compared to that in the control group ([Fig 3B](#pone.0231998.g003){ref-type="fig"}). At 4d-post-*S*. Heidelberg challenge, birds vaccinated with the commerical vaccine or the CNP vaccine had 66.3% and 71.8% lower (P\<0.05) cloacal anti-*S*. Heidelberg OMPs IgA amounts, respectively, compared to that in the control group ([Fig 3B](#pone.0231998.g003){ref-type="fig"}). At 4d-post-*S*. Heidelberg challenge, birds vaccinated with the commerical vaccine or CNP vaccine had 64.36% and 51.2% lower (*P\<*0.05) cloacal anti-*S*. Heidelberg flagellin IgA amounts, respectively, compared to that in the control group ([Fig 3C](#pone.0231998.g003){ref-type="fig"}). In addition, at 8h-post-*S*. Heidelberg challenge, birds vaccinated with the commerical vaccine or CNP vaccine had 6.36% and 4.59% higher (*P\<0*.05) bile anti-*S*. Heidelberg flagellin IgA amounts, respectively, compared to that in the control group ([Fig 3C](#pone.0231998.g003){ref-type="fig"}). At 8h-post-*S*. Heidelberg challenge, birds vaccinated with the commerical vaccine or CNP vaccine had 253% and 173% higher (*P\<0*.05) serum anti-*S*. Heidelberg OMPs IgG amounts, respectively, compared to that in the control group ([Fig 3A](#pone.0231998.g003){ref-type="fig"}). These results demonstrate that CNP can elicit an antigen-specfic and cross-protective immune response against *S*. Heidelberg in broilers. Effect of CNP vaccine on *S*. Enteritidis loads in ceca at 2d-post-challenge {#sec016} ---------------------------------------------------------------------------- *Salmonella* Enteritidis frequently colonizes the GIT of poultry \[[@pone.0231998.ref002]\]; hence *Salmonella* loads in cecal contents was quantified by real-time qPCR. In Experiment I, there were no significant differences in *S*. Enteritidis loads in cecal content of birds in different treatment groups at d25 (11d-post-challenge) (*P\>*0.05). In Experiment II, *S*. Heidelberg and *S*. Enteritidis loads in liver and spleen was analyzed using MPN method while *S*. Heidelberg and *S*. Enteritidis loads in the cecal content was analyzed by real-time qPCR. At 4d post-challenge, there were no significant differences (*P\>*0.05) in *S*. Heidelberg loads in the liver, spleen (Tables [2](#pone.0231998.t002){ref-type="table"} and [3](#pone.0231998.t003){ref-type="table"}) or ceca among birds in different treatment groups. However, *S*. Heidelberg loads in liver and spleen was numerically lower in birds vaccianted with commerical vaccine and CNP vaccine compared to that in the control group. 10.1371/journal.pone.0231998.t002 ###### *S*. Heidelberg loads in liver of birds in different treatment groups at 4d-post-*S*. Heidelberg challenge. {#pone.0231998.t002g} Sample 100 μl 10 μl 1 μl Colonies counted (10^−5^) MPN index/g Log10 Average MPN value SEM -------------------- ----------- ------- ------ --------------------------- ------------- ------- ------------------- ------ Control 2 1 0 0 0 0 0.57 0.36 3 0 2 63 64 1.8 3 0 0 0 0 0 2 2 1 0 0 0 2 2 2 0 0 0 3 0 1 37 38 1.6 CNP 1 2 1 13 15 1.2 0.20 0.20 2 0 0 0 0 0 1 0 0 0 0 0 1 1 0 0 0 0 2 0 2 0 0 0 3 0 0 0 0 0 Commercial Vaccine 1 1 0 0 0 0 0.22 0.22 0 1 0 0 0 0 1 1 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 2 0 2 20 20 1.3 *P-*value *P\>*0.05 At 1d of age, chickens were orally vaccinated with either PBS (NV), 1000 μg chitosan nanoparticle vaccine (CNP) or a commercial vaccine (CV). The same route of delivery and doses was repeated at 7d of age. At 14d of age, birds were challenged with 1 X 10^5^ CFU/bird of either *S*. Enteritidis or *S*. Heidelberg. At 4d-post-challenge *Salmonella* loads in the liver samples was determined by the MPN method. Samples with no bacterial growth were assigned an MPN of 0. Chitosan Nanoparticle (CNP). n = 6. Means with no common superscript differ (*P\<*0.05). 10.1371/journal.pone.0231998.t003 ###### *S*. Heidelberg loads in spleen of birds in different treatment groups at 4d-post-*S*. Heidelberg challenge. {#pone.0231998.t003g} Sample 100 μl 10 μl 1 μl Colonies counted (10^−5^) MPN index/g Log10 Average MPN value SEM --------------------- ----------- ------- ------ --------------------------- ------------- ------- ------------------- ------ Control 3 3 1 0 0 0 1.15 0.38 3 1 2 25 120 2.1 3 1 1 4 75 1.9 2 2 2 15 35 1.5 2 2 2 0 0 0 3 0 0 14 23 1.4 CNP 3 2 1 9 150 2.2 0.58 0.39 2 2 0 4 21 1.3 1 0 0 0 0 0 3 2 1 0 0 0 2 2 2 0 0 0 2 0 0 0 0 0 Commericial Vaccine 1 1 1 2 11 1.0 0.57 0.40 2 1 0 0 0 0 1 1 0 0 0 0 1 0 0 0 0 0 3 3 0 12 240 2.4 2 0 0 0 0 0 *P-*value *P\>*0.05 At 1d of age, chickens were orally vaccinated with either PBS (NV), 1000 μg chitosan nanoparticle vaccine (CNP) or a commercial vaccine (CV). The same route of delivery and doses was repeated at 7d of age. At 14d of age, birds were challenged with 1 X 10^5^ CFU/bird of either *S*. Enteritidis or *S*. Heidelberg. At 4d-post-challenge *S*. Heidelberg loads in the spleen samples was determined by the MPN method. Samples with no bacterial growth were assigned an MPN of 0. Chitosan Nanoparticle (CNP). n = 6. Means with no common superscript differ (*P\<*0.05). In Experiment II, at d18 (4d-post-challenge) there was no *S*. Enteritidis colonization in the liver and spleen. At d16 (2d-post-challenge) birds vaccinated with commerical vaccine or CNP vaccine had lower (*P\<*0.05) *Salmonella* load by 82.7% and 65.9%, respectively, when compared to that in the control group ([Fig 4](#pone.0231998.g004){ref-type="fig"}). Similar results were observed in a recent 2018 study, as a PNP vaccination cleared *Salmonella* in the ceca of 33% of birds in the study \[[@pone.0231998.ref029]\]. This identified that the synthesized vaccine decreased *Samonell*a loads in the ceca of broilers. {#pone.0231998.g004} Effect of CNP vaccine on cytokine gene expression by qPCR {#sec017} --------------------------------------------------------- The effect of CNP vaccine on four key cytokines that influence both innate and adaptive immune responses, IL-1β ([Fig 5A](#pone.0231998.g005){ref-type="fig"}), IL-10 ([Fig 5B](#pone.0231998.g005){ref-type="fig"}), IL-4 ([Fig 5C](#pone.0231998.g005){ref-type="fig"}), and IFNγ ([Fig 5D](#pone.0231998.g005){ref-type="fig"}) were studied. In Experiment I, liver and spleen samples were analyzed to identify any adverse effects of CNP vaccine on liver and spleen, as *S*. Enteritidis colonizes in liver and spleen at 1d post- inoculation \[[@pone.0231998.ref043],[@pone.0231998.ref044]\]. {#pone.0231998.g005} In Experiment I, at d25 (11d-post-challenge), birds immunized with 500μg, 1000μg or 2000μg CNP had no effect on cecal tonsil IL-4 and IFNγ mRNA amounts (*P*\>0.05), but had higher IL-1β mRNA amounts in cecal tonsils, spleen, and liver samples (*P\<*0.05). Cecal tonsil IL-1β mRNA amount was 10-fold lower (*P\<*0.05) in the 500μg CNP-vaccinated birds while the 1000μg and 2000μg CNP vaccinated birds had two-fold higher (*P\<*0.05) IL-1β, compared to that in the control group. Spleen IL-1β mRNA amount was two-fold higher (*P\<*0.05) in the 1000μg CNP-vaccinated birds, compared to that in the control group. Liver IL-1β mRNA amount in the 500μg CNP-vaccinated group was 2.6-fold higher (*P\<*0.05), compared to that in the control group, while the 1000μg-vaccinated birds had 0.7-fold lower (*P\<*0.05) IL-1β mRNA amounts compared to that in the control group. A possible explanation for increased mRNA content of pro-inflammatory cytokine IL-1β is that the adjuvant composition of the vaccine can induce a predominantly Th1 type inflammatory response \[[@pone.0231998.ref026]\]. The only vaccine dose that increased the anti-inflammatory cytokine IL-10 amounts was the 1000μg CNP vaccine. Cecal tonsil IL-10 mRNA content was two-fold lower in the 500μg CNP-vaccinated birds, while cecal tonsil IL-10 mRNA content was 3.3-fold higher (*P\<*0.05) in the 1000μg CNP treatment groups, compared to that in the control group. IL-10 cytokine reduces host tissue damage in response to inflammation caused by bacterial infections \[[@pone.0231998.ref045]\] and hence 1000μg CNP vaccine dose was chosen for Experiment II. Results show that birds vaccinated with 500μg, 1000μg or 2000μg CNP had higher spleen IL-4 mRNA amounts, compared to that of the control (*P\<*0.05). Liver IL-4 mRNA amounts were three-fold lower in the 1000μg vaccinated group, compared to that in the control group (*P\<*0.05). IL-4, a Th2 cytokine, can decrease the production of IFNγ cytokine and decrease the activation of macrophages \[[@pone.0231998.ref038]\]. However, no significant differences were observed on IFNγ mRNA content among birds in different treatment groups. At d25 (11d-post-challenge), birds vaccinated with 500μg, 1000μg or 2000μg CNP had no differences in spleen and liver IL-10 and IFNγ mRNA amounts (*P\>*0.05), compared to that in the control groups. In Experiment II, at d18 (4d-post-*S*. Heidelberg or *S*. Enteritidis challenge) there were no significant differences in cecal tonsil IL-1β or IL-10 mRNA amounts (*P\>*0.05). Results from Experiment I and II demonstrated that the CNP vaccine had no adverse effects on the bird's health. Conclusion {#sec018} ========== The vaccine under study has previously shown to induce substantially higher antigen-specific IgA response in bile, serum, cloacal swab, and tracheal wash samples of layer birds \[[@pone.0231998.ref026]\]. This study identified that the 1000μg CNP-*Salmonell*a vaccine can substantially increase antigen-specific anti-*Salmonella* IgG and IgA amounts in broilers infected with either *S*. Enteritidis or *S*. Heidelberg, providing cross-protection against both *Salmonella* enterica serovars. The 1000μg CNP vaccine did not affect the production performance of birds and induced substantial IL-1β and IL-10 cytokines, and nitrite productuon in response to *S*. Enteritidis infection, thereby identifying the CNP vaccine as a potential vaccine candidate. The 1000μg CNP vaccine significantly decreased the cecal colonization of *S*. Enteritidis as well as *S*. Heidelberg in vacinated broilers, idnetifying the CNP vaccine as a potential candicate that can mitigate *Salmonella* loads. Results show that the CNP vaccine is a potential vaccine candidate that can be deliveed orally to control *Salmonella* infections in poultry. Supporting information {#sec019} ====================== ###### Effect of CNP vaccine on performance parameters in *Salmonella* challenged birds. A\) Final body weight gain (BWG) and feed consumption ratio (FCR) of d25 broilers in Experiment I. At 1d and 7d of age, chickens were orally vaccinated with 0.1 mL of PBS (control-challenge) or 500μg, 1000μg, or 2000μg CNP vaccine. At 14d of age, birds were challenged using an oral gavage with 1 X 10^5^ CFU/bird of *S*. Enteritidis. BWG and FCR was calculated on d25 of age. B) Final BWG and FCR of d18 broilers in Experiment II---At 1d and 7d of age, chickens were orally vaccinated with 0.1 mL of PBS (control-challenge) or 1000μg CNP vaccine, or a live commercial *Salmonell*a vaccine at d1 and d7. At 14d of age, birds were challenged using an oral gavage with 1 X 10^5^ CFU/bird of either live *S*. Enteritidis or live *S*. Heidelberg. BWG and FCR was calculated on d18 of age. n = 6. Means (SEM) with no common superscript differ (*P\<*0.05). (PDF) ###### Click here for additional data file. We acknowledge Dr. Woo Kim and Dr. Manpreet Singh, from the University of Georgia, for their assistance throughout the research. We also acknowledge every member of the Selvaraj laboratory, for their hard work and support: Theros Ng, Mohamad Mortada, Gabriel Akerele, Jarred Oxford, Nour Ramadan, and Ragini Reddyvari. [^1]: **Competing Interests:**The authors have declared that no competing interests exist.

Introduction {#s1} ============ Cellular proteins are subject to oxidative damage and must either be repaired or removed to maintain protein homeostasis. Cells possess two main mechanisms to deal with damaged protein. The first mechanism is the chaperone system. Chaperones such as Hsp70 and Hsp90 bind to exposed hydrophobic regions of misfolded proteins and assist in the refolding process. Secondly, if refolding is unsuccessful, the misfolded protein is removed by the cellular proteolytic systems. The main system for removal of damaged and/or misfolded proteins is the proteasome, although autophagy also plays a role. If either the chaperone system or degradation pathways are perturbed, damaged protein may accumulate and start to form aggregates. Aggregated protein has been implicated in age-related neurodegenerative disorders although there is still controversy over whether it is a cause or consequence of the disease process [@pone.0022038-Lansbury1]. Small aggregates bind to proteasomes but are unable to be degraded and so inhibit proteasomal function [@pone.0022038-Bence1]. This will lead to a reduction in the clearance of misfolded protein and a vicious cycle might ensue. Levels of reactive oxygen species (ROS) increase with age due to either an increase in dysfunctional mitochondria or a decline in the antioxidant system [@pone.0022038-Cadenas1], [@pone.0022038-Zhu1] leading to increase damage to cellular components. In addition it has been observed that there is a progressive decline in the overall proteolytic capacity of the cell with age [@pone.0022038-Friguet1], [@pone.0022038-Szweda1] resulting in an accumulation of oxidized and cross-linked proteins [@pone.0022038-Sitte1], [@pone.0022038-Sitte2]. Chaperone function has also been shown to decline with age [@pone.0022038-Nardai1], [@pone.0022038-Sti1]. There are several links between the chaperone and degradation pathways. For example, the molecular co-chaperone CHIP (C terminus of Hsc70-interacting protein, (Q9UNE7)) serves as a catalyst for the ubiquitination of several Hsp70 and Hsp90 client proteins that have been linked to neurodegenerative disorders [@pone.0022038-Dickey1]. Therefore, interventions which target part of one pathway will also impact the rest of the system. A stochastic model of the chaperone Hsp90 was previously developed to examine the role of chaperones in the ageing process [@pone.0022038-Proctor1]. This model showed that under conditions of low or transient stress, chaperone capacity is sufficient to maintain protein homeostasis. Even under conditions of increasing stress with age, normal chaperone capacity is able to deal with the increasing overload due to the mechanism of upregulation of Hsps after stress. However, if the model includes a decline in chaperone function with age, then the system becomes overwhelmed and protein homeostasis is lost. This model only included Hsp90 and one client of Hsp90, namely Heat Shock Factor-1 (Hsf1, (Q00613)) but was constructed in the Systems Biology Mark-up Language (SBML) [@pone.0022038-Hucka1] which allows easy extension as required. The role of apoptosis in ageing has received considerable interest but the exact relationship between ageing and apoptosis has yet to be established [@pone.0022038-Zheng1]. Apoptosis has a critical role in tissue homeostasis, and in mitotic tissue is important for preventing tumorigenesis. In post-mitotic tissue, ageing is associated with increased apoptosis to remove dysfunctional cells and since these cells are irreplaceable, this will have an impact on tissue function. For example, neuronal loss is a common feature of neurodegeneration and results in pathology such as memory loss, movement disorders and/or visual impairment, depending on the specific part of the brain which is affected. Therefore it is important to extend the model to include apoptotic pathways. Many chaperones are involved in apoptotic pathways such as Hsp27 and Hsp70 which are antiapoptotic. There are many pathways involved in apoptosis but our model will concentrate on three pathways that are particularly relevant to neurodegeneration. The first pathway involves the c-Jun N-terminal kinase (JNK). JNK (MAPK8) is activated by phosphorylation and in turn, JNK phosphorylates a number of proteins involved in various signalling pathways including apoptosis. It is maintained in an inactive state by the phosphatase Mkp1 (DUSP1, (P28562)) which requires Hsp70 for its activation. When levels of free Hsp70 are high, Mkp1 is activated and JNK activation is inhibited. Under conditions of stress, misfolded protein binds to Hsp70, diminishing pools of free Hsp70, so less Mkp1 is activated. Therefore, JNK remains phosphorylated and can induce one of the signalling pathways which lead to apoptosis. In this model we assume that the probability of cell death depends on the level of activated JNK, with the understanding that future models could incorporate more additional regulatory elements. Upregulation of Hsp70 by the stress response maintains pools of free Hsp70 to prevent activation of this apoptotic pathway. Mkp1 is prone to oxidation in its catalytic domain resulting in inactivation of its phosphatase activity [@pone.0022038-Denu1]. This means that ROS elevation has opposing effects on Mkp1 activity. On the one hand, increased pools of Hsp70 leads to activation of Mkp1, but on the other hand oxidation of Mkp1 by ROS leads to its inactivation. These opposing effects of ROS are incorporated into the model. The second apoptotic pathway considered involves the mitogen-activated protein kinase p38 (MAPK14). It has been shown that p38 is activated by an increase in levels of ROS [@pone.0022038-Passos1], and that p38 is also involved in generating ROS, so providing a positive feedback loop [@pone.0022038-Passos1], [@pone.0022038-Koli1], [@pone.0022038-Torres1]. Similarly to JNK, p38 is de-phosphorylated by Mkp1 to keep it in an inactive state. As for JNK, we model this apoptotic pathway by assuming that the probability of cell death depends on the level of activated p38. The third apoptotic pathway that we include involves the transcription factor p53 (P04637) which is normally present in low levels as it is rapidly turned over by the ubiquitin-proteasome pathway. However, if the proteasome is inhibited, levels of p53 will rise and either cell cycle arrest (relevant only for dividing cells) or apoptosis takes place. We do not include details of the p53 apoptotic pathway in this model as this pathway is very complex and would also have needed to include detail of p53 turnover. This extra complexity would have made stochastic simulations and model analysis impractical to do in terms of the computer simulation time required. Therefore, for simplicity we assume that the probability of cell death depends on the level of aggregated protein inhibiting the proteasome. Akt (P31749) is a kinase which plays an important role in cellular signalling and it is involved in survival pathways. It is a client of Hsp90 and is destabilised by inhibitors of Hsp90 function [@pone.0022038-Basso1]. It phosphorylates glycogen synthase kinase-3 beta (GSK3β, (P49841)) [@pone.0022038-Cross1], one of the main kinases responsible for tau (P10636) hyperphosphorylation. A model of tau aggregation is also being developed with the intention of linking it to this model of chaperones and so we include Akt as a common link for these models. Hsp90 inhibitors, such as geldamycin, bind tightly to the Hsp90 ATP/ADP pocket and prevent ATP binding and the completion of client protein refolding. Instead proteins are degraded. For example, Basso *et al.* [@pone.0022038-Basso1] found that Hsp90 inhibitors do not alter association of Akt with Hsp90 but result in ubiquitination of Akt and its subsequent degradation. They found that the half-life of Akt was shortened from 36 hours to 12 hours in cells exposed to the Hsp90 inhibitor 17-AAG. Here we have extended the chaperone model of Proctor *et al.* [@pone.0022038-Proctor1] to include details of Hsp70 and its role in apoptosis, as well as its chaperoning activity. We also include detail of some of the client proteins of Hsp70 and Hsp90. We hypothesize that: (1) under normal conditions, basal levels of chaperones are able to maintain protein homeostasis; (2) under conditions of low or moderate stress, there is a transient increase in protein misfolding but chaperones are upregulated and protein homeostasis is restored; and (3) under conditions of prolonged high stress, the chaperone system is overwhelmed and apoptosis takes place. We now give more details of each of these scenarios. (1) Normal conditions -unstressed cells {#s1a} --------------------------------------- Under normal cellular conditions ([Figure 1](#pone-0022038-g001){ref-type="fig"}) there are very low levels of damaged/misfolded protein which are either refolded via the chaperone system or eliminated from the cell via the ubiquitin-proteasome system. The majority of Hsf1 is bound to Hsp90 which prevents Hsf1 from becoming transcriptionally active. There is basal transcription of heat shock proteins to maintain pools at a steady state level but there is no upregulation of Hsps. JNK and p38 are maintained in their unphosphorylated state due to activity of the phosphatase Mkp1, which requires Hsp70 for its activity. Since there are sufficient pools of free Hsp70, apoptosis is inhibited. [Figure 1](#pone-0022038-g001){ref-type="fig"} omits detail of JNK for clarity. (Note that [Figures 1](#pone-0022038-g001){ref-type="fig"}, [2](#pone-0022038-g002){ref-type="fig"}, [3](#pone-0022038-g003){ref-type="fig"} are biological diagrams that summarise the key components of the model. A diagram of the full model network is shown in [Figure S1](#pone.0022038.s001){ref-type="supplementary-material"}). {#pone-0022038-g001} {#pone-0022038-g002} {#pone-0022038-g003} (2) Low -- moderate levels of stress {#s1b} ------------------------------------ When cells are stressed ([Figure 2](#pone-0022038-g002){ref-type="fig"}), damaged/misfolded proteins accumulate and bind to Hsp70 and Hsp90. Once bound to chaperones, they may be refolded or sent for degradation. Binding of misfolded proteins to chaperones also prevents their aggregation. Due to competition from misfolded proteins, Hsf1 is released from Hsp90 and its concentration is now sufficient for it to form trimers. Hsf1 trimers are phosphorylated and become transcriptionally active so Hsps are upregulated. Hsp70 also binds to a phosphatase (currently unknown and so is referred to as PPX) which will dephosphorylate the trimers. This occurs when the concentration of free Hsp70 is sufficiently high and provides a negative feedback loop to prevent further upregulation of Hsps when they are present in sufficient number. This also provides a mechanism to switch off transcription of Hsps once the stress is over. There is sufficient Hsp70, despite some of the pool binding to misfolded protein, to continue to keep Mkp1 active and so prevent apoptosis. (3) High levels of stress {#s1c} ------------------------- If cells are subject to high levels of stress over long time periods ([Figure 3](#pone-0022038-g003){ref-type="fig"}), there will be a large increase in misfolded proteins which bind to the pool of free Hsp70 and Hsp90. Upregulation of Hsps will occur but eventually the chaperone system is overwhelmed and free pools are severely depleted. The idea of an overload of the chaperone system has been previously proposed and so we will use our model to test this hypothesis [@pone.0022038-Csermely1]. There is also experimental evidence that chaperone systems are overwhelmed during periods of high stress. For example, Sangster *et al*. have shown that elevated temperature can phenocopy the effects of Hsp90 inhibition on genetic trait selection in *Drosophila*. This is a consequence of heat-induced widespread misfolding of proteins overwhelming the Hsp90 chaperone system and thereby preventing it from assisting in a subset of proteins that require Hsp90 to fold correctly under non-stressed conditions [@pone.0022038-Sangster1]. Cowen and Lindquist have made similar observations in yeast cells [@pone.0022038-Cowen1]. An overload of the chaperone system leads to depletion of free pools of Hsp70 which allows phosphorylation of JNK and p38 and subsequently leads to apoptosis. There may also be an increase in aggregated protein if there is insufficient Hsp70 and Hsp90 to bind to the misfolded proteins. In addition, activation of p38 leads to increased ROS and so even more protein misfolding occurs. It has been shown that damaged protein accumulates with age in post-mitotic cells, and in addition, there is also an age-related increase in dysfunctional mitochondria leading to increased levels of ROS. All these factors may explain why the risk of neurodegenerative disorders increases with age. Results {#s2} ======= The purpose of building the model is to test the following hypotheses: (1) protein homeostasis is maintained during periods of low or moderate levels of stress; (2) protein homeostasis is disturbed after prolonged high levels of stress leading to cell death due to an overload of the chaperone system. Therefore we ran the model under three different scenarios: normal conditions (no stress); a transient increase in ROS (moderate stress); and with increasing stress with time (high stress). Normal conditions (no stress) {#s2a} ----------------------------- We set the model parameters so that under normal conditions levels of all proteins remain fairly constant, there is no upregulation of Hsps and no activation of JNK or p38. A typical simulation result is shown in [Figure 4](#pone-0022038-g004){ref-type="fig"}. As we have used stochastic simulation, we carried out multiple runs and plotted the mean values for 100 simulations ([Figure S2](#pone.0022038.s002){ref-type="supplementary-material"}). The model for unstressed conditions was also run in a deterministic simulator and produces very similar results to the stochastic model ([Figure S3](#pone.0022038.s003){ref-type="supplementary-material"}). {#pone-0022038-g004} Transient increase in ROS (moderate stress) {#s2b} ------------------------------------------- We ran stochastic simulations in which we increased ROS levels four-fold for a period of 1 hour ([Figure 5](#pone-0022038-g005){ref-type="fig"}). Deterministic simulation produced similar results to the stochastic model ([Figure S4](#pone.0022038.s004){ref-type="supplementary-material"}). After ROS levels increase (at time = 4 hours), there is an increase in misfolded protein which binds to Hsp70 and Hsp90. The model predicts that initially there are sufficient pools of unbound Hsps and so clients including Hsf1 remain bound to Hsp70 and Hsp90. As pools of unbound Hsps become depleted, misfolded protein competes with the client proteins but very little Hsf1 is released from its complex since the binding affinity of Hsf1 to Hsp90 is much greater than the binding affinity of other clients. So the model predicts only a very small increase in total Hsp levels. The misfolded protein is either refolded or degraded and replaced with newly synthesised protein so that by 2 hours after ROS levels have returned to normal, protein homeostasis is restored. We also looked at the effect in varying the amount and duration of stress by modifying the events for increasing and decreasing stress so that ROS was increased by a factor of 2, 4 or 8 for a period of 1, 2, 3 or 4 hours. The time to recovery under each condition was calculated and was assumed to be the time taken for the total level of misfolded protein after the stress event to return to within one standard deviation of the mean level of misfolded protein before the stress event. The model predicts that increasing the duration of stress leads to a shorter time of recovery for all stress levels ([Table 1](#pone-0022038-t001){ref-type="table"}). This was due to greater upregulation of Hsps which persisted even after the stress was over. Increasing the amount of stress from two-fold to four-fold increased the amount of misfolded protein but also increased the amount of heat shock proteins so that the time for recovery was also shortened. However, an 8-fold increase in the amount of ROS led to longer recovery times compared to a 4-fold increase when the duration of stress was 1, 2 or 3 hours. This was due to a much larger increase in misfolded proteins which outnumbered the pool of heat shock proteins. {#pone-0022038-g005} 10.1371/journal.pone.0022038.t001 ###### Simulation results of varying the amount and duration of a transient increase in ROS levels. {#pone-0022038-t001-1} Duration of stress (h) Fold increase Time for recovery (h) Maximum level of misfolded protein Maximum level of total Hsps ------------------------ --------------- ----------------------- ------------------------------------ ----------------------------- 1 2 3.24 1554.3 4377.2 2 2 2.76 1682.9 4383.6 3 2 2.46 1699.3 4394.4 4 2 2.27 1699.3 4404.5 1 4 3.06 2771.6 4398.9 2 4 2.69 3082.5 4435.8 3 4 2.43 3101.1 4470.8 4 4 2.24 3101.1 4503.6 1 8 3.12 5060.5 4467.1 2 8 3.02 6025.4 4713.0 3 8 2.74 6104.0 4939.1 4 8 2.06 6104.0 5030.8 Increasing ROS with time (high stress) {#s2c} -------------------------------------- We replaced a constant rate of ROS production with a rate that increases linearly with time ([Figure S5](#pone.0022038.s005){ref-type="supplementary-material"}). This was achieved by including the time variable in the rate law (see [Methods](#s4){ref-type="sec"}). As the results for this situation are very variable it was necessary to do 100 simulations. We ran the simulation for 48 hours (simulated time) and the model predicted that 74% cells died during this period. The cause of death was recorded and in this set of simulations, 32% of cell deaths are due to the p38 pathway and 42% due to the JNK pathway. Since we assume that the reaction rates for both pathways are equal, we would expect the number of deaths by each pathway to be equal to 37. If we assume that the number of deaths in 100 simulations follows a Poisson distribution with mean 37, then the standard deviation is 6.1. So obtaining 32 and 42 cell deaths via the p38 and JNK pathways respectively is due to stochastic effects. Cell death takes place due to insufficient free Hsp70 for the activation of Mkp1 which is required to prevent phosphorylation of both JNK and p38. There are no cell deaths due to proteasome inhibition, as even in cells which survive, 48 hours is insufficient for aggregates to start accumulating. Cell death occurs at variable times ranging from 0.4 to 47.4 hours with the median time to death equal to 34.4 hours which corresponds to the time at which ROS levels reached about 10 times the basal level ([Figure 6](#pone-0022038-g006){ref-type="fig"}). The model predicts that as ROS levels increase, there is also an increase in misfolded protein and levels of total protein decrease due to an increase in degradation of misfolded protein. The pools of free Hsp70 and Hsp90 decline to low levels but then start to increase again due to upregulation of Hsps. However, there are insufficient chaperones to deal with the overwhelming burden of misfolded protein or to inhibit cell death pathways. {#pone-0022038-g006} Deterministic model -- ROS increasing with time {#s2d} ----------------------------------------------- Deterministic simulation for the model with increasing ROS with time was carried out using CellDesigner and COPASI. For this set of conditions the deterministic and stochastic models give incomparable results. The deterministic model predicts that the time to cell death is 38.8 h ([Figure S6](#pone.0022038.s006){ref-type="supplementary-material"}) and gives no information on the variability in the time to death. It should be noted that the way the deterministic simulator models cell death is quite different to the stochastic method. In the deterministic simulation, the species variable CellDeath represents the sum of p38Death, JNKDeath and PIDeath and gradually increases with time until it reaches a value of 1.0 when the simulation stops. In the stochastic simulation, CellDeath remains at zero until a cell death pathway is activated by p38, JNK or proteasome inhibition and then changes to 1.0 and the simulation stops. Despite these differences, the deterministic version is useful for looking at the effect of each parameter on the timing of cell death (see below). Inhibition of JNK and p38 cell death pathways {#s2e} --------------------------------------------- Simulations in which ROS levels increase with time were carried out with inhibition of cell death pathways via JNK and p38 in order to examine if protein aggregation occurs over longer time periods. The model predicts that aggregates start to form and bind to the proteasome at about 24 hours but it takes four or five days before levels at the proteasome are high enough to cause cell death ([Figure 7](#pone-0022038-g007){ref-type="fig"}). Cell death by 8 days occurs in about 80% of simulations, with the earliest time of death occurring at 55.7 hours and median time to death equal to about 5.7 days. Interestingly, in some simulation runs, aggregates are sequestered into inclusion bodies and if this starts to happen before aggregates bind to the proteasome, then all further aggregates get sequestered and cell death via proteasome inhibition is prevented ([Figure 7F](#pone-0022038-g007){ref-type="fig"}). If the parameter for upregulation of Hsps is increased, then the model predicts that aggregates take much longer to form as more misfolded proteins are able to bind to Hsp70 and Hsp90 which prevents their aggregation. Alternatively, if the parameter for sequestering of aggregates is increased by a factor of two or more, then proteasome inhibition is prevented. Therefore the model predicts that inhibition of cell death pathways via p38 or JNK may not be beneficial unless levels of misfolded and aggregated protein can also be reduced. Since *k~upregHsp~* and *k~seqagg~* have an effect on protein aggregation, we also checked whether changing these parameters had an effect on the time to cell death. In the model which includes JNK and p38 death pathways, neither parameter has an effect on the time to cell death. However, when both JNK and p38 death pathways are inhibited, varying *k~seqagg~*, from half to double its initial value, did affect the predicted time to cell death in the deterministic model ([Figure S7](#pone.0022038.s007){ref-type="supplementary-material"}). Varying *k~upregHsp~* over the same range had little effect on time to cell death, and it was necessary to decrease (and increase) its value over two orders of magnitude to see an effect ([Figure S8](#pone.0022038.s008){ref-type="supplementary-material"}). {#pone-0022038-g007} Effect of relative change in the JNK and p38 rate constants {#s2f} ----------------------------------------------------------- We assumed that cell death pathways via p38 and JNK have equal probability of being activated and which pathway was activated was determined by stochastic effects. This may not be the case in reality and so we investigated the effect of relative changes in the JNK and p38 rate constants. We lowered *k~p38death~* by 50% and increased *k~Jnkdeath~* by 50% so that the overall death rate remained the same and carried out 100 simulations. The model predicts that 58 simulated cells die due to the JNK pathway, 22 simulated cells die due to the p38 pathway and 20 simulated cells remain alive by 48 hours. The median time to cell death is 27.3 hours (range 0.4--47.7 hours). The cells which are still alive are included in the calculation for the median since we assume that if we carried out simulations for longer time periods all cells would die and the time of later deaths does not affect the median value (provided more than 50% of cells have died) unlike the effect on the mean. The median time to cell death when the JNK and p38 constant rates are equal is 34.4 hours, so changing the relative rates results in earlier cell death. In addition, 100 simulations were carried out in which *k~p38death~* was increased by 50% and *k~Jnkdeat~* ~h~ decreased by 50%. In this scenario, the model predicts 19 deaths due to JNK pathway, 64 deaths due to p38 pathway and 17 cells still surviving by 48 h with median time to cell death equal to 31.9 hours (range 0.4--47.7 hours). So again the model predicts earlier cell death if one pathway has a greater probability of being activated than when both death pathways have equal probabilities. Hsp90 inhibition {#s2g} ---------------- The use of Hsp90 inhibitors have been proposed as a potential beneficial therapy in neurodegenerative diseases (recently reviewed in [@pone.0022038-Luo1]). Hsp90 inhibitors release Hsf1 from Hsp90 resulting in upregulation of heat shock proteins, and in addition they promote the degradation of Hsp90 clients. This works by blocking the ATPase activity of Hsp90 so that instead of clients being refolded they are targeted for degradation. As already mentioned, the half-life of Akt is reduced from about 36 hours to 12 hours after the addition of an Hsp90 inhibitor [@pone.0022038-Basso1]. To model the effects of Hsp90 inhibition we set the parameters for Hsp90/Hsf1 binding and for the release of Hsp90 from its clients (Hsp90client and Akt) to zero after one hour of simulation time using an event structure (see [Table 2](#pone-0022038-t002){ref-type="table"}). In order to check that Hsp90 affects the half-life of Akt we set the synthesis rate for Akt to zero and ran the model with and without Hsp90 inhibition. The model predicts that the half-life of Akt is reduced from 36.2 hours to 13.7 hours when Hsp90 is inhibited, in agreement with the experimental data. When we restore Akt synthesis, the model predicts a decline in Akt pools if Hsp90 is inhibited. There is also an increase in Hsp70 and Hsp90 pools due to the upregulation of heat shock proteins after Hsp90 inhibition. The effects of Hsp90 inhibition under conditions of high stress (ROS increases linearly with time) was also examined. In this case, the model predicts that there is no significant effect on the percentage of cell deaths or the timing of cell death. However, the current model does not specifically include Hsp90 clients such as GSK3β or mutant tau which like Akt are regulated by Hsp90 via CHIP. We would expect that inhibition of Hsp90 would lead to a decline in GSK3β and mutant tau levels and so we would predict a reduction in cell death. On the other hand Akt inhibits GSK3β activity and so GSK3β may become more active even if actual protein levels are reduced. Therefore it will be of interest to add further detail to the model to test these predictions. 10.1371/journal.pone.0022038.t002 ###### Model Events. {#pone-0022038-t002-2} Name Trigger Assignments ----------------- -------------------- ---------------------------------------------------------------------- DeathOfCell CellDeath \> = 1.0 *k~alive~* = 0 incROS t\>14400.0 ROS = 40; *k~genROS~*  = 0.04 decROS t\>18000.0 ROS = 10; *k~genROS~* = 0.01 Hsp90Inhibition t\>3600.0 *k~binHsf1Hsp90~* = 0; *k~relHsp90client~* = 0; *k~relAktHsp90~* = 0 Model analysis {#s2h} -------------- ### 1. Parameter scan for model under normal conditions {#s2h1} Many of the parameters in the model are estimated based on knowledge of protein half-lives or on knowledge of the rate of different types of reactions in cells (e.g. phosphorylation occurs over a time-scale of minutes). Since it is not possible to obtain exact measurements for the parameters in this model we carried out a parameter scan to see which parameters affected the steady state of the system under normal conditions using a deterministic version of the model in COPASI [@pone.0022038-Hoops1]. Each parameter was scanned over ten values ranging from half to double its initial value. The results of the parameter scans indicate that altering values of the majority of parameters over a two-fold range have little effect on the steady state of the system under normal conditions. The parameters which did affect the system are shown in [Table 3](#pone-0022038-t003){ref-type="table"} and are those involved in protein turnover, misfolding and folding, basal turnover of Hsp70 and Hsp90, and turnover of ROS. The parameters which have the greatest effect on the levels of misfolded protein are *k~misfold~*, *k~refold~*, *k~genROS~* and *k~remROS~*. The model suggests that the best way to reduce levels of misfolded protein is to lower levels of ROS. The kinetics of p38 phosphorylation/de-phosphorylation also have a small effect on pools of NatP, MisP and free pools of Hsps due to its effect on ROS levels ([Table 3](#pone-0022038-t003){ref-type="table"}). The parameters which had more than a 10% effect on protein pools were also examined using the stochastic model ([Table 4](#pone-0022038-t004){ref-type="table"}). As it is necessary to do repeat simulations, we just examined the effect of doubling the value of each parameter and calculated the mean values from 20 repeat simulations. The results are very similar to the deterministic model (compare [Tables 3](#pone-0022038-t003){ref-type="table"} and [4](#pone-0022038-t004){ref-type="table"}). 10.1371/journal.pone.0022038.t003 ###### Results of parameter scan for deterministic model under normal conditions. {#pone-0022038-t003-3} -------------------- ------------- ------------- -------------------- -------------------- Parameter NatP Total MisP Free pool of Hsp70 Free pool of Hsp90 *k~synNatP~* **↑** 0.079 **↑** 0.075 ↓ 0.018 ↓ 0.018 *k~misfold~* ↓ **0.110** ↑ **0.798** ↓ **0.172** ↓**0.173** *k~binMisPProt~* ↓ 0.058 ↓**0.146** ↑ 0.037 ↑ 0.036 *k~refold~* ↑ 0.060 ↓ **0.432** **↑** 0.096 **↑** 0.096 *k~basalsynHsp70~* ↔ 0.000 ↓ 0.002 **↑ 0.160** ↑ 0.014 *k~basalsynHsp90~* ↔ 0.000 ↓ 0.002 ↑ 0.014 **↑ 0.131** *k~binHsp70Prot~* ↔ 0.000 ↑ 0.002 ↓ **0.152** ↓ 0.015 *k~binHsp90Prot~* ↔ 0.000 ↑ 0.002 ↓ 0.016 ↓ **0.131** *k~genROS~* ↓ **0.108** **↑ 0.799** ↓ **0.176** **↓ 0.177** *k~remROS~* **↑** 0.064 ↓ **0.481** ↑ **0.108** **↑ 0.109** *k~phosp38~* ↓ 0.005 ↑ 0.033 ↓ 0.007 ↓ 0.007 *k~dephosp38Mkp1~* **↑** 0.002 ↓ 0.017 ↑ 0.004 ↑ 0.004 -------------------- ------------- ------------- -------------------- -------------------- Each parameter was scanned over 10 values ranging from half to double its initial value over a period of 10 hours. The results show the effect of increasing the parameter from its initial value on the pools of NatP, MisP (including MisP bound to Hsps and the proteasome), free pools of Hsp70 and free pools of Hsp90. The arrows show the direction of change, the numbers indicate the proportional change in the species value when the parameter value is doubled (), where the denominator is equal to one, if the parameter value *k* is double). Effects which are greater than 10% are indicated in boldface. 10.1371/journal.pone.0022038.t004 ###### Results of parameter scan for stochastic model under normal conditions. {#pone-0022038-t004-4} Parameter NatP Total MisP Free pool of Hsp70 Free pool of Hsp90 -------------------- ------------- ------------- -------------------- -------------------- *k~misfold~* ↓ **0.114** ↑ **0.864** ↓ **0.172** ↓**0.173** *k~binMisPProt~* ↓ 0.063 ↓**0.173** ↑ 0.037 ↑ 0.036 *k~refold~* ↑ 0.057 ↓ **0.444** **↑** 0.096 **↑** 0.096 *k~basalsynHsp70~* ↓ 0.006 ↑0.027 **↑ 0.157** ↔ 0.000 *k~basalsynHsp90~* ↓ 0.006 ↑0.027 ↑ 0.014 **↑ 0.131** *k~binHsp70Prot~* ↓ 0.006 ↑ 0.032 ↓ **0.164** ↓ 0.028 *k~binHsp90Prot~* ↓ 0.006 ↑ 0.008 ↓ 0.014 ↓ **0.135** *k~genROS~* ↓ **0.108** **↑ 0.752** ↓ **0.164** **↓ 0.169** *k~remROS~* **↑** 0.063 ↓ **0.492** ↑ **0.107** **↑ 0.107** Each parameter was increased to double its initial value and 20 repeat simulations were carried out over a 10 hour period. The results show the effect on the mean values of NatP, MisP (including MisP bound to Hsps and the proteasome), free pools of Hsp70 and free pools of Hsp90. The arrows show the direction of change, the numbers indicate the proportional change in the species value when the parameter value is doubled (), where the denominator is equal to one, if the parameter value *k* is double). Effects which are greater than 10% are indicated in boldface. ### 2. Parameter scan for model under conditions of high stress {#s2h2} We also looked at the effect of parameters under conditions of high stress by doing a parameter scan on the model with increasing ROS levels. As this model does not produce a steady state until death occurs, we looked at the effect of a two-fold increase in the parameter value on the time to cell death which the deterministic model predicts to take place at 38.8 hours. (Note that this does not equal the median time to cell death in the stochastic model as mentioned previously). The parameters which have the greatest effect on the time to cell death are *k~p38death~*, *k~Jnkdeath~*, *k~p38act~* and *k~remROS~* ([Table 5](#pone-0022038-t005){ref-type="table"}). The first three of these parameters decrease the time to cell death by more than 20% (i.e death occurs before 31.3 hours), whereas a two-fold increase in *k~remROS~* delays cell death by about 40% (death occurs at about 53.8 hours). [Table 5](#pone-0022038-t005){ref-type="table"} shows all the parameters which have more than a 1% effect on the time to cell death. These parameters are either involved in ROS turnover, cell death pathways, or protein misfolding and aggregation. 10.1371/journal.pone.0022038.t005 ###### Results of parameter scan for deterministic model with increasing ROS over time. {#pone-0022038-t005-5} Size of effect Parameters which delay cell death Predicted time of cell death (hours) Parameters which hasten cell death Predicted time of cell death (hours) ---------------- ----------------------------------- -------------------------------------- ------------------------------------ -------------------------------------- 30--40% *k~remROS~* 53.8 20--30% *k~p38death~* 30.4 20--30% *k~Jnkdeath~* 30.5 20--30% *k~p38act~* 30.1 10--20% *k~dephosJnkMkp1~* 44.9 *k~phosJnk~* 33.1 10--20% *k~dephosp38Mkp1~* 45.0 *k~phosp38~* 32.9 5--10% *k~misfold~* 36.7 5--10% *k~genROS~* 36.6 1--5% *k~binMisPProt~* 39.3 *k~agg~* 38.5 1--5% *k~genROSp38~* 38.4 1--5% *k~PIdeath~* 38.6 The results show the effect of a two-fold increase in the parameter value on time to cell death (predicted time to cell death for default parameters  = 38.8 hours). ### 3. Global parameter scan for model under normal conditions {#s2h3} In order to further examine the sensitivity of the model to changes in parameter values a global parameter scan in which each parameter was varied simultaneously was carried out using the deterministic model. As the model contains 60 parameters, we generated 50 sets of parameters in which each parameter was randomly assigned a value between half and double its initial value. Each model was run in COPASI and the results were analysed and plotted in R. The full results are given in the supplementary information ([Text S1](#pone.0022038.s011){ref-type="supplementary-material"} and [Table S1](#pone.0022038.s009){ref-type="supplementary-material"}) and summarised in [Table 6](#pone-0022038-t006){ref-type="table"}. This shows that there are two main types of effect, either increases in Hsps with an increase in NatP and decrease in total MisP or decreases in Hsps with a decrease in NatP and increase in total MisP. The first group is normally associated with a decrease in ROS and the second with an increase in ROS. Closer examination of the parameter sets that produced these results reveal that in the sets with large changes in the variables, there are large changes in some of the more sensitive parameters in the model which are given in [Table 4](#pone-0022038-t004){ref-type="table"}. For example, the sets in which levels of NatP and Hsps decrease and MisP increases have larger values for the rate of misfolding, and the amount of MisP increases further if the parameter for ROS generation is increased and/or the parameter for ROS removal is decreased (compare parameter Set 1 and Set 6 in the supplementary information [Text S1](#pone.0022038.s011){ref-type="supplementary-material"}). 10.1371/journal.pone.0022038.t006 ###### Summary of global parameter scan results. {#pone-0022038-t006-6} ROS ↓ ROS ↑ ROS ↔ (\<5% change) --------------------------------------------- ------- ------- --------------------- NatP ↑, Hsps ↑, MisP ↓ 19 2 2 NatP ↓, Hsps ↓, MisP ↑ 3 15 1 NatP ↔, Hsps ↔, MisP ↑ 1 1 0 NatP ↔, Hsps ↔, MisP ↓ 1 1 0 Less than 10% change in NatP, Hsps and MisP 4 0 0 Discussion {#s3} ========== Neurodegeneration is an age-related disorder which is characterised by the loss of a subset of neurons in specific regions of the brain and the accumulation of aggregated protein. The proteins involved and the regions of the brain affected are dependent on the particular disease which suggests that stochastic effects may play an important role. Despite individual differences, there seems to be common mechanisms involved which involve loss of protein homeostasis due to cellular systems for removing and repairing damage being overwhelmed. We built a stochastic model to investigate the processes involved, focussing on the role of Hsp70 and Hsp90 in preventing both protein aggregation and cell death. We extended the chaperone model of Proctor *et al.* [@pone.0022038-Proctor1] to include the molecular chaperone Hsp70, clients of Hsp70 and Hsp90 and apoptotic pathways. We used the model to examine the effects of transient moderate stress and the effects of a gradual increase in stress over time to reflect the ageing process. The model predicts that protein homeostasis can be maintained during short periods of stress due to the abundance of chaperones in cells and the fact that misfolded protein competes with clients for Hsps when free pools become diminished. So upregulation of Hsps is not necessary for short periods of stress. This was an unexpected result as we had predicted that chaperones would be upregulated after transient moderate stress ([Figure 2](#pone-0022038-g002){ref-type="fig"}). However, there are large pools of free Hsps under normal conditions making upregulation unnecessary unless the stress is severe or prolonged. Under very long periods of stress, the chaperone system eventually becomes overwhelmed as the upregulation of Hsps cannot keep pace with the increasing demand of protein misfolding. This is consistent with experimental data described in the [Introduction](#s1){ref-type="sec"}, which show that the capacity of chaperone systems can be exceeded under conditions of high external stress [@pone.0022038-Sangster1], [@pone.0022038-Cowen1]. As the chaperone system becomes overwhelmed there is a decline in free pools of the anti-apoptotic Hsp70 and so the probability of cell death pathways being activated increases leading to neuronal loss. Since neuronal loss has many adverse effects on brain function, a potential therapeutic target is to inhibit cell death pathways. This could be achieved by the use of JNK or p38 inhibitors. However, the model shows that inhibiting either or both of these pathways may delay cell death but that cell death will still ultimately occur via other apoptotic pathways. For example, p53 which is normally rapidly turned over by the proteasome may start to accumulate if aggregated protein inhibits proteasomal function. Indeed, the model predicts that preventing cell death does not stop the aggregation process and eventually cells will die due to proteasome inhibition. The model predicts that if aggregates are sequestered into inclusion bodies, then cell death is less likely to occur although it is necessary for inclusions to form before aggregates start binding to the proteasome. However, it is likely, that high levels of inclusions will also result in cell death [@pone.0022038-Tang1], and so even if it is possible to manipulate this pathway, it may not be beneficial. Therefore, we suggest that any intervention to inhibit cell death pathways also needs to be able to reduce protein misfolding and aggregation either by enhancing clearance systems or by reducing levels of stress within cells. Our models indicate that the latter strategy may be particularly beneficial. Although it is generally accepted that an increase in ROS-mediated macromolecular damage contributes to ageing, a causal relationship between ROS and proteotoxicity is still controversial. It has been suggested that an increase in oxidative damage with age could be due to an age-related increase in ROS production [@pone.0022038-Beckman1], although other possibilities include a decline in the efficiency of antioxidant systems or an increase in damaged mitochondria which produce more ROS. An increase in ROS with time can be modelled by either increasing the rate of ROS production or decreasing the rate of ROS removal over time. The outcome in the model would be the same and we chose to allow the level of ROS to slowly increase linearly with time. It might be preferable to include mitochondria in the model and then to use ROS generation from the mitochondria as an input instead of the reaction for ROS generation. This would add further complexity to the model but would provide a more complete picture of the processes involved. In addition a constant pool of ATP could be replaced by reactions of ATP production and consumption where the production rate depends on the state of the mitochondria, e.g. damaged mitochondria produce less ATP. This would be an interesting future development of the current model. Other perturbations which could be carried out in the current model include examining a lowered efficiency of Hsf1 transcription or a decline in proteasome efficiency with age. The model developed here makes several predictions that are potentially testable using cell culture and animal models. In general there is relatively little information in the chaperone literature on the extent to which various chaperone pools are engaged in the folding process [@pone.0022038-Morimoto1]. Kamal *et al*. have compared the extent to which Hsp90 is complexed with cochaperones in normal and tumor cells, using this as a marker of Hsp90 engagement with client proteins [@pone.0022038-Kamal1]. These experiments suggested that only a small fraction of Hsp90 in normal cells is actively engaged in protein folding. It would be interesting to perform similar analyses on cells in which ROS levels have been manipulated experimentally. This might give quantitative experimental data on levels of total and free Hsp90 which could then be compared with model predictions. Many potential therapeutic strategies for neurodegenerative disorders are targeting JNK and p38. However, our model predictions suggest that although these therapies may be useful in delaying the onset of disease pathology, they will not prevent the early stages of disease or its progression. For example the model predicts that inhibition of p38 and JNK will delay cell death but that eventually inhibition of proteasomes by aggregated protein will cause death in a significant proportion of cells. Previous studies on the role of p38 and JNK have only assessed apoptosis at a short time interval after initiation of protein misfolding in cells and thus are compatible with this model. Assessing apoptosis at longer intervals after initiating protein misfolding would provide a test of this model prediction. This experiment could be performed in the absence and presence of constitutively-active Hsf1 to determine whether a combination of stress-activated protein kinase inhibition and enhancement of chaperone activity is more effective at inhibiting apoptosis than either treatment alone. This would give insights into whether a combination of therapeutic strategies such as increasing chaperone activity to reduce protein aggregation and inhibiting p38 and JNK pathways would be beneficial. Other therapies that could be tested in combination with JNK or p38 inhibitors are the use of antioxidants to reduce protein damage and aggregation, or Hsp90 inhibitors which would reduce aggregation by increasing chaperone activity and reducing levels of mutant tau and may also have effects on apoptosis by reducing the level GSK3β. As recently suggested by Lindner & Demarez, merging predictive modelling and quantitative experimentation will bring new breakthroughs in our understanding of the ageing process [@pone.0022038-Lindner1]. Although the model is quite complex, it is still a simplification of the cellular system and it may be desirable to add in further detail in order to determine the main pathways involved in maintaining protein homeostasis. The level of detail chosen in the current model was based on the current knowledge of the system and the relevance to ageing especially in the context of neurodegeneration. The model was built in SBML, a computer readable format for representing computer models of biological processes. SBML models can be used by an extensive variety of software tools without the need for recoding the model which means that models can be easily shared and adapted by other users. Therefore, the model can easily be adapted by other users and it will be straight forward to add further detail as new information about the system emerges. It is also possible to remove some of the current detail. For example the reactions involving Akt could be removed and then we would assume that Akt is one of the generic Hsp90 clients. In fact removing these reactions did not affect the qualitative behaviour of the model but we chose to include them as it gave insights into the effect of Hsp90 inhibition on Hsp90 clients. Since Akt levels declined when Hsp90 is inhibited we would expect that other clients such as GSK3β and mutant tau would also decline. Therefore we predict that inhibition of Hsp90 might lead to less cell death even though the current model does not show this. On the other hand Akt inhibits GSK3β activity and so the effect of Hsp90 inhibition on GSK3β is not entirely clear and requires further investigation. An important future extension to the model is the addition of GSK3β and tau. Both proteins are linked to Akt since Akt phosphorylates GSK3β and it has been shown that Akt plays an important role in regulating tau degradation [@pone.0022038-Dickey2]. As well as the chaperone system, the ubiquitin-proteasome system plays an important role by removing damaged and misfolded proteins. Our model includes protein degradation but does not have detail of the ubiquitination steps. A model of the ubiquitin-proteasome has also been developed by Proctor *et al.* [@pone.0022038-Proctor2] and since this model was also coded in SBML, it will be possible to link the models together in order to explore the interactions between the chaperone and proteasome pathways in more detail. Hsp70 and Hsp90 are both degraded via CHIP and it may be important to add further detail of these degradation pathways. The molecular chaperone Hsp27 (P04792) is also important in preventing protein aggregation and is also anti-apoptotic. It predominantly exists as oligomers which have chaperone activity, but in stressed conditions oligomers are disrupted by phosphorylation of Hsp27 to form monomers and dimers. One of the kinases involved is p38 which as previously mentioned is itself activated by stress. Hsp27 oligomers also have a role in modulating ROS levels by induction of glutathione levels. A separate model of Hsp27 is currently being developed and this will be incorporated into the model of Hsp70 and Hsp90 to provide a more complete model of the chaperone system. Our current model includes generic pools of Hsp70 and Hsp90 clients and also some specific clients such as Akt. As already mentioned, another client of Hsp90 which plays an important role in neurodegeneration is (GSK3β) which is involved in phosphorylation of tau, production of amyloid beta (Aβ (P05067)), and modulates apoptotic pathways (reviewed in [@pone.0022038-Hernandez1], [@pone.0022038-Hooper1]). It also interacts with p53 and activity of both proteins is increased as a result of this interaction [@pone.0022038-Bijur1]. A stochastic model of the role of GSK3β and p53 has been developed [@pone.0022038-Proctor3] and so it will be of great interest to add GSK3β to the chaperone model and then link the models together to examine the effect of Hsp90 inhibition on the aggregation kinetics of tau and Aβ. We assumed that JNK and p38 were both directly activated by ROS. Recent data suggests that the catalytic Cys residue Mkp1 is oxidised and so deactivated by ROS [@pone.0022038-Liu1]. Therefore we also assumed that Mkp1 deactivation depends on ROS levels and so an increase in ROS may lead to reduced dephosphorylation of JNK and p38. However, ROS also induces upregulation of Hsp70 which results in more active Mkp1. Our model predicts that increasing ROS leads to inactivation of Mkp1 as the increased pools of Hsp70 are required by the increased levels of misfolded protein. Our model uncovers three main principles about the system. The first is that oxidative stress is the main trigger for the loss in protein homeostasis and so reducing stress is the key to preventing the initiation of the disease process. The second principle is that the ability of the cell to deal with stress via the chaperone and degradation pathways is important especially over long time scales as misfolded and aggregated protein will accumulate and then disease progression is difficult to halt. Finally, inhibiting cell death pathways allows longer cell survival but cannot prevent the aggregation process which will also ultimately lead to cell death. To summarise, we have developed a mechanistic model of the chaperone system to investigate how loss of protein homeostasis may lead to protein aggregation and cell death, both characteristics of neurodegeneration. We mainly used stochastic simulation so that cellular variability can be examined and the relationship between protein aggregation and cell death can be explored. The model made several predictions and the next stage will be to test these experimentally using cell culture or animal models. One important prediction is that inhibiting either the JNK or p38 cell death pathway may delay cell death but does not stop the aggregation process so that eventually cells die due to aggregated protein inhibiting proteasomal function. In addition to the chaperone system modelled here, the ubiquitin-proteasome system, lysosomal pathways, and mitochondria are involved in age-related neurodegeneration. Mathematical models of these other cellular mechanisms are currently being developed by the authors and will provide the building blocks of an integrative model. This will increase our understanding about how different processes interact to produce systemic outcomes. Methods {#s4} ======= Building the stochastic model {#s4a} ----------------------------- We modified our previous model of Hsp90 [@pone.0022038-Proctor1] which modelled the role of Hsp90 in maintaining protein homeostasis. The earlier model was encoded in the SBML [@pone.0022038-Hucka1] and so is easily modified. As in our previous model, we assume that there is a pool of proteins which are in their native conformation (NatP) but that these proteins are continuously subjected to damage which results in misfolding. There are three outcomes for misfolded proteins (MisP). Firstly, the molecular chaperones Hsp70 and Hsp90 bind to MisP to prevent their aggregation and an attempt is made to refold the proteins in an ATP-dependent reaction. Secondly, Hsp70 or Hsp90 will transport misfolded protein to the ubiquitin-proteasome system for degradation, although we do not include details of the ubiquitination steps as previously modelled [@pone.0022038-Proctor2]. Lastly the misfolded proteins may aggregate to form a small aggregate (AggP). Further misfolded protein may then bind to a small aggregate to form a larger aggregate (Seq AggP) which we will also refer to as an inclusion body. Small aggregates may bind to the proteasome and inhibit proteasomal function (AggP_Proteasome) and may also increase the generation of ROS. We assume that SeqAggP does not interfere with the cellular machinery and is a means to isolating the toxic aggregates. We also assume that Hsp70 and Hsp90 may be damaged by ROS and that damaged forms lose their chaperone activity and may form small aggregates or be sequestered into inclusion bodies in a similar way to misfolded protein. Hsp70 and Hsp90 both have many client proteins which bind with high affinity to form complexes, but are constantly undergoing cycles of assembly and disassembly. We represent these clients with generic pools named Hsp70Clients and Hsp90Clients. We also include the Hsp90 client Akt since this kinase is important in many pathways and so provides a link to other models. In addition, there is experimental data on the half-life of Akt and how this is affected by Hsp90 inhibition providing useful information for the parameters involved in Akt turnover. Hsp90 binds Heat Shock transcription Factor-1 (Hsf1) which keeps Hsf1 in its inactive monomeric state which we include as a separate species. Under conditions of stress, an increase in misfolded protein sequesters Hsf1 from Hsp90 and the increase in the pool of unbound Hsf1 enables dimerisation and trimerisation to take place. Hsf1 trimers are then phosphorylated by a protein kinase (PKC, (IPR015745)) which makes Hsf1 transcriptionally active. Hsf1 trimers can bind to the Heat Shock Element (HSE) in either their phosphorylated or un-phosphorylated state but only the former binding leads to transcription of heat shock proteins. We also include basal synthesis and degradation of Hsp70 and Hsp90. Among the clients for Hsp70 is the currently unknown phosphatase responsible for dephosphorylating Hsf1 trimers, (which we name PPX). We assume that under normal conditions PPX is in complex with Hsp70 and so Hsf1 trimers are in their unphosphorylated state. Hsp70 also activates the phosphatase Mkp1 which is responsible for dephosphorylating JNK and p38 [@pone.0022038-Lee1]. When free pools of Hsp70 are high, Mkp1 is in its active state, but if Hsp70 pools are depleted, Mkp1 becomes inactivated and so JNK and p38 are more likely to be in a phosphorylated state. Under conditions of stress misfolded protein binds preferentially to Hsp70 leading to an increase in phosphorylated JNK and Hsf1 trimers. This leads to an increase in the transcription of heat shock genes. If JNK remains in a phosphorylated state, it will contribute to apoptotic signalling. However, if Hsp70 levels increase due to the transcriptional activity of Hsf1, then JNK will be de-phosphorylated before apoptotic signalling takes place. We assume that JNK and p38 are both phosphorylated in response to signalling via ROS, so if ROS levels are high as a result of an increase in stress, the rate of phosphorylation increases. A list of species and reactions are shown in [Tables 7](#pone-0022038-t007){ref-type="table"} and [S2](#pone.0022038.s010){ref-type="supplementary-material"} respectively. The model was encoded in SBML shorthand and converted to full SBML [@pone.0022038-Wilkinson1]. It was then imported into the BASIS modelling system [@pone.0022038-Kirkwood1] and simulations were run using a stochastic simulator based on the Gillespie algorithm [@pone.0022038-Gillespie1]. To keep the model simple, we have not included detail of the apoptotic pathways. Models of apoptosis will be developed separately and then linked to this model if required. Instead, we assume that a cell will die if any of the following three conditions occur. The first two conditions involve activation of either JNK or p38, where we assume that the probability of cell death increases as pools of activated JNK or p38 increase. The third condition is inhibition of the proteasome where we assume that as the pool of AggP_Proteasome increases (and hence the available pool of proteasomes decreases) the probability of cell death increases. These conditions are modelled by reactions in SBML and the details are given in [Table S2](#pone.0022038.s010){ref-type="supplementary-material"}. Note that since the model is stochastic, cell death may occur even when pools of activated JNK or p38 are fairly low but with a very low probability. When a death reaction occurs, a dummy species is set to one, so that cell deaths can be counted according to type. A dummy parameter *kalive* is present in all the reactions and is initially set to one. When a cell death occurs, the parameter is set to zero so that no more reactions can take place in the simulation. This is achieved by using SBML event structures ([Table 2](#pone-0022038-t002){ref-type="table"}). The SBML code is available from the BASIS website [@pone.0022038-BASIS1], and a fully annotated version can be obtained from the Biomodels database (MODEL1005280000) [@pone.0022038-Biomodels1], [@pone.0022038-LeNovere1]. The code is also available in [Code S1](#pone.0022038.s012){ref-type="supplementary-material"} in the supplementary materials. Simulation results were analysed and plotted using R. 10.1371/journal.pone.0022038.t007 ###### List of Species. {#pone-0022038-t007-7} Species description Species Name Database term Initial Amount ----------------------------------------------- --------------------------- --------------------------- ---------------- Native protein NatP CHEBI:36080 17600 Misfolded protein MisP CHEBI:36080 0 Aggregated protein AggP CHEBI:36080 0 MisP bound to Hsp70 Hsp70_MisP IPR001023, CHEBI:36080 470 MisP bound to Hsp90 Hsp90_MisP IPR001404, CHEBI:36080 410 Hsp70 Hsp70 IPR001023 1400 Hsp90 Hsp90 IPR001404 1850 Damaged Hsp70 Hsp70_dam IPR001023 0 Damaged Hsp90 Hsp90_dam IPR001404 0 Hsp70 bound to proteasome Hsp70_Proteasome IPR001023, GO:0000502 0 Hsp90 bound to proteasome Hsp90_Proteasome IPR001404 GO:0000502 0 Hsf1 Hsf1 Q00613 5 Hsf1 bound to Hsp90 Hsf1_Hsp90 Q00613, IPR001404 95 Hsf1 dimers Hsf1_Hsf1 Q00613 0 Hsf1 trimers Hsf1_Hsf1_Hsf1 Q00613 0 Phosphorylated Hsf1 trimers Hsf1_Hsf1_Hsf1_P Q00613 0 Hsp70 heat shock element HSEHsp70 SBO:0000369 2 Hsp90 heat shock element HSEHsp90 SBO:0000369 2 Phosphorylated Hsf1 bound to HSEHsp70 HSEHsp70_Hsf1_Hsf1_Hsf1_P SBO:0000369, Q00613 0 Unphosphorylated Hsf1 bound to HSEHsp70 HSEHsp70_Hsf1_Hsf1_Hsf1 SBO:0000369, Q00613 0 Phosphorylated Hsf1 bound to HSEHsp90 HSEHsp90_Hsf1_Hsf1_Hsf1_P SBO:0000369, Q00613 0 Unphosphorylated Hsf1 bound to HSEHsp90 HSEHsp90_Hsf1_Hsf1_Hsf1 SBO:0000369, Q00613 0 Hsp70 client proteins Hsp70Client CHEBI:36080 490 Hsp90 client proteins Hsp90Client CHEBI:36080 590 Hsp70 bound to clients Hsp70_Hsp70Client IPR001023, CHEBI:36080 10 Hsp90 bound to clients Hsp90_Hsp90Client IPR001404, CHEBI:36080 10 Akt (protein kinase B) Akt P31749 340 Akt bound to Hsp90 Akt_Hsp90 P31749, IPR001404 30 Carboxy terminus of Hsp70-interacting protein CHIP Q9UNE7 255 Akt bound to CHIP/Hsp90 complex Akt_CHIP_Hsp90 P31749, Q9UNE7, IPR001404 80 Akt bound to proteasome Akt_Proteasome P31749, GO:0000502 0 Proteasome Proteasome GO:0000502 500 MisP bound to proteasome MisP_Proteasome CHEBI:36080,GO:0000502 0 Aggregated protein bound to proteasome AggP_Proteasome CHEBI:36080, GO:0000502 0 Sequestered aggregated protein SeqAggP CHEBI:36080 0 Phosphatase Mkp1 (DUSP1) Mkp1 P28562 0 Phosphorylated Mkp1 Mkp1_P P28562 100 Mkp1 bound to proteasome Mkp1_Proteasome P28562, GO:0000502 0 Phosphatase for Hsf1 Ppx GO:0008287 0 PPX bound to Hsp70 Hsp70_Ppx IPR001023, GO:0008287 100 JNK (MAPK8) Jnk P45983 100 Phosphorylated JNK Jnk_P P45983 0 p38MAPK (MAPK14) p38 Q16539 100 Phosphorylated p38MAPK p38_P Q16539 0 Protein kinase C Pkc IPR015745 100 Reactive oxygen species ROS CHEBI:26523 10 Adenosine triphosphate ATP CHEBI:15422 10000 Adenosine diphosphate ADP CHEBI:16761 1000 IPR: InterPro [@pone.0022038-Interpro1]. GO: Gene ontology [@pone.0022038-Gene1]. CHEBI: Chemical Entities of Biological Interest database [@pone.0022038-Chemical1]. P and Q: Uniprot [@pone.0022038-UniProtKBSwissProt1]. SBO: Systems Biology Ontology [@pone.0022038-Systems1]. Use of stochastic and deterministic models {#s4b} ------------------------------------------ Stochastic effects are very important in this model under conditions of moderate or high stress. In particular the timing of when the aggregation process starts, the destination of aggregates (inhibition of the proteasome or sequestered into inclusion bodies), the levels of ROS and the timing of cell death show large variations between simulation runs. However, stochastic simulations are very computer intensive and for conditions of high stress a set of 100 repeat runs takes about one week if run on a PC or about 24 hours if run on the BASIS cluster. Therefore it is not feasible to use a stochastic model to carry out a full parameter scan. To overcome this problem we also developed a deterministic model and ran simulations in CellDesigner and COPASI. We used the same initial conditions and parameter values for the deterministic model so that direct comparisons could be made. The rate laws for dimerisation reactions were adjusted, for example *k~dimerHsf1~*\<\#Hsf1\>\<\#Hsf1-1\>/2.0 was changed to *k~dimerHsf1~*Hsf1^2^. Initial amounts of species {#s4c} -------------------------- Proteins are present in very large numbers in cells. For example, experimental values of the number of p38 molecules per cell has been estimated as 10^6^ molecules per cell [@pone.0022038-Hendriks1]. As we are using stochastic simulation, it is not practical to have starting values in this order of magnitude due to the amount of time that would be required to carry out even a single simulation. Therefore it is necessary to scale down the values of model species. The motivation for the model was to examine the qualitative behaviour of the system and as information on initial amounts for the majority of species is not available we have assumed that most species are present in relatively similar abundances but that Hsps are an order of magnitude more abundant than kinases, and the generic pool of NatP was set to be two orders of magniture more abundant as this represents many different proteins. We used a constant value of 10^4^ molecules/cell for ATP since our model includes only a small fraction of the reactions requiring ATP and it does not seem appropriate to assume that ATP levels would be rate limiting in this model. We assumed that the cell volume is equal to one so that all initial amounts represent the number of molecules per cell and so we used the same initial values in the deterministic simulations. Parameter values {#s4d} ---------------- One of the most difficult parts in the model building process is finding values for all the parameters as kinetic data is often not available. Where kinetic data is available, it is often from *in vitro* systems which may not accurately reflect the *in vivo* kinetics. Also kinetic parameters are often dependent on cell type. However, even if exact values are not easy to obtain, it is often satisfactory to know the relative time scales of the reactions. For example, kinase/phosphatase reactions are very fast reactions occurring over timescales that range from fraction of a second to seconds, whereas protein synthesis takes several minutes. Degradation rates are dependent on protein type but fortunately, there is usually information available on protein half-lives which can be used to calculate the degradation rate. Note that the half-life of a protein is dependent on the rate at which the protein is targeted for degradation rather than on the rate of proteasome activity which we assume to be independent of protein type (apart from aggregated protein which we assume is totally resistant to degradation by proteasomes). So we set the rate at which a particular protein binds to the proteasome to be equal to --ln(0.5)/t~0.5~, where t~0.5~ is the protein half-life. As an example we assume that the half-life of Hsp70 is about 30 hours, so that *k~binHsp70Prot~* \* Proteasome  = −ln(0.5)/108000s. Since the pool of proteasomes is about 500, then *k~binHsp70Prot~*  = 1.2×10^−8^. We then set the basal synthesis rate, so that total Hsp70 pools remain constant under normal conditions. We set the parameters for aggregation to be very low, since it has been shown that there is a very long lag phase before aggregates start to form and also we would not expect aggregation to take place under normal conditions. Once we had chosen a set of parameters, we then simulated the model under normal conditions and checked that all species remained fairly constant over time. As the model is stochastic we would expect variations in protein levels over time but there should be no obvious trend of an increase or decrease in these levels. If such a trend was observed, we made the necessary adjustments to the parameter values. We also checked the model by doing 100 repeat simulations and plotting the mean values, and in addition ran the model in a deterministic simulator using CellDesigner [@pone.0022038-Funahashi1]. We carried out parameter scans for all the parameter values using COPASI [@pone.0022038-Hoops1] (see [Results](#s2){ref-type="sec"} section). Model validation {#s4e} ---------------- After choosing the parameters for normal conditions, we then validated the model against experimental data for the heat shock response. We do not have temperature as a variable in our model, however the transient increase in ROS levels has a similar effect to increasing the temperature for a specified time period. Kline & Morimoto et al measured the dynamics of phosphorylated Hsf1, binding of Hsf1 to HSE and the transcription rate of Hsp70 after heat-shock of HeLa cells at 42°C for 250 minutes [@pone.0022038-Kline1]. They observed a rapid activation of all three variables between 0 and 35 minutes after the stress with a attenuation back to basal levels over the next ∼200 minutes. Note that the attenuation phase begins during the continued heat-shock exposure. We examined the dynamics of the same variables from our simulation output under conditions of ROS increase for 4 hours. The model predicts that there is a rapid activation of all three variables which begins about 1 minute after the start of the stress response, reaches a maximum after about 1--2 hours and then attenuates back to basal levels over the next 2 hours with the attenuation phase beginning during the period of elevated ROS (data not shown). The attenuation phase starts even though ROS is still elevated due to the increase in levels of Hsp70 which will bind to PPX resulting in dephosphorylation of Hsf1 trimers. The qualitative behaviour of our model is in good agreement with the experimental data showing that the model captures the important features of the stress response. Supporting Information {#s5} ====================== ###### Diagram of the model. Numbers on the arrows refer to the reaction numbers in [Table S2](#pone.0022038.s010){ref-type="supplementary-material"}. Note that some reactions are omitted for clarity where similar reactions occur as noted in the footnotes to [Table S2](#pone.0022038.s010){ref-type="supplementary-material"}. (TIF) ###### Click here for additional data file. ###### Mean of 100 runs for normal conditions. A Native protein, total misfolded protein (includes misfolded bound by Hsps), and reactive oxygen species (ROS). ROS are scaled x100 to allow easier visualisation. B Total Hsp90 (free pools plus all complexes), Free Hsp90 (unbound Hsp90) and Hsp90_MisP (Hsp90 bound to misfolded protein). C Total Hsp70 (free pools plus all complexes), Free Hsp70 (unbound Hsp70) and Hsp70_MisP (Hsp70 bound to misfolded protein). (TIF) ###### Click here for additional data file. ###### Deterministic simulation for normal model. A Native protein, total misfolded protein (includes misfolded bound by Hsps), and reactive oxygen species (ROS). ROS are scaled x100 to allow easier visualisation. B Total Hsp90 (free pools plus all complexes), Free Hsp90 (unbound Hsp90) and Hsp90_MisP (Hsp90 bound to misfolded protein). C Total Hsp70 (free pools plus all complexes), Free Hsp70 (unbound Hsp70) and Hsp70_MisP (Hsp70 bound to misfolded protein). (TIF) ###### Click here for additional data file. ###### Deterministic simulation for model with transient stress. A Native protein, total misfolded protein (includes misfolded bound by Hsps), and reactive oxygen species (ROS). ROS are scaled x100 to allow easier visualisation. B Total Hsp90 (free pools plus all complexes), Free Hsp90 (unbound Hsp90) and Hsp90_MisP (Hsp90 bound to misfolded protein). C Total Hsp70 (free pools plus all complexes), Free Hsp70 (unbound Hsp70) and Hsp70_MisP (Hsp70 bound to misfolded protein). (TIF) ###### Click here for additional data file. ###### Graph to show how ROS levels increase with time. Black line shows rate of ROS production versus time, red and green curves show ROS levels for two stochastic simulations. Horizontal part of curve corresponds to cell death. (TIF) ###### Click here for additional data file. ###### Deterministic model for ROS increasing with time. A Native protein, total misfolded protein (includes misfolded bound by Hsps), and reactive oxygen species (ROS). ROS are scaled x100 to allow easier visualisation. B Total Hsp90 (free pools plus all complexes), Free Hsp90 (unbound Hsp90) and Hsp90_MisP (Hsp90 bound to misfolded protein). C Total Hsp70 (free pools plus all complexes), Free Hsp70 (unbound Hsp70) and Hsp70_MisP (Hsp70 bound to misfolded protein). (TIF) ###### Click here for additional data file. ###### Effect of varying *k~seqagg~*. The parameter *k~seqagg~* was varied from half to double of its initial value in the deterministic model with ROS increasing with time and inhibition of JNK and p38 death pathways. The parameter scan was carried out in COPASI and the results plotted in R. (TIF) ###### Click here for additional data file. ###### Effect of varying *k~upregHsp~*. The parameter *k~upregHsp~* was varied over two orders of magnitude in the deterministic model with ROS increasing with time and inhibition of JNK and p38 death pathways. The scan was carried out in COPASI and the results plotted in R. (TIF) ###### Click here for additional data file. ###### Results of the global parameter scan. Excel spreadsheet containing full results of 50 randomly chosen parameter sets and the effects on the model predictions. Sheet 1 contains all the parameter values for each set. Sheet 2 shows the percentage difference between each parameter and the default value. Sheet 3 shows the predicted percentage change in mean value for native protein, total misfolded protein, free pools of Hsp70 and Hsp90, and ROS for each parameter set. (XLS) ###### Click here for additional data file. ###### List of reactions. List of all the reactions in the model including kinetic rate laws and parameter values. (DOC) ###### Click here for additional data file. ###### Figures for the global parameter scan. Plots showing levels of native protein, misfolded protein and free pools of Hsp70 and Hsp90 for each of the 50 randomly chosen parameter sets. The number below each graph corresponds to the parameter set shown in [Table S1](#pone.0022038.s009){ref-type="supplementary-material"}. (PDF) ###### Click here for additional data file. ###### (XML) ###### Click here for additional data file. We thank Dr Graham Smith for reading the manuscript and for helpful suggestions. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**CJP is funded by Alzheimer Scotland and Alzheimer\'s Research UK (ART/RF2008/3). IAJL is supported by a grant from the Canadian Cancer Society (grant \#020121). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: CJP IAJL. Performed the experiments: CJP. Analyzed the data: CJP. Wrote the paper: CJP IAJL.

Tomczyńska M, Salata I, Bijak M, Saluk‐Bijak J. The potential contribution and role of a blood platelets in autoimmune thyroid diseases. J Cell Mol Med. 2018;22:6386--6390. 10.1111/jcmm.13862 1. INTRODUCTION {#jcmm13862-sec-0001} =============== The involvement of blood platelets in autoimmune diseases is well documented, and disturbance in their functioning is a hallmark of idiopathic thrombocytopenic purpura (ITP),[1](#jcmm13862-bib-0001){ref-type="ref"} rheumatoid arthritis (RA),[2](#jcmm13862-bib-0002){ref-type="ref"} systemic lupus erythematosus (SLE),[3](#jcmm13862-bib-0003){ref-type="ref"} antiphospholipid syndrome [4](#jcmm13862-bib-0004){ref-type="ref"} and multiple sclerosis.[5](#jcmm13862-bib-0005){ref-type="ref"} However, there is no data about platelet activity in autoimmune thyroid diseases (AITDs): Hashimoto\'s thyroiditis (HT) and Graves\' disease (GD). The pro‐inflammatory activity of platelets leads to disturbance of the haemostatic balance and can increase the risk of cardiovascular disease.[6](#jcmm13862-bib-0006){ref-type="ref"}, [7](#jcmm13862-bib-0007){ref-type="ref"}, [8](#jcmm13862-bib-0008){ref-type="ref"}, [9](#jcmm13862-bib-0009){ref-type="ref"} Studies indicate that people with AITDs are prone to the development of other autoimmune diseases and cardiovascular diseases. These include ITP and thrombocytopenia associated with disorders in the structure and function of blood platelets.[10](#jcmm13862-bib-0010){ref-type="ref"} The potential mechanism that may be associated with AITDs is the immunogenic destruction of platelets by circulating autoantibodies, which react with both target thyroid antigens and epitopes located on the platelets\' surface, mainly the typical glycoprotein receptors.[10](#jcmm13862-bib-0010){ref-type="ref"} In view of the proven significant contribution of platelets to the mechanisms of inflammation and to autoimmune processes, the aim of this current study was to answer the simple question of whether platelets present in the circulation of people with autoimmune hyperthyroidism (GD) or hypothyroidism (HT) exhibit the characteristics of increased activity. 2. MATERIALS AND METHODS {#jcmm13862-sec-0002} ======================== 2.1. Demographic and clinical characteristics {#jcmm13862-sec-0003} --------------------------------------------- The study population included 25 HT patients positive for both TPO‐Ab, Tg‐Ab and with elevated level of TSH and 50 GD patients with elevated concentration of T4 and/or T3, and suppressed TSH level, diffusely increased thyroidal uptake of iodine‐131, presence of TSHR antibodies and/or antimicrosomal antibodies. All subjects were without other autoimmune or acute and chronic inflammatory diseases. The healthy control (HC) consisted of 40 donors without any autoimmune or chronic inflammatory disease. All subjects were characterized by the correct number of platelets and did not use antiplatelet or immunomodulate drugs for at least 14 days prior to blood collection. All procedures were carried out according to the Helsinki Declaration and were approved by the Bioethics Committee of the Faculty of Biology and Environmental Protection of the University of Lodz, Poland, with Resolution No. 12/KBBN‐UŁ/II/2014. 2.2. Platelet aggregation {#jcmm13862-sec-0004} ------------------------- Platelet aggregation determined in response to physiological agonists: ADP (10 μmol/L), collagen (2 μg/mL) or arachidonic acid (0.5 mmol/L) was measured in platelet‐rich plasma using the turbidimetric method on the optical Chrono‐Log Aggregometer. 2.3. Flow cytometry analysis {#jcmm13862-sec-0005} ---------------------------- The resting or agonist‐stimulated (ADP 20 μmol/L, collagen 20 μg/mL) platelets were analysed using a flow cytometer ‐ LSR II Flow Cytometer (Becton Dickinson, San Diego, CA, USA). After fixation (1% Cellfix solution) blood samples were stained with saturating concentrations of murine monoclonal IgG1 antibodies: peridinin--chlorophyllprotein complex (PerCp)---a conjugated antibody against CD61 (constitutive platelets\' surface receptor, that distinguishes platelets from other cells), and fluorescein isothiocyanate (FITC)---a conjugated PAC‐1 antibody binds to the activated conformation of GPIIb/IIIa receptor. The fluorescence of 10 000 platelets (CD61/PerCP‐positive objects) was measured each time. In each sample, FITC fluorescence was detected and the percentage of PAC‐1‐positive platelets was determined relative to the total number of platelets (10,000 CD61/PerCP‐positive cells). Based on size and granularity, we determined forward light scatter (FSC) *vs*. side light scatter plots (SSC) in CD61/PerCP‐positive objects, formation of platelet subpopulations dependent on the degree of cell activation: aggregates (PAs) and platelet‐derived microparticles (PMPs). Using reference beads, we estimated FSC gates. CD61PerCP‐positive objects with an FSC lower than 10^2.3^ were characterized as PMPs, while objects with FSC higher than 10^4^ were considered PAs. All data analysis was performed in FACSDiva version 6.1.2. 2.4. Statistical analysis {#jcmm13862-sec-0006} ------------------------- The results were analysed for normality with a Shapiro‐Wilk test. The significance of the differences between the values was determined by normality using a *U*‐Mann‐Whitney test (for data deviating from normal distribution). 3. RESULTS {#jcmm13862-sec-0007} ========== In blood with non‐stimulated platelets, we observed an augmented basal level of PAs in the both GD and HT groups (about 2‐fold vs HC; *P *\< 0.001) as well as an increased level of PMPs in GD (2‐fold vs HC; *P *\< 0.001) and HT (2.5‐fold vs HC; *P *\< 0.001). Platelet activation was also measured through surface binding of PAC‐1 antibodies complementary only to the active form of GPIIb/IIIa responsible for platelet aggregation. Objects with a level of FITC fluorescence greater than 10^3.05^ were characterized as platelets with PAC‐1 antibody binding. Their number was 2.5‐fold higher in GD patients (*P *\< 0.001), and 2‐fold higher in HT patients (*P *\< 0.001), than in HC. The analysis of blood platelet responsiveness to the action of physiological agonists: ADP (20 μmol/L) or 20 μg/mL of collagen, showed the elevated PAC‐1 binding (almost 1.5‐fold increase for GD, *P *\< 0.001; and 2‐fold for HT, *P *\< 0.001), relative to HC. The pool of PAs in GD patients was 1.5‐fold greater vs HC; *P *\< 0.001, and in the HT patients 1.2‐fold greater vs HC; *P *\< 0.001. Similarly, the proportion of PMPs in GD patients was about 2.5‐fold larger vs HC; *P *\< 0.001 and in HT patients was almost 2‐fold larger vs HC; *P *\< 0.01 (Figure [1](#jcmm13862-fig-0001){ref-type="fig"}). {#jcmm13862-fig-0001} We also monitored the kinetic course of the aggregation process. Examples of aggregation curves recorded in an optical aggregate are shown in Figure [2](#jcmm13862-fig-0002){ref-type="fig"}. The platelet aggregation upon ADP stimulation was 10% higher for GD (*P* \< 0.005), and 8% for HT (*P* \< 0.05), compared to HC. Collagen caused 10% growth of control for GD and 11% for HT (*P* \< 0.005), while aggregation induced by arachidonic acid was 11% greater for GD (*P* \< 0.05) and 20% for HT (*P* \< 0.005), than in HC (Figure [2](#jcmm13862-fig-0002){ref-type="fig"}). {#jcmm13862-fig-0002} 4. DISCUSSION {#jcmm13862-sec-0008} ============= Our studies, for the first time demonstrated changes in the haemostatic function of platelets in HT and GD. We proved the elevated levels of platelet aggregation and generation of PMPs (vesicular structures mainly produced during activation and cell death) as well a greater sensitivity to agonists, which is crucial for platelet haemostatic function. The excessive production of microparticles induced by permanent cell activation may contribute to chronic inflammatory processes[11](#jcmm13862-bib-0011){ref-type="ref"} and predispose to autoimmune diseases.[12](#jcmm13862-bib-0012){ref-type="ref"} Our findings are in line with results indicating an increased level of PMPs in autoimmune diseases, such as RA and SLE, which have been associated with disease activity.[12](#jcmm13862-bib-0012){ref-type="ref"}, [13](#jcmm13862-bib-0013){ref-type="ref"} It has been proposed that PMPs can interact with circulating autoantibodies and C1q, participating in the formation of immune complexes, which could trigger immune responses in autoimmune diseases.[14](#jcmm13862-bib-0014){ref-type="ref"} GPIIb/IIIa receptors function as constituent antigens, but after platelet activation the number of GPIIb/IIIa copies grows and receptors change their conformation. Therefore, as surface antigens, they are a good marker for platelet activation. The conformational changes in the GPIIb/IIIa complex upon the platelets\' activation allow binding of fibrinogen, and in consequence platelet aggregation.[15](#jcmm13862-bib-0015){ref-type="ref"} We showed the increased surface expression of the active form of GPIIb/IIIa on platelets in AITDs. Aggregation is the final stage of platelet activation applicable to the cellular processes of haemostasis. We demonstrated significantly higher platelet aggregation in AITDs. 5. CONCLUSIONS {#jcmm13862-sec-0009} ============== Our analysis in whole blood samples without isolation of platelets significantly reduces the risk of creating artefacts and may illustrate the activation state of platelets in circulation. Therefore, we can conclude that the platelet hyperactivity is a phenomenon occurring in the vascular system of patients with AITDs. Because of the lack of differences in the studied groups (HT vs GD), we postulate that in AITDs, platelet abnormalities result from inflammation and autoimmune processes, more than hormone disorders. AUTHOR CONTRIBUTIONS {#jcmm13862-sec-0011} ==================== Małgorzata Tomczyńska and Michał Bijak conceived and designed the experiments; Małgorzata Tomczyńska performed the experiments; Małgorzata Tomczyńska and Joanna Saluk‐Bijak analysed the data; Ireneusz Salata contributed reagents/materials/analysis tools; Małgorzata Tomczyńska, Joanna Saluk‐Bijak and Michał Bijak wrote the manuscript. All authors approved the final version of the manuscript. CONFLICTS OF INTEREST {#jcmm13862-sec-0012} ===================== The authors confirm that there are no conflicts of interest. This work was supported by a grant from the Polish National Science Centre (no. 2014/13/N/NZ5/01389).

Introduction {#sec1-1} ============ According to recent global cancer statistics, gallbladder cancer ranks 20th among all cancers, thus it is a relatively uncommon cancer (International Agency for Research on Cancer, 2017). However, previous studies have reported a high incidence of gallbladder cancer in specific countries (Chile, Bolivia, and India) or confined areas (southern Chile and northern India) (Wistuba and Gazdar, 2004; Andia et al., 2008). Thus, the development of gallbladder cancer is suggested to be associated with environmental and genetic factors. India has a high incidence of gallbladder cancer, especially in the northern states of Punjab, Uttar Pradesh, Bihar, West Bengal, and Assam (Kapoor and McMichael, 2003; Nandakumar et al., 2005). Some risk factors for the development of gallbladder cancer in Indian people have been identified (Khan et al., 2013; Mhatre et al., 2016). According to a guideline provided by the Indian Council of Medical Research (2014), ethnicity, sex, age, gallstones, chronic inflammation, genetic factors, gallbladder polyps, and lifestyle factors are risk factors for gallbladder cancer in Indian people. However, the onset mechanism for gallbladder cancer in Indian people has not been sufficiently explained by these disclosed factors alone. Although cholelithiasis or presence of gallstones is the strongest risk factor for gallbladder cancer (Randi et al., 2006; Ishiguro et al., 2008), less than 3% of patients with cholelithiasis develop gallbladder cancer (Hundal and Shaffer, 2014) and 69--100% of Indian patients with gallbladder cancer have gallstones (Khan et al., 2013; Indian Council of Medical Research, 2014). These data suggest not only gallstones but also other factors are associated with developing gallbladder cancer. Some studies have reported an association between infections and gallbladder cancer risk (Kumar et al., 2006), but findings have been inconsistent because of differences in methodology and samples used. Some researchers reported that *Helicobacter pylori* (*H. pylori*) infection is associated with increased risk of gallbladder cancer (de Martel et al., 2009; Mishra et al., 2010), however, another study did not observe such association (Bohr et al., 2007). At least three studies have reported an association between *H. pylori* infection and gallbladder cancer in Indian people (Sharma et al., 2007; Mishra et al., 2011; Mishra et al., 2013). *H. pylori* DNA was detected in gallbladder tissue of gallbladder cancer patients, but *H. pylori* DNA detection rates did not differ between gallbladder cancer patients and cholelithiasis patients (Mishra et al., 2011). A serological test to measure serum or plasma *H. pylori* antibody titer has been developed. This test is useful for population-based *H. pylori* screening and treatment program because it is non-invasiveness, convenient, and inexpensive (Inui et al., 2017). However, the test cannot differentiate between *H. pylori* infection in the gallbladder and infection in other organs, such as the stomach, liver, and biliary epithelium. Although the antibody titer is not specific for the gallbladder, Deeba et al., (2010) reported that the titer in cholelithiasis and cholecystitis patients is significantly higher than that in healthy subjects. According to this evidence, we hypothesized that serum or plasma *H. pylori* antibody titer in gallbladder cancer patients with gallstones would be higher than that in cholelithiasis patients because the presence of gallstones is a major risk factor for gallbladder cancer. To our knowledge, no study has examined the association between *H. pylori* infection and gallbladder cancer risk using serological tests. Therefore, we conducted a hospital-based case-control study to clarify the role of *H. pylori* infection in the development of gallbladder cancer in Indian people. Materials and Methods {#sec1-2} ===================== Subjects and plasma collection {#sec2-1} ------------------------------ We conducted a hospital-based case-control study to evaluate the association between *H. pylori* infection and gallbladder cancer risk in Indian patients from May 2014 through July 2017. A total of 100 gallbladder cancer patients with gallstones (cases) and 100 cholelithiasis patients (controls) participated in this study. All patients had a diagnosis of gallbladder cancer or cholelithiasis at Sanjay Gandhi Post Graduate Institute of Medical Sciences in Lucknow in northern India, a high incidence area, from May 2014 through July 2017. Informed consent was obtained from all participants for the use of plasma samples. This study was approved by the Ethical Committees of Sanjay Gandhi Post Graduate Institute of Medical Sciences and Niigata University of Health and Welfare (No. 17809-170517). Plasma samples were collected from all participants before surgical treatment. Measurement of plasma H. pylori antibody titer {#sec2-2} ---------------------------------------------- Plasma *H. pylori* antibody titer was measured using a commercial kit (LZ Eiken *H. pylori* antibody; Eiken Chemical Co. Ltd., Tochigi, Japan) and an autoanalyzer (BM 9130, JEOL Ltd., Tokyo, Japan). *H. pylori* infection was defined as plasma antibody titers ≥10 U/mL according to the kit's manual (Eiken Chemical Co., Ltd., 2015). Statistical analysis {#sec2-3} -------------------- All statistical analyses were performed using Stata 14 software (StataCorp LLC, College Station, TX, USA). Differences in mean ages and antibody titers between cases and controls were analyzed by the chi-square test or Fisher's exact test. To compare differences in female proportions and *H. pylori* infection positivity rates of cases and controls, odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using two-way contingency table analysis. P values \<0.05 (two-tailed) were considered statistically significant. Results {#sec1-3} ======= [Table 1](#T1){ref-type="table"} shows participant characteristics. The proportion of female patients was 72% in cases and 65% in controls, which was not significantly different (P = 0.29). No significant difference was found in mean age between male cases (52.6 years; standard deviation \[SD\], 11.2; range, 33--73) and controls (47.1 years; SD, 12.5; range, 23--73), P = 0.07. However, significant differences in mean age were observed between female cases (53.2 years; SD, 11.3; range, 32--79) and controls (43.7 years; SD, 14.0; range, 14--75), P\<0.001, or between both sexes (cases: 53.0 years; SD, 11.2; range, 32--79 vs. controls: 44.9 years; SD, 13.5; range, 14--75), P\<0.001. ###### Participant Characteristics Characteristic Cases Controls  P value ------------------------- ------------- ------------- ---------- No. of participants 100 100 No. of female patients 72 65 0.29 Mean age  Women (SD), years 53.2 (11.3) 43.7 (14.0) \<0.001  Men (SD), years 52.6 (11.2) 47.1 (12.5) 0.07  Both sexes (SD), years 53.0 (11.2) 44.9 (13.5) \<0.001 SD, standard deviation; P values show statistical significance between cases and controls [Table 2](#T2){ref-type="table"} shows plasma *H. pylori* antibody titers of cases and controls by sex and age. No significant difference was found in mean antibody titer (standard deviation) between cases and controls by sex; women: 10.7 U/mL (12.5) vs. 15.9 U/mL (27.8) (P = 0.15); men: 12.1 U/mL (9.2) vs. 9.4 U/mL (7.4) (P = 0.21). Additionally, no significant difference was observed in mean antibody titer between cases and controls by age, and antibody titer was not increased with older age. ###### Mean Plasma *Helicobacter pylori* Antibody Titers by Sex and Age Cases  Controls P value ------------ ------- ------------- --------- ------------- ------ Women  ≤39 11 17.8 (17.4) 24 22.7 (41.6) 0.71  40 -- 49 14 8.9 (6.6) 16 16.8 (18.8) 0.14  50 -- 59 22 9.4 (8.9) 18 7.7 (8.6) 0.54  60≤ 25 9.7 (14.7) 7 11.7 (10.6) 0.74  Total 72 10.7 (12.5) 65 15.9 (27.8) 0.15 Men  ≤39 6 11.7 (9.7) 9 6.2 (2.8) 0.13  40 -- 49 2 21.5 (10.6) 9 8.9 (8.9) 0.11  50 -- 59 10 10.6 (9.3) 12 10.0 (8.6) 0.88  60≤ 10 11.9 (9.1) 5 14.8 (4.4) 0.52  Total 28 12.1 (9.2) 35 9.4 (7.4) 0.21 Both sexes  ≤39 17 15.6 (15.1) 33 18.2 (36.0) 0.78  40 -- 49 16 10.4 (8.0) 25 14.0 (16.2) 0.43  50 -- 59 32 9.8 (8.9) 30 8.6 (8.5) 0.60  60≤ 35 10.3 (13.2) 12 13.0 (8.4) 0.52  Total 100 11.1 (11.6) 100 13.6 (23.0) 0.32 Plasma H. pylori antibody titers are expressed as U/mL; SD, standard deviation. [Table 3](#T3){ref-type="table"} shows *H. pylori* infection positivity rates evaluated based on antibody titers. No significant differences in *H. pylori* infection positivity rates were observed between cases and controls grouped according to women only (P = 0.32), men only (P = 0.18), or both sexes (P = 0.89). ###### *Helicobacter pylori* Infection Positivity Rates in Cases and Controls *Helicobacter pylori* infection ------------ ----- --------------------------------- --------- --------- ------------ ------ Women  Controls 65 37 (57) 28 (43) 1 (ref)  Cases 72 47 (65) 25 (35) 0.7 0.3 -- 1.4 0.32 Men  Controls 35 21 (60) 14 (40) 1 (ref)  Cases 28 12 (43) 16 (57) 2.0 0.8 -- 5.4 0.18 Both sexes  Controls 100 58 (58) 42 (42) 1 (ref)  Cases 100 59 (59) 41 (41) 1.0 0.5 -- 1.7 0.89 *H. pylori* infection was defined as antibody titer ≥10 U/mL; OR, odds ratio; CI, confidence interval. Discussion {#sec1-4} ========== This hospital-based case-control study demonstrated that plasma *H. pylori* antibody titer in cases is similar to that in controls. Furthermore, no significant difference was found in *H. pylori* infection positivity rates between cases and controls. Our data suggest that *H. pylori* infection does not play a significant role in the development of gallbladder cancer in Indian people. Kanthan et al., (2015) proposed 4 broad categories of risk factors for gallbladder cancer: patient demographics, gallbladder abnormality, patient exposure, and *Salmonella* and *Helicobacter* infections. Previous studies have reported the association between *H. pylori* infection and gallbladder cancer risk, but findings have been inconsistent (de Martel et al., 2009; Mishra et al., 2010). Of the 3 studies including Indian patients with gallbladder cancer (Sharma et al., 2007; Mishra et al., 2011; Mishra et al., 2013), one case-control study including gallbladder cancer and cholelithiasis patients (Mishra et al., 2011) reported no association between the two patient groups, although the other 2 studies reported the detection of *H. pylori* DNA from gallbladder tissue of cancer patients. Conflicting evidence on the association between *H. pylori* infection and gallbladder cancer risk might have been caused by differences in samples and methods used in each study. Culture, serological examination, histological examination, and PCR analysis are common techniques used to evaluate *H. pylori* infection. Of these methods, PCR analysis is the most sensitive for detecting *H. pylori* in human biological samples. In fact, as summarized in a recent review (de Martel et al., 2009), *H. pylori* DNA was detected in 7 of 8 studies using PCR methods, but not in 2 studies using culture methods. A serological test for evaluating the presence or absence of *H. pylori* infection based on *H. pylori* antibody titer has recently been developed and used. This method is non-invasiveness, convenient, inexpensive, and mainly used as a gastric cancer screening method (Inui et al., 2017). In the present study, to reveal whether *H. pylori* infection is a risk factor for gallbladder cancer development in Indian people, we conducted a hospital-based case-control study of gallbladder cancer patients with gallstones and cholelithiasis patients. We used patients with gallstones as our cases because gallstones are a major risk factor for gallbladder cancer. We found that mean plasma *H. pylori* antibody titer and infection positivity rate in cases were not higher than those in controls. Our findings support the results of Mishra et al., (2011) that *H. pylori* infection does not play a role in gallbladder cancer development in Indian people. They conducted a case-control study including 54 gallbladder cancer patients, 54 cholelithiasis patients. However, of the 54 gallbladder cancer patients, only 40 (74%) had gallstones. To our knowledge, our study is the first gallstone-matched case-control study to examine the association between *H. pylori* infection and risk of gallbladder cancer in Indian people. Thus, further studies are required to confirm our findings. The following limitations must be considered when interpreting our data. First, serological test results are not specific to the gallbladder, therefore, data may reflect *H. pylori* infection in the gallbladder and other organs. Because *H. pylori* has been detected in the stomach, bile, liver, heart, and biliary epithelium (Lacy and Rosemore, 2001; Griniatsos et al., 2009), our data may indicate *H. pylori* infection in these organs. Moreover, no significant difference in *H. pylori* infection was reported between women and men (Singh et al., 2002; Shukla and Tewari, 2012; Adlekha et al., 2013), however, a significant difference among age groups has been reported (Singh et al., 2002; Shukla and Tewari, 2012). In the present study, no significant differences in *H. pylori* infection in cases and controls were observed by sex and age. Thus, *H. pylori* infection in organs other than the gallbladder is unlikely to have made a substantial contribution to the results. If *H. pylori* infection is a risk factor for gallbladder cancer, then mean plasma *H. pylori* antibody titers or infection positivity rates should be higher in cases than those in controls. Second, we defined *H. pylori* infection as plasma antibody titers ≥10 U/mL in accordance with the kit's instructions. The sensitivity and specificity of this method for detecting *H. pylori* are 90--95% (Burucoa et al., 2013). In addition, a recent study conducted in Japan demonstrated that this kit has a gray-zone between 5.6 U/mL and 10 U/mL, where infected patients are detected (Inui et al., 2017). Therefore, our results may not accurately reflect *H. pylori* infection status. However, no significant difference in *H. pylori* infection positivity rate was found between cases and controls, even if we defined positive infection as antibody titers ≥6 U/mL (cases, 65%; controls, 62%; P = 0.66). Third, *Helicobacter* species other than *H. pylori* have been detected from gallbladder tissue using PCR (de Martel et al., 2009; Hamada et al., 2009; Pandey et al., 2010). This evidence suggests *Helicobacter* species may be associated with gallbladder cancer risk, however, we were unable to reveal a role of these species in gallbladder cancer development. Further work will clarify the association between these species and gallbladder cancer risk. Fourth, we selected both gallbladder cancer and cholelithiasis patients as study subjects. In a previous case-control study including cholelithiasis patients and healthy subjects, mean serum *H. pylori* antibody titer in cholelithiasis patients was significantly higher than that in healthy subjects (Deeba et al., 2010). Moreover, an association between *H. pylori* infection and gallstone formation has been reported (Kawaguchi et al., 1996; Lee et al., 2010). If plasma or serum *H. pylori* antibody titer for gallbladder cancer patients, cholelithiasis patients, and healthy subjects are measured, the role of *H. pylori* infection in gallbladder cancer development might be more clearly defined. In summary, our data showed *H. pylori* infection is not related to increased gallbladder cancer risk in Indian people. While our findings require further confirmation, they provide evidence that *H. pylori* infection is not an important risk factor for gallbladder cancer in Indian people. We are indebted to Mr. Jyosaku Kawamura of ALP Inc. for performing the serological assays, and we thank Dr. Christina Croney of Edanz Group China for editing a draft of this manuscript. This work was supported by JSPS KAKENHI (grant number 16K09080).

Introduction {#s1} ============ Sexual attractiveness and individual fitness are intimately linked as animals evolve to recognize the traits that indicate high fitness and reproductive potential in their potential mates [@pone.0049799-Andersson1], [@pone.0049799-MaynardSmith1]. Diet is one of the primary environmental factors that influences fitness, and it has long been known that dietary restriction without malnutrition affects lifespan and reproductive output across a wide range of organisms from nematode worms to mammals [@pone.0049799-Fontana1]. Emerging evidence suggests that it is diet composition, such as the ratio of protein to carbohydrate, rather than caloric content *per se,* that is responsible for the observed effects [@pone.0049799-Piper1], [@pone.0049799-Fanson1], [@pone.0049799-Simpson1]. This may not be surprising as various physiological and sex-specific tasks require particular nutrients, and certain allocation decisions, such as those directed towards survival versus reproduction, are often optimized by different diets [@pone.0049799-Fricke1], [@pone.0049799-Maklakov1], [@pone.0049799-Vargas1], [@pone.0049799-Gosden1]. Because diet composition can vary widely over an individual's lifetime, natural selection is likely to favor biological mechanisms that rapidly alter allocation decisions in response to nutrient availability as well as mechanisms in individuals of the opposite sex to evaluate such decisions in their potential mates. Many insects assess reproductive value of potential mates based on the chemical signature of cuticular hydrocarbons (CHCs), which are long-chain lipids synthesized from fatty acid precursors and deposited on the insect cuticle. Their presumed ancestral function is desiccation resistance, but they also play a major role in insect social communication and recognition of species, sex, dominance, and reproductive status [@pone.0049799-Howard1]. CHC composition has been shown to respond to diet and other environmental changes in a variety of insects [@pone.0049799-Gosden1], [@pone.0049799-Savarit1], [@pone.0049799-Kent1], suggesting interactions among major metabolic pathways and CHC biosynthesis. In *D. melanogaster*, a few of the cuticular hydrocarbons, namely female-specific 7,11-heptacosadiene (7,11-HD) and 7,11-nonacosadiene (7,11-ND), have been shown to act as sex pheromones [@pone.0049799-Antony1], with 7,11-HD being the most abundant CHC in non-African strains and also the most potent inducer of male courtship [@pone.0049799-Antony2]. Among the remaining thirty-plus identified CHCs, some have been suggested to exhibit small additive effects on sexual attractiveness [@pone.0049799-Ferveur1], but most have unidentified functions. Very long-chain CHCs (C31--35) are abundant in immature adult flies, when both sexes are vigorously courted by mature males, but these CHCs greatly diminish within a day after eclosion [@pone.0049799-Arienti1]. Recent studies have shown that CHC composition and sexual attractiveness in *Drosophila melanogaster* are strongly influenced by aging [@pone.0049799-Kuo1] as well as by genetic manipulations of the insulin-insulin-like signaling (IIS) and target-of-rapamycin (TOR) pathways [@pone.0049799-Kuo2]. CHCs of young females and of females with increased IIS are more attractive to males than CHCs of old females or of females with decreased IIS [@pone.0049799-Kuo1], [@pone.0049799-Kuo2]. Aging and IIS are both strongly affected by diet [@pone.0049799-Partridge1], but how CHCs are modified by diet composition and whether any observed changes will impact the attractiveness of female pheromone profiles are unclear. While, in general, increased dietary sugar is thought to promote IIS [@pone.0049799-Tirone1] and increased dietary protein putatively activates the TOR pathway [@pone.0049799-Avruch1], these two major nutrient-sensing pathways are known to interact [@pone.0049799-Radimerski1], [@pone.0049799-Lizcano1], making it difficult to predict a resulting phenotype when the activity of both pathways might be altered. For instance, dietary yeast and sugar produce opposing effects on a number of physiological traits in *Drosophila*, including fecundity, triglyceride levels, sleep, and activity patterns, while having similar effects on others, such as longevity [@pone.0049799-Catterson1], [@pone.0049799-Skorupa1]. Female fecundity, in particular, is known to be compromised by aging, reduced insulin signaling, and yeast restriction [@pone.0049799-Skorupa1], [@pone.0049799-Clancy1]. Any one or more of these traits may serve as indicators of individual fitness and therefore may contribute to mate choice. This study set out to test whether dietary protein and sugar affect CHC composition and female attractiveness across different ages. Based on the above considerations, we predicted that increasing dietary yeast and sugar would have significant but opposing effects on these phenotypes. Under the assumption that males would prefer females with higher immediate reproductive potential, we further anticipated that females fed high-yeast diets would be more attractive because they are more fecund [@pone.0049799-Skorupa1]. We find that CHCs are differentially affected by increases in dietary sugar and yeast, and that the differences are exacerbated by age. Surprisingly, however, females fed different diets showed no detectable difference in sexual attractiveness at any age. We discuss potential explanations for these results and draw parallels among the effects of diet, age, and insulin signaling on female CHCs and attractiveness. Materials and Methods {#s2} ===================== Fly Stocks and Husbandry {#s2a} ------------------------ Canton-S flies were obtained from the Bloomington Stock Center. For all experiments, larvae were cultured in cornmeal-sugar-yeast "larval" media. Virgin adults were collected shortly after eclosion and supplied with fresh S10Y10 food (represented in terms of sugar%yeast%) every 2--3 days. A balanced diet in the range of S5Y5 to S10Y10 has been determined to maximize longevity of *D. melanogaster* females in the laboratory [@pone.0049799-Skorupa1]. S5Y5 diet was, therefore, chosen as a base (control) diet for this study, while S20Y5, S5Y20, and S20Y20 represented diets with increased sugar, yeast, or both, respectively. These diets were chosen based on extensive characterization of their effects on fly lifespan and physiology [@pone.0049799-Skorupa1]. Notably, yeast is a major source of protein. However, it also provides the dietary source of other nutrients including lipids, fatty acids, and vitamins [@pone.0049799-Bass1]. Amino acid supplementation alone has been shown to reverse the effects of a low-yeast diet on lifespan and reproductive output; micronutrients and lipids had little effect [@pone.0049799-Fanson1], [@pone.0049799-Grandison1]. On the other hand, even though flies are able to synthesize all the necessary fatty acids *de novo*, they do effectively incorporate dietary sources of fatty acids into a variety of tissues [@pone.0049799-Carvalho1], making it possible that dietary yeast may impact CHC production in a variety of ways. All flies were maintained at 25°C and 60% relative humidity in a 12∶12 h light:dark cycle. Details concerning media recipes can be found in [@pone.0049799-Skorupa1]. We chose to study female CHCs because the effects of diet, age, and IIS are larger in magnitude and better understood mechanistically in this sex [@pone.0049799-Kuo1], [@pone.0049799-Kuo2], [@pone.0049799-Magwere1]. Furthermore, female fecundity is more tightly linked to diet than male sexual performance in a single mating event [@pone.0049799-Skorupa1], [@pone.0049799-Fricke2], [@pone.0049799-Fedina1], [@pone.0049799-Lewis1], suggesting that mate choice based on current nutritional status is an especially advantageous strategy for males. Cuticular Hydrocarbon Analysis {#s2b} ------------------------------ CHCs were analyzed by gas chromatography mass spectrometry (GC-MS) and by laser desorption/ionization orthogonal time-of-flight mass spectrometry (LDI-MS) as described in more detail in [@pone.0049799-Kuo1], [@pone.0049799-Kuo2]. GC-MS analysis provides quantification of individual alkanes, alkenes, and branched hydrocarbon species while LDI-MS is better suited for analysis of more polar lipid components such as hydrocarbons with oxygen functional groups. LDI-MS does not detect alkanes. For GC-MS analysis, a large cohort of flies was established by placing same-age virgin females into 95×25 mm^2^ vials containing 10 ml of one of 4 different media: S5Y5, S20Y20, S20Y5 and S5Y20. Three replicates of 5 flies were sampled at 6, 24, 47 and 61 days of adult age (i.e., days since eclosion). Each sample of 5 flies from a single vial was placed in 100 µl of hexane, which contained 10 µg/ml of hexacosane (Sigma-Aldrich) as an internal standard. Following incubation at room temperature (23°C) for 30 min, the cuticular extract was removed and placed in a clean glass vial. The solvent was then evaporated under a chemical hood for 1--2 h, and the dry extracts were stored at −80°C. After all samples had been collected, each one was re-dissolved in 30 µl of heptane and analyzed by GC-MS analysis. All samples were processed on the same day. The GC-MS analysis was performed as described in [@pone.0049799-Kuo1], [@pone.0049799-Kuo2]. Compounds were identified on the basis of retention time and electron ionization (EI) mass spectra. The signal intensity of each compound was calculated as the area under its corresponding peak. Relative CHC profiles were derived by dividing individual peak intensities by the sum of the intensities of all identified hydrocarbons. To obtain a measure for the total amount of CHC, the sum of the signal intensities of all identified compounds was divided by that of the internal standard. For LDI-MS, multiple independent cohorts were established every 2--3 weeks. Virgin females were sorted onto the four different food media and were sampled on the same day for CHC analysis (described in more detail in [@pone.0049799-Kuo1]). Flies were anesthetized and mounted with fine forceps onto adhesive tape. Mass-spectra were acquired from both the foreleg tarsi and ano-genital regions. These two measures were highly correlated [@pone.0049799-Kuo1], and we chose to focus on the tarsal data for further analysis. The orthogonal mass spectrometer was equipped with an N~2~ laser emitting 3 ns long pulses at a wavelength of 337 nm and a repetition rate of 30 Hz. The laser beam spot size on the sample is ∼200 µm in diameter and has an approximately flat-top intensity profile. Ions were generated in a buffer gas environment using 2 mbar of Argon gas. For acquisition of a mass spectrum, 900 laser pulses were applied to one spot for 30 s. Laser fluence (light energy per pulse and area) was adjusted to values moderately above the ion detection threshold, corresponding to values between 100--200 J/m^2^. All data were acquired in positive ion mode, and mass spectra were processed using the MoverZ software (v. 2001.02.13, Genomic Solutions). Potassiated molecules \[M+K\]^+^ formed the dominant hydrocarbon ion signals in all LDI-MS mass spectra. Elemental composition assignments are based on the assumption that the observed and theoretical mass values agree within +/−0.02 Da. Relative quantification was expressed using a normalized intensity (also referred to as relative abundance, relative amount), derived by dividing individual signal intensities by the sum of the intensities of all identified hydrocarbon signals. Behavioral Assay {#s2c} ---------------- Female attractiveness was tested using a two-choice behavioral assay in which flies were video recorded and then analyzed using fly tracking software [@pone.0049799-Kuo1]. In this assay, two subject females were decapitated and embedded (legs only) in agar 15--20 mm apart and 7--10 mm away from the side of the dish. After the agar solidified, a single, 4--8 day-old virgin Canton-S male was released in the arena and given 10 min to acclimate to the new environment. Video recording was then started and continued for 30 min. Videos were recorded at 2 frames per second and converted to AVI file format, which was analyzed with VideoFly software that was developed in our laboratory. The software was written in C and C\# and was built around the OpenCV image analysis library (<http://opencv.willowgarage.com/wiki/>); it is freely available from our laboratory website (<http://sitemaker.umich.edu/pletcherlab/>). VideoFly calculates the amount of time spent by a focal fly inside a circle of 3 mm radius centered on each decapitated subject fly. Instances where the total time spent inside a defined region was less than 50 s were removed from further analysis to standardize the denominator for proportional data. As with our courtship assay, male preference was calculated as the percentage of time males spent in the circles centered on one of the subject females divided by the total time spent in both. Statistical Analysis {#s2d} -------------------- To visualize and interpret general changes in CHC profiles with age and diet, principal component analysis (PCA [@pone.0049799-Jolliffe1]) was used for the two datasets obtained by GC-MS (26 CHCs) and LDI-MS (12 CHCs). To increase the power of PCA, previous data on CHC changes with age in two fly strains (a standard laboratory Canton-S strain and a recently wild-caught strain called Fv [@pone.0049799-Kuo1]) on a balanced S10Y10 diet were also used to extract principal components (PCs). This resulted in 78 total CHC samples for GC-MS dataset and 135 samples for LDI-MS dataset. PCA was performed on the correlation matrix; three PCs were retained; and Varimax rotation was applied to improve interpretability of component loadings. Because LDI-MS analysis was performed on individual flies, it resulted in much tighter correlation among individual CHCs compared to GC-MS data (where 1 sample = CHCs of 5 flies). Consequently, ANOVA on principal components scores from LDI-MS dataset was used to test the effects of diet, age, and their interaction on each principal component (PC). For individual CHCs, two-factor ANOVAs were applied to test the effects of diet, age, and their interaction, with transformation applied when required to conform the data to a normal distribution. Results of behavioral trials, expressed as a mate preference (i.e., the percentage time a male spent in the courtship area of one of the two females over total courtship time directed to both females), were analyzed using Wilcoxon signed rank test for difference from 50% ( = no choice). All analyses were performed in JMP 9 (SAS Institute Inc). A sequential Bonferroni correction was applied when assessing the impact of high dietary sugar versus high yeast on individual CHC, and differences were considered significant in the context of an experiment-wise α \<0.05. Qualitative meta-analysis comparing directions of changes in individual CHCs in response to age, diet and insulin signaling was done using the data from this study and from our two previous studies [@pone.0049799-Kuo1], [@pone.0049799-Kuo2]. When two or more datasets were used to derive P-values for CHC response to the same factor, direction and statistical significance of change in each individual CHC was tabulated based on the prevalent direction and significance. Results and Discussion {#s3} ====================== Global Responses of CHC Profiles to Diet {#s3a} ---------------------------------------- The total amount of CHCs detected by GC-MS was affected strongly by diet. While flies from all diets exhibited similar levels of CHCs early in life, there was a marked (nearly two-fold) increase in total CHCs on high-yeast diets with age, irrespective of the amount of dietary sugar ([Fig. 1](#pone-0049799-g001){ref-type="fig"}). Flies fed low-yeast diets maintained roughly equivalent levels of CHC throughout their life. The amount of dietary sugar had no effect on total CHC levels or their change with age. One possible interpretation of these results is that the trends reflect the ability of reduced protein to slow many aspects of the aging process and reduced dietary yeast to extend lifespan [@pone.0049799-Skorupa1], [@pone.0049799-Ambuhl1], [@pone.0049799-Vigne1]. However, given the observation that the low-yeast diet does not simply slow age-dependent changes in total CHC levels but eliminates them entirely, it is more likely that production of some CHCs, particularly those that experience increased synthesis with age, is stimulated by dietary protein, or possibly, by other compounds in the yeast. Whether these CHCs have defined functions and dietary yeast is required for their synthesis or whether these CHCs are just by-products of changes in other metabolic pathways remain open questions. {#pone-0049799-g001} To gain insight into the global response of CHC profiles to diet, we first visualized relative amounts of individual CHCs using principal component analysis. For 26 CHCs identified by GC-MS analysis, three retained PCs explained 56% of all variation ([Fig. 2](#pone-0049799-g002){ref-type="fig"}, [Fig. S1](#pone.0049799.s001){ref-type="supplementary-material"}). To compare the age-dependent changes in each of the diets, we plotted PC scores in 3D space and identified the major trends with an "aging cones". These cones resulted from fitting an orthogonal regression line through all samples of the same diet treatment, with the base of the cone representing the youngest ages. The plot provides several useful insights. First, it illustrates that females fed the S20Y5 diet exhibited CHC profiles that were clearly distinct from the profiles measured on females from the other diets ([Fig. 2A](#pone-0049799-g002){ref-type="fig"}, focus on the blue cone). Second, the differences among the diets increased with age (i.e., the Euclidean distance between the cones increased towards their tips). Third, the two unbalanced diets, S20Y5 and S5Y20, diverged most dramatically ([Fig. 2A](#pone-0049799-g002){ref-type="fig"}, compare red and blue cones), and the balanced diets of S5Y5 and S20Y20 (as well as S10Y10 in [Fig. S1](#pone.0049799.s001){ref-type="supplementary-material"}) affected CHCs in an intermediate fashion between sugar- and yeast-biased diets. Fourth, the S20Y20 cone more closely followed changes induced by S5Y20 diet ([Fig. 2A](#pone-0049799-g002){ref-type="fig"}, the green cone is closest to red), which indicates a dominance of yeast effects over sugar. Similarities between S20Y20 and S5Y20 can be partly explained by the increase in total CHCs on high-yeast diets ([Fig. 1](#pone-0049799-g001){ref-type="fig"}). However, changes in total CHC levels cannot explain all of the observed responses to diet because the CHC profiles of S5Y5 females ([Fig. 2A](#pone-0049799-g002){ref-type="fig"}, grey cone) and of S20Y5 females are significantly different ([Fig. 2A](#pone-0049799-g002){ref-type="fig"}, blue cone) despite exhibiting similar total CHC amounts ([Fig. 1](#pone-0049799-g001){ref-type="fig"}). {ref-type="supplementary-material"}.](pone.0049799.g002){#pone-0049799-g002} Some interpretation of the meaning of each PC in terms of the original CHCs can be provided by examining the PC loadings ([Fig. 2B](#pone-0049799-g002){ref-type="fig"}). Thus, PC1 is characterized by positively loading short-chain CHCs (C21--24) and by negatively loading Me-group CHCs. Females fed a S20Y5 diet ([Fig. 2A](#pone-0049799-g002){ref-type="fig"}, blue cone) have high PC1 values, which decline slowly with age, while females fed a S5Y20 diet ([Fig. 2A](#pone-0049799-g002){ref-type="fig"}, red cone) have lower PC1 values early in life that decline rapidly in an age-dependent manner. PC2 is represented primarily by medium-long-chain CHCs, and it includes positively loading dienes 7,11-TD, 9,13--C25∶2, 7,11-PD, C26∶2, and 7,11-HD (the major identified female sex pheromone) and negatively loading alkanes C25∶0 and C27∶0. PC2 scores for all diets except S20Y5 decrease with age. PC3 is represented by a number of mostly long-chain CHCs, including 5,9-HD, 9--C27∶1, 7-H, and 7,11-ND, and it distinguishes high-yeast diets (S5Y20 and S20Y20) from low-yeast diets (S5Y5 and S20Y5). For flies from high-yeast diets, PC3 scores are not significantly different at different ages, while they increase with age for flies from low-yeast diets ([Fig. 2A](#pone-0049799-g002){ref-type="fig"}). PCA was also applied to a dataset comprised of the 12 CHCs identified by LDI-MS analysis [@pone.0049799-Yew1]. The three retained PCs explained 80% of the total CHC variation, and the PC structure was easily interpretable--most individual CHCs loaded heavily on one of the PCs ([Fig. 3A](#pone-0049799-g003){ref-type="fig"}). PC1 is strongly represented by three O-containing compounds, C~23~H~46~O, C~25~H~50~O, C~29~H~58~O, and two very long-chain CHCs, C~33~H~64~ and C~35~H~68~. S5Y20 females exhibited the lowest value for PC1, while females from the S20Y5 diet had the highest PC1 values. Statistically, PC1 effectively separates females fed a S5Y20 diet from the remaining three diet treatments ([Fig. 3B](#pone-0049799-g003){ref-type="fig"}, top panel). Very long-chain CHCs (C31--35) are prevalent in young flies, but decreased significantly within few days after eclosion [@pone.0049799-Arienti1], suggesting that their further loss in aging females fed S5Y20 may reflect accelerated senescence induced by the unbalanced and protein-rich nature of the diet ([Fig. 3B](#pone-0049799-g003){ref-type="fig"}, red line in top panel). PC2 ([Fig. 3B](#pone-0049799-g003){ref-type="fig"}, middle panel) is represented by positively loading C~25~H~48~ (presumably, 7,11-PD) and C~27~H~52~ (7,11-HD) and by negatively loading C~29~H~56~ (7,11-ND). This PC was not particularly associated with any one diet, but its scores universally decreased with age. Two Oxygen-containing compounds, C~27~H~54~O and C~27~H~54~O~2~, load heavily on PC3, and this PC clearly distinguishes females fed high- and low-yeast diets ([Fig. 3B](#pone-0049799-g003){ref-type="fig"}, bottom panel). Several insects have been shown to metabolize CHCs by hydroxylation and oxidation into O-containing compounds that serve as contact pheromones [@pone.0049799-Blomquist1]. The functions of Oxygen-containing CHCs in *D. melanogaster* females, however, are unknown (but see [@pone.0049799-Yew2] for possible functions in males). {ref-type="supplementary-material"}.](pone.0049799.g003){#pone-0049799-g003} In summary, the PCA analysis suggests several global trends. First, most individual CHCs can be assigned to one of several groups, each of which responded characteristically to diet. Second, the most dramatic differences in CHC profiles were between sugar- and protein-rich diets. Third, diet-mediated differences in CHCs profiles increased with age. Fourth, dietary yeast tended to exert dominant effects over sugar on the relative abundances of individual CHCs as well as on total CHC amount. Contrasting Effects of Dietary Yeast and Sugar on Individual CHCs {#s3b} ----------------------------------------------------------------- Examination of diet- and age-dependent changes in individual CHCs reveals a number of interesting trends. Consistent with our interpretation of the PCA, we found that dietary yeast and sugar produced contrasting effects on many individual CHCs ([Fig. 4A and B](#pone-0049799-g004){ref-type="fig"}, note the log-scale on the Y axes). Following correction for multiple testing, several CHCs, namely C22∶0, 7,11-TD, 9--C23∶1, 9-P, 7,11-HD, and C~25~H~48~ significantly increased in relative abundance following an increase in dietary sugar (S5Y5 vs. S20Y5; [Fig. 4A](#pone-0049799-g004){ref-type="fig"}). These initial differences persisted throughout the lifespan. On the other hand, the relative abundances of several CHCs were reduced in females fed high-yeast diets (S5Y5 vs. S5Y20); 8/26 CHCs from the GC-MS dataset and 4/12 CHCs from the LDI-MS dataset were significantly decreased, with effects often magnified by aging. Unlike most other CHCs, a methyl-group compound, 2-MeC~26~, that constitutes between 10--40% of the total amount of CHCs detected by GC-MS, experienced a roughly two-fold increase in its relative abundance with age on the high-yeast (S5Y20) diet. Levels of this compound were significantly reduced and independent of age in females fed a sugar-biased diet ([Fig. 4B](#pone-0049799-g004){ref-type="fig"}, [Fig. S2](#pone.0049799.s002){ref-type="supplementary-material"}). These data complement a growing body of literature suggesting that dietary yeast and sugar have opposing effects on a range of phenotypes including triglyceride levels, fecundity, and sleep/activity patterns [@pone.0049799-Catterson1], [@pone.0049799-Skorupa1]. It seems likely that this represents a general phenomenon. {#pone-0049799-g004} LDI-MS analysis revealed that increased dietary yeast resulted in a significant upregulation of the oxygen-containing compound C~27~H~54~O~2~, which increased 3-fold with age on S5Y20 diet ([Fig. 4B](#pone-0049799-g004){ref-type="fig"} and [Fig. S3](#pone.0049799.s003){ref-type="supplementary-material"}). This compound may, therefore, serve as an indicator of the amount of protein or lipid in female diet, the former of which is known to correlate strongly with fecundity in mated females [@pone.0049799-Skorupa1]. LDI-MS analysis was generally consistent with the GC-MS results for those compounds that were detected with both methods (e.g. C~27~H~52~ = 7,11-HD, and C~29~H~56~ = 7,11-ND). In addition to the main effects of dietary sugar and yeast, our experimental design provides the ability to determine the effects of balanced versus unbalanced diets and to assess specific CHCs for which the dietary components exhibit dominance or additivity in their effects. For the majority of CHCs analyzed using GC-MS, relative abundances on balanced diets were roughly intermediate to the relative abundances observed on yeast- and sugar-rich diets (i.e., grey lines/square symbols and green lines/inverted triangle symbols lie close to each other and between red lines/round symbols and bluelines/triangle symbol, [Fig. S2](#pone.0049799.s002){ref-type="supplementary-material"}). For these compounds, it appears that individual dietary components exert independent/additive effects on CHC profiles, perhaps by altering substrate availability for different CHC synthesis pathways (see below). On the other hand, the S20Y20 diet often resulted in relative CHC abundances that were more similar to those observed in females from the yeast-rich (5S/20Y) diet than they were to females from the sugar-rich (S20Y5) diet (green line is closer to the red than to the blue line in [Fig. S2](#pone.0049799.s002){ref-type="supplementary-material"}); these compounds are regulated predominantly by dietary yeast. Two pairs of CHCs -- 2-MeC~26~ with 7,11-HD, and 2-MeC~28~ with 7,11-ND -- stand out because of the similarity in molecular weight within each pair and because they are clearly the most abundant set of compounds detected by the GC-MS analysis [@pone.0049799-Kuo1]. These two pairs exhibited interesting responses to balanced and unbalanced diets. Within each pair, levels of the two CHCs exhibited opposing responses to yeast-rich and sugar-rich diets but similar responses to balanced diets of different nutritional value (e.g., S5Y5 *vs*. S20Y20; [Fig. S4](#pone.0049799.s004){ref-type="supplementary-material"}, see also [@pone.0049799-Kuo1], [@pone.0049799-Kuo2]). It has been shown that Me-branched and unsaturated CHCs are synthesized via two different pathways, both using the same substrate (Acetyl-CoA), and that 2Me-CHC synthesis utilizes several amino acids as its initial substrate [@pone.0049799-Howard1], [@pone.0049799-Blomquist1]. These two pathways may, therefore, compete for a common substrate, with their balance shifted in favor of the methyl-group compounds when dietary protein is replete and to the unsaturated CHC when it is limited ([Fig. S4](#pone.0049799.s004){ref-type="supplementary-material"}). Regardless of the mechanism, the relative representation of Me-branched CHCs compared to their unsaturated straight-chain CHC partners, 7,11-HD and 7,11-ND, seem to be a key trait that reflects the ratio of sugar versus yeast in the female diet. It will be important to determine whether Me-group compounds constitute attractive or repellent CHCs as well as whether they interact with 7,11-HD and 7,11-ND to determine overall female attractiveness. For instance in long-horned beetles, 2-MeC~26~ and 2-MeC~28~ have been shown to be the major female sex pheromones [@pone.0049799-Spikes1]. In *D. melanogaster*, however, the relative abundance of 2-MeC~26~ is increased (decreased) in females with decreased (increased) IIS, which are less (more) attractive to males ([Fig. 5](#pone-0049799-g005){ref-type="fig"}, see also [@pone.0049799-Kuo2]). ![Comparison of diet, age, and insulin signaling effects on CHCs.\ Arrows indicate an increase or decrease in the relative abundance of individual CHCs: 1) for aging - from young to old age on balanced diets [@pone.0049799-Kuo1], 2) for protein -- based on change from S5Y5 to S5Y20, and for sugar -- based on change from S5Y5 to S20Y5, 3) for insulin signaling (IIS) - change in transgenic flies with increased insulin signaling (via insulin receptor overexpression - InR^OX^) or decreased insulin signaling (insulin substrate mutants *chico*) compared to controls [@pone.0049799-Kuo2]. Yellow and blue background indicates statistically significant (at α = 0.05) decrease and increase, respectively. Our previous studies determined that increased insulin signaling or young age makes females more attractive to males and that decreased insulin signaling or old age makes them less attractive [@pone.0049799-Antony1], [@pone.0049799-Antony2].](pone.0049799.g005){#pone-0049799-g005} A comparison of how diet-dependent changes in CHC profiles correspond to those previously observed following aging and alteration of IIS may provide insights into the function of individual CHCs as well as the mechanisms of their regulation ([Fig. 5](#pone-0049799-g005){ref-type="fig"}). For example, the decreased relative abundance of short-chain CHCs that we observed in females fed a yeast-rich diet is also characteristic of old females kept on balanced diets [@pone.0049799-Kuo1] and of transgenic females with decreased IIS [@pone.0049799-Kuo2], suggesting that these physiological states may share some underlying regulatory mechanisms. Increased dietary protein may promote TOR signaling, which would result in increased protein synthesis and more rapid senescence [@pone.0049799-Fanson1], [@pone.0049799-Hietakangas1], [@pone.0049799-Katewa1]. Therefore, the decrease in short-chain CHCs in flies from high-yeast diets might reflect the idea that they are physiologically older than their siblings of the same chronological age fed low-yeast diets. Alternatively, ovarian activity may represent one of the important regulators of female CHC profiles, as has been shown for social insects [@pone.0049799-Blomquist1]. Ovarian activity is low in aged females, females with decreased IIS, and females fed a low-yeast diet [@pone.0049799-Richard1]. Therefore, compounds changed similarly by all these three conditions, such as increased levels of 7,11-ND, may indicate lower ovarian activity. Similarly, for short-chain CHCs, a high-sugar diet mimics the effects of increased insulin signaling ([Fig. 5](#pone-0049799-g005){ref-type="fig"}), which is consistent with modulation of this pathway by dietary carbohydrates [@pone.0049799-Lin1]. While changes in short-chain CHC may be indicative of common regulatory mechanisms, there is less correspondence among changes in long-chain CHCs in response to age, diet, and IIS. For example, while the changes in shorter-chain (C21--26) and Me-branched CHCs in response to increased dietary yeast are similar to those induced by decreased IIS and aging, changes in the levels of long-chain dienes (7,11-HD and 7,11-ND) to the same diet mimic increased IIS ([Fig. 5](#pone-0049799-g005){ref-type="fig"}). Similarly, for sugar-rich diets changes in shorter-chain CHCs (C21--26) and 2Me-C~26~ phenocopy those caused by increased IIS, while changes in 7,11-HD and 7,11-ND phenocopy decreased insulin signaling. Interestingly, increased dietary sugar has been shown to inhibit elongation and promote desaturation of fatty acids in *Drosophila* larvae [@pone.0049799-Geer1]. Effect of Diet on Female Attractiveness {#s3c} --------------------------------------- We have shown previously that the changes in CHC profiles induced by aging and by manipulations of IIS impact sexual attractiveness [@pone.0049799-Kuo1], [@pone.0049799-Kuo2]. In those studies, males exhibited a 70--80% preference towards young (vs. old) females and a 60--75% preference for transgenic females with higher IIS. In the current study, we found that diet significantly impacts CHC profiles, suggesting that it might affect attractiveness as well. Surprisingly, we obtained no evidence for significant dietary effects on attractiveness. Females fed yeast-rich (S5Y20) or sugar-rich diets (S20Y5) did not consistently differ in attractiveness to males when tested against S5Y5 females ([Fig. 6A,B](#pone-0049799-g006){ref-type="fig"}). Because females fed the two unbalanced diets exhibited the largest difference in their CHC profiles, they were also placed in competition with one another in our choice assay. However, we failed to observe consistent differences over measures spanning most of female lifetime ([Fig. 6C](#pone-0049799-g006){ref-type="fig"}). A few trials indicated modest statistically significant differences between dietary treatments, but the direction of the effect varied from trial to trial and none remained significant after a Bonferonni correction for multiple testing. Taken together these data indicate that the broad range of dietary conditions that was used in these experiments does not affect female attractiveness. {#pone-0049799-g006} Interestingly, a positive effect of sugar (or negative effect of yeast) on female attractiveness has been reported for *D. melanogaster* fed sugar-only (S10) or S10Y10 diets [@pone.0049799-Cook1]. Despite using similar conditions, however, we failed to observe differences in attractiveness between females fed sugar-only diet and females fed a balanced S10Y10 diet ([Fig. S6](#pone.0049799.s006){ref-type="supplementary-material"}). This inconsistency may result from differences in fly husbandry or in experimental conditions or protocols. For instance, Cook and Cook [@pone.0049799-Cook1] used decapitated females that were not fixed in place, which allowed them to express a limited range of normal behaviors (such as preening or basic responses to male courtship; *personal observation*). Our protocol requires the females' legs to be fixed in agar, and we have shown that this behavioral assay results in male choice based on CHC perception and not on other female traits or behaviors [@pone.0049799-Kuo1], [@pone.0049799-Kuo2]. Finally, combined analysis of changes in attractiveness in response to diet, aging, and IIS, indicates that attractiveness is likely to be determined by a combination of several attractive and repulsive CHCs. Previous studies suggested that it is female unsaturated CHCs (especially female-specific dienes), and, in particular 7,11-HD, that are biologically active and that alkanes do not induce male courtship [@pone.0049799-Antony1], [@pone.0049799-Antony2]. Therefore, we asked whether diet composition is comparable to aging and IIS manipulation in its effects on the prevalence of major CHC classes, such as those defined by their carbon chain length or the number of double bonds or Me-group present. While all these manipulations seem to affect the proportion of the four major CHC classes, there is no pattern consistent with effects on attractiveness ([Fig. S5A](#pone.0049799.s005){ref-type="supplementary-material"}). Similar to what we observed previously [@pone.0049799-Kuo1], [@pone.0049799-Kuo2], however, we did observe a noticeable trend in the effects of diet on CHCs of different carbon chain length ([Fig. S5B](#pone.0049799.s005){ref-type="supplementary-material"}). In the yeast-rich diet we observed an age-dependent increase in the representation of long-chain CHCs and a corresponding decrease in short-chain CHCs, as evidenced by significant age\*chain-length interaction. The sugar-rich diet resulted in an increased prevalence of short-chain CHCs and a decrease in long-chain CHCs, which was largely unchanged with age ([Fig. S5B](#pone.0049799.s005){ref-type="supplementary-material"}). However, this similarity in the effect of diet and age on CHC chain length did not result in similar effect of these manipulations on attractiveness, suggesting that other characteristics of CHC profiles must be important. In addition, the lack of correlation between attractiveness and relative abundance of 7,11-HD in females with altered IIS [@pone.0049799-Kuo2] suggests that other CHCs must influence attractiveness, perhaps by masking or opposing the effect of 7,11-HD. Notably, changes in 2-MeC~26~ and 2-MeC~28~ on unbalanced diets of variable quality are accompanied by opposing changes in 7,11-HD and 7,11-ND, while aging and reduced IIS on balanced diets each lead to positively correlated changes in these CHCs. We should note that shorter-chain CHCs respond similarly to increased dietary yeast, aging, and decreased IIS, but long-chain CHCs respond quite differently across these treatments ([Fig. 5](#pone-0049799-g005){ref-type="fig"}). It is possible, therefore, that changes in individual CHCs work in opposition, effectively cancelling each other out, and leaving no overall effect on attractiveness. In other words, it may be that the benefits of increased levels of certain attractive CHCs may be limited by reductions in other attractive CHC or increases in repulsive CHCs. Conclusions {#s3d} ----------- This study found consistent and significant effects of diet composition on female CHC profiles. PCA analyses and comparison of CHC levels across a range of different diets suggested that 1) most individual CHCs tend to cluster into a small number of groups, each of which respond to diet differently; 2) dietary sugar and yeast generate opposite changes in CHC profiles; 3) yeast effects are generally dominant over those of sugar; and 4) the magnitude of diet-dependent effects is increased with age. A simple meta-analysis of diet-induced changes in CHCs together with those caused by aging and IIS [@pone.0049799-Kuo1], [@pone.0049799-Kuo2] provide potentially valuable insight into the function of individual compounds and their underlying regulatory mechanism. Our analysis suggests, for example, that responses of short-chain CHCs to diet might be mediated by the IIS pathway, while long-chain CHCs might have a separate regulatory mechanism. Despite the profound changes in CHC profiles caused by different diets, these changes failed to lead to differences in male mating preference for females fed different diets. The lack of dietary effects on attractiveness is particularly surprising in the light of our previous data, which established that changes in CHCs profiles following aging and genetic manipulation of IIS strongly affect female attractiveness [@pone.0049799-Kuo1], [@pone.0049799-Kuo2]. One potential explanation for these observations is that diet-related effects are complicated, and integrated over multiple, nutrient-dependent molecular pathways. This may lead to changes in some CHCs that reduce attractiveness and others that potentiate it, thereby sending conflicting messages to potential mates. In support of this hypothesis, changes in short-chain and Me-branched CHCs in response to increased dietary protein phenocopy changes with aging and decreased IIS, while changes in the long-chain CHCs--specifically the most abundant compounds 7,11-HD, and 7,11-ND--do not exhibit similar trends. In fact, these often change in the opposite direction from that predicted based on changes in short-chain CHCs. Our results have several implications for CHC biology and evolutionary theory. First, we have identified several compounds, including predominant Me-branched and oxygen-containing CHCs, that respond differentially to diet. These CHCs deserve further investigation as potential attractive or repulsive pheromones in *Drosophila*. Second, in terms of evolutionary theory, it is known that dietary changes in natural and laboratory insect populations at both larval and adult stages often lead to changes in CHC composition, nestmate recognition, and mate preferences [@pone.0049799-Gosden1], [@pone.0049799-Etges1], [@pone.0049799-Etges2], [@pone.0049799-Liang1]. In the present study, we have not observed consistent effects of diet-induced changes in CHCs on female attractiveness, suggesting that not all dietary-induced CHC changes generate differences in attractiveness. Finally, the experimental and analytical approaches used in this and our previous studies, which include genetic and environmental manipulations that break down correlations among individual and groups of CHCs, followed by multivariate analysis of these changes, may be valuable for identifying CHCs that are important for attractiveness, and at the same time, for determining the underlying biosynthetic pathways responsible for the observed CHC changes. Together, these considerations may clarify why certain CHCs are used in mate choice. CHCs that define attractiveness may be the very ones that accurately reflect the activity of the evolutionarily conserved molecular pathways that are critical determinants of reproductive success and overall fitness [@pone.0049799-Kuo2]. Supporting Information {#s4} ====================== ###### **Principal component analysis of CHCs detected by GC-MS in** ***D. melanogaster*** **females fed different diets.** The plot represents the same data as in [Fig. 2](#pone-0049799-g002){ref-type="fig"} with the addition of two more aging cones for Canton-S (purple) and a second wild type strain called Fv (yellow), which we used to increase the power of the PCA analysis. Both strains were fed S10Y10 diet; additional details are presented elsewhere [@pone.0049799-Kuo1]. The similar directions of the purple and yellow cones in 3D space indicate that genetic background has minimal effect on CHC aging dynamics and that the measures are highly repeatable. The intermediate position of these two cones between the balanced treatments of S5Y5 (grey) and S20Y20 (green) suggests a dose-dependent effect of diet. (TIF) ###### Click here for additional data file. ###### **Aging dynamics of individual CHCs identified with GC-MS analysis in** ***D. melanogaster*** **females fed four different diets**. X-axis indicates the age of females (7, 23, 49, and 65 days). P-values for the effects of diet, age, and diet by age interaction are coded as \*(P\<0.05), \*\*(P\<0.01), \*\*\*(P\<0.001), and ns (P\>0.05). Colored symbols indicate food treatments: Grey squares = S5Y5, Red circles = S5Y20, Blue up-triangles = S20Y5, Green down-triangles = S20Y20, and lower-case letters on the right, where different, indicate statistically significant difference (at α\<0.05) between diets by Tukey HSD tests. (TIF) ###### Click here for additional data file. ###### **Aging dynamics of individual CHCs identified with LDI-MS analysis in** ***D. melanogaster*** **females fed four different diets**. X-axis indicates the age of females (7, 28, 43, and 58 days). P-values for the effects of diet, age, and diet by age interaction are coded as \*(P\<0.05), \*\*(P\<0.01), \*\*\*(P\<0.001), and ns (P\>0.05). Colored symbols indicate food treatments: Grey squares = S5Y5, Red circles = S5Y20, Blue up-triangles = S20Y5, Green down-triangles = S20Y20, and letters, where different, specify statistically significant difference (at α\<0.05) between diets by Tukey HSD tests. (TIF) ###### Click here for additional data file. ###### **Relationships between relative abundance of 2Me-C~26~ and 7,11-HD on different diets.** There is a significant negative correlation between the relative abundances of these compounds, which may reflect competing substrate use by the two biosynthetic pathways for dienes and Me-group CHCs. The numbers designate fly ages: 1 = 7d, 2 = 23d, 3 = 49d, 4 = 65d, and colors indicate diet treatment (Grey = S5Y5, Red = S5Y20, Blue = S20Y5, Green = S20Y20). (TIF) ###### Click here for additional data file. ###### **Effects of diet on major CHC classes and chain length:** (**A**) A proportional representation of the major CHC classes (alkanes, monoenes, dienes, and Me-alkanes) in CHC profiles of *D. melanogster* females in response to differences in (left to right): reduced insulin signaling through mutation of *chico*; increased insulin signaling through overexpression of the insulin receptor, *InR*; aging (7 *vs* 49 day old flies); diets rich in either sugar or protein. (**B**) Change in the relative abundance of individual CHCs of variable chain length in response to high-sugar (S20Y5 minus S5Y5) or high-yeast (S5Y20 minus S5Y5) diets. P-values associated with regression analysis of % change on carbon chain length are presented in each age panel. A multiple regression analysis indicates significant effects of age (P = 0.006), chain length (P = 0.02), and their interaction (P = 0.03) on the change in the proportions of individual CHCs in response to high-protein diet. High-sugar diet significantly altered CHC chain length (P = 0.0003), but neither age (P = 0.31) nor the interaction between age and chain length (P = 0.96) were statistically significant. (TIF) ###### Click here for additional data file. ###### **Effect of regular S10Y10 diet versus sugar-only (S10) diet on female attractiveness**. No consistent male preference was observed for 4--6 day old females fed either diet competed against each other in the two-choice attractiveness assay. Three replicate experiments were performed, and P-values indicate difference from 50% based on Wilcoxon signed rank test. (TIF) ###### Click here for additional data file. We thank K. Marney for her extensive help with the video analysis, A. Dervisefendic and the other members of the Pletcher laboratory for help with *Drosophila* husbandry and comments on the experimental design. We thank H. Luftmann (University of Münster) for help with GC/MS analysis and interpretation and A. Roy (Temasek Life Sciences Laboratory) for technical assistance. We are also grateful to W. Etges and two anonymous reviewers for their insightful comments on the final version of the manuscript. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: TYF T-HK HAD JYY SDP. Performed the experiments: TYF T-HK JYY. Analyzed the data: TYF T-HK JYY SDP. Contributed reagents/materials/analysis tools: TYF KD HAD SDP. Wrote the paper: TYF SDP.

Introduction {#s1} ============ Airborne bacteria in the indoor environment are the confirmed or presumed causative agents of several infectious diseases, and their components are linked to the development and exacerbation of chronic respiratory illness including asthma [@pone.0034867-Fields1], [@pone.0034867-Douwes1], [@pone.0034867-Li1], [@pone.0034867-Peccia1], [@pone.0034867-Falkinham1], [@pone.0034867-Lamoth1]. These associations are important in industrialized countries and in cities of emerging nations where people spend at least 85% of their time indoors [@pone.0034867-Klepeis1], [@pone.0034867-Brasche1], [@pone.0034867-Yang1]. Developing a fundamental understanding of the origins and character of biological aerosols is therefore a research priority for reducing human exposure to airborne pathogens and bacterial toxins in the indoor environment [@pone.0034867-National1]. Studies based on indoor/outdoor mass balance and receptor-based source apportionment models have demonstrated that, in addition to particles suspended in outdoor air, material resuspended from surfaces as a result of human activities is an important source of indoor airborne particles [@pone.0034867-Ferro1], [@pone.0034867-Koistinen1], [@pone.0034867-Koperrud1]. Other significant sources of indoor airborne bacteria may be human oral and respiratory fluid emitted via coughing, sneezing, talking, and breathing [@pone.0034867-Nicas1], [@pone.0034867-Johnson1], [@pone.0034867-Xie1] or the direct shedding of skin-associated microbiota [@pone.0034867-Noble1], [@pone.0034867-Mackintosh1], [@pone.0034867-Fox1]. Ribosomal rRNA sequences that are homologous to the sequences of bacteria commonly present on human skin have been found in indoor floor dust [@pone.0034867-Taubel1] suggesting that resuspension of this dust may also act as a human-associated source of airborne bacteria or bacterial constituents. Previous studies have extended characterization of these potential sources to indoor air content by estimating occupancy-associated emission rates of culturable bacteria [@pone.0034867-Scheff1], and --- through the use of biomarker analysis --- tracking the sources of some bacteria isolated from indoor air back to human origin [@pone.0034867-Fox1], [@pone.0034867-Fox2]. However, to date, there are no reports in the literature that directly compare phylogenetically derived indoor air bacterial populations with populations from potential sources including human occupants, ventilation duct air and floor dust. The purpose of the research reported in this paper is to investigate the sources and origins of bacteria in indoor air in a university classroom. We hypothesize that through resuspension and direct shedding, human occupancy strongly influences the concentration and character of bacteria in indoor air. To test this hypothesis, quantitative measurements of airborne particle mass and bacteria concentrations were performed in an instrumented classroom during occupied and vacant conditions. To determine the contributions of other sources and to elucidate the origin (human or environmental) of bacteria suspended in indoor air, phylogenetic libraries were produced for indoor aerosols during occupancy and for potential indoor aerosol sources that include floor dust, ventilation duct air entering the room, and particles collected on the building\'s HVAC filter. The microbial ecology results were further compared to published phylogenetic libraries of the human skin microbiome, outdoor bioaerosols, and indoor floor dust to help assess the relative abundances of source-associated bacteria found in indoor air and to determine if the trends observed in the classroom studied herein would be more generally applicable. This work integrates knowledge of physical indoor aerosol processes with molecular biology-based tools to determine the origins of bacteria in indoor air and complements a recent study that reports the size-distributed emission rates of airborne bacterial populations in classrooms owing to human occupancy [@pone.0034867-Qian1]. Study results provide insight into how humans are exposed to indoor microorganisms originating from the environment and other humans. Such insight can help inform how buildings might be designed, operated, and occupied to reduce human exposure to bacteria that cause adverse health effects. Results {#s2} ======= Room Conditions and Ventilation Configuration {#s2a} --------------------------------------------- Experiments were conducted in a small university classroom during four days under vacant conditions and during three additional days under occupied conditions. Continuously monitored environmental parameters included temperature, relative humidity (RH), and CO~2~ concentrations. For the seven sampling days, the outdoor temperature and relative humidity (mean ± standard error) during the time of sampling were 13.4±0.9°C and 45±6.0%, respectively. Corresponding indoor temperature and relative humidity were 23.5±0.4°C and 28±2.6%. The measured outdoor CO~2~ concentration was consistent with the tropospheric background concentration (390 ppm) and the indoor difference between occupied and vacant periods averaged 230 ppm. Investigation into the building air handling system revealed that the ventilation duct air that supplied the room was a mixture of outdoor air and building return air from other classrooms and offices in the building. The proportion of outdoor air to total air flow varies from 25% to 100% depending on the building\'s heating and cooling needs, and would have been near 50% during the experiments conducted herein based on the outdoor temperature. Before entering the classroom, the air mixture passes through a HVAC filter with a MERV 8 rating. The efficient removal by this filter of airborne particles \>3 µm was confirmed through optical counter measurements ([Figure S1](#pone.0034867.s001){ref-type="supplementary-material"}). Thus, the HVAC filter dust results presented herein represent a cumulative sample of airborne particulate matter dominated by the coarse fraction of outdoor air and indoor return air particles collected over the filter operation period of 6 months from August to February. The ventilation duct supply sample represents mainly the smaller particles that pass through the filter and enter the study room. Aerosol Measurements {#s2b} -------------------- Aerosol samples in this study include indoor, outdoor, and ventilation duct supply PM~10~ (mass of particulate matter in particles 10 µm in aerodynamic diameter or less) and PM~2.5~ (mass of particulate matter in particles 2.5 µm in aerodynamic diameter or less) mass concentrations ([Table 1](#pone-0034867-t001){ref-type="table"}). The samples were obtained during occupied and vacant periods to characterize the influence of occupancy on airborne particles and airborne bacteria. For phylogenetic comparisons, floor dust and HVAC filter dust from the HVAC system\'s filter (from hereon referred to as "HVAC filter dust") were mechanically extracted and then sieved and resuspended to obtain PM~37~ (mass of particulate matter in particles 37 µm in aerodynamic diameter or less), PM~10~, and PM~2.5~ size fractions. A full description of samples collected and analyzed is presented in [Table 1](#pone-0034867-t001){ref-type="table"}. Total mass and bacterial genome concentrations for the PM~10~ and PM~2.5~ size fractions in air are shown in [Figure 1](#pone-0034867-g001){ref-type="fig"}. Occupancy results in an increase in airborne concentrations of both total particle mass and bacterial genome copy numbers (GCN). For indoor PM~10~, mass increased by 15 times (*p* = 0.00001) and GCN increased by 66 times (*p* = 0.001) for occupied conditions compared with the vacant case, while smaller increases of 2.5 times (*p* = 0.015) and 16 times (*p* = 0.02) occurred in PM~2.5~ mass and GCN, respectively. Ratios ± standard error of PM~10~ to PM~2.5~ mass concentrations were 4.9±0.3 for indoor occupied air, 0.8±1.2 for indoor unoccupied air, 1.2±0.9 for occupied outdoor air and 1.1±0.9 for vacant outdoor air, respectively, indicating a strong influence on respirable particles larger than 2.5 µm for the indoor environment when occupied, and substantiating the expectation that occupancy is an important contributor to suspended coarse particulate matter [@pone.0034867-Thatcher1]. {#pone-0034867-g001} 10.1371/journal.pone.0034867.t001 ###### Airborne particulate matter, filter dust, and floor dust samples acquired and analyzed in this study. {#pone-0034867-t001-1} Sample category Sample description Processing No. collected No. used in mass analyses No. used in qPCR analysis No. used in sequencing ----------------------------- ------------------------------------------------ -------------------------------------------------- --------------- --------------------------- --------------------------- ----------------------------------------- Indoor air Indoor air, occupied, PM~10~ Sampled onto PCTE filters 6 6 6 5 Indoor air, occupied, PM~2.5~ Sampled onto PCTE filters 6 6 6 \_ Indoor air, vacant, PM~10~ Sampled onto PCTE filters 8 8 8 \_ Indoor air, vacant, PM~2.5~ Sampled onto PCTE filters 8 8 8 \_ Ventilation duct supply air Ventilation duct supply air, occupied, PM~10~ Sampled onto PCTE filters 3 3 3 4 (3 samples, one sequencing duplicate) Ventilation duct supply air, occupied, PM~2.5~ Sampled onto PCTE filters 3 3 3 \_ Ventilation duct supply air, vacant, PM~10~ Sampled onto PCTE filters 4 4 4 \_ Ventilation duct supply air, vacant, PM~2.5~ Sampled onto PCTE filters 4 4 4 \_ Outdoor air Outdoor air, occupied, PM~10~ Sampled onto PCTE filters 3 3 3 \_ Outdoor air, occupied, PM~2.5~ Sampled onto PCTE filters 3 3 3 \_ Outdoor air, vacant, PM~10~ Sampled onto PCTE filters 4 4 4 \_ Outdoor air, vacant, PM~2.5~ Sampled onto PCTE filters 4 4 4 \_ Floor dust PM~37~ Sieved 12 − − 3 PM~10~ Sieved, resuspended, and sampled on PCTE filters 12 12 12 1 PM~2.5~ Sieved, resuspended, and sampled on PCTE filters 12 12 12 \_ HVAC filter dust PM~37~ Sieved 4 \_ \_ \_ PM~10~ Sieved, resuspended, and sampled on PCTE filters 4 \_ \_ 3 To elucidate potential sources of increased aerosol concentrations during occupancy, experiments were conducted to investigate separately the impacts of resuspension from the carpet during walking and direct shedding from humans. The ratio of the indoor particle number concentration to the outdoor particle number concentration for three experimental conditions are presented in [Figure 2](#pone-0034867-g002){ref-type="fig"}. These conditions included (a) one person walking on the carpet, (b) one person walking on the same floor covered with plastic sheeting to eliminate resuspension of particles from the carpet, and (c) 30 adults occupying the room while the carpet was covered with plastic sheeting. For condition (c), occupants were allowed to moved freely about the room and activities centered on talking, reading, and writing. The results in [Figure 2](#pone-0034867-g002){ref-type="fig"} suggest that significant particle generation in the occupied test room may occur through resuspension of dust deposited on the floor, through direct shedding of particles from human occupants, or both. Cases (a) and (c) resulted in particle number concentrations that were greater than the outdoor concentrations for all size ranges. In case (a) when the carpet was not covered with plastic sheeting (resuspension), these increases were 1.2 to 11 times across the range of particle sizes with an average increase of 5.2 times (*p* = 0.05). In case (c), proportional increases of 1.2--4.5 times with an average of 2.7 times (*p* = 0.18) were observed for the floor covered with plastic when the occupancy level was 30 people. Case (c) is suggestive of shedding rather than resuspension. In both cases, the proportional extent of particle concentration increase rose monotonically with increasing optical particle size throughout the instrument\'s measurement range. {#pone-0034867-g002} A comparison of the bacterial mass percentage of airborne particle and floor dust samples is shown in [Figure 3](#pone-0034867-g003){ref-type="fig"}. Estimates of bacterial mass were computed assuming the mass of a bacterium to be 655 femtograms [@pone.0034867-Ilic1], and an average 16 S rDNA gene copy number of four per bacterium [@pone.0034867-Lee1]. Results displayed in [Figure 3](#pone-0034867-g003){ref-type="fig"} demonstrate that the PM~10~ and PM~2.5~ fractions of resuspended floor dust are enriched with bacteria, compared to indoor air, ventilation duct supply air, and outdoor air. The median bacterial mass percentages of indoor and outdoor airborne particles were less than 0.3%, whereas the bacterial proportion of aerosolized floor dust exceeded 2.2% in both size fractions. Based on a Tukey\'s range test, resulting ranks for bacterial abundance in both PM~2.5~ and PM~10~ cases are resuspended floor dust≫outdoor air\>duct supply air\>indoor air. However, only differences between resuspended floor dust and the three air samples were statistically significant at a 95% confidence level. ![Enrichment of bacteria in airborne particulate matter and floor dust.\ Bacterial mass percentage (100×bacterial mass divided by total particle mass) in indoor air, outdoor air, and duct supply air samples and in the PM~2.5~ and PM~10~ size fraction of resuspended floor dust samples. Mass fractions were estimated assuming an average mass of 655 fg per bacterium [@pone.0034867-Ilic1]. Box and whisker plots have the same interpretation as in [Figure 1](#pone-0034867-g001){ref-type="fig"}.](pone.0034867.g003){#pone-0034867-g003} Phylogenetic Analysis {#s2c} --------------------- Sixteen samples from indoor air, ventilation duct supply air, floor dust, and HVAC filter dust ([Table 1](#pone-0034867-t001){ref-type="table"}) were analyzed for bacterial population composition using the 454 GS-FLX pyrosequencing platform with multiplex identifiers (MIDs). Sample MIDs are presented in [Table S1](#pone.0034867.s004){ref-type="supplementary-material"}. After machine- and method-based quality control, denoising, and chimera checking, 10,675 partial 16 S rDNA gene sequences were generated at an average trimmed length of 500 base pairs (bp). Rarefaction values based on 97% similarity were produced for each sample and were then averaged based on their sample type ([Figure S2](#pone.0034867.s002){ref-type="supplementary-material"}). Rarefaction curves of observed OTUs continued to rise with increasing numbers of sequences, suggesting that further increases in sample size would yield more species. Chao1 richness estimator predicted 3720, 1260, 2990, and 640 OTUs, respectively, for floor dust, HVAC filter dust, indoor air, and ventilation duct supply air ([Figure S1](#pone.0034867.s001){ref-type="supplementary-material"}). The diversity metrics reported here are higher than those previously determined for floor dust, which ranged from 83 to 464 based on the Chao1 approach in conjunction with a cloning and Sanger sequencing method [@pone.0034867-Taubel1], [@pone.0034867-Rintala1]. The relative abundances of the 20 most prominent bacterial taxa from indoor air, ventilation duct supply air, HVAC filter dust, and floor dust are shown in [Figure 4](#pone-0034867-g004){ref-type="fig"}. (Phyla level data are presented in [Figure S3](#pone.0034867.s003){ref-type="supplementary-material"}.) Indoor air, ventilation duct supply air, and floor dust samples show heavy representation from the dominant bacteria previously found to be associated with human skin, hair, and nostrils [@pone.0034867-Roth1], [@pone.0034867-Fierer1], [@pone.0034867-Grice1], [@pone.0034867-Costello1], [@pone.0034867-Frank1]. These five human associated taxa --- *Proprionibacterineae*, *Staphylococcus*, *Streptococcus*, *Enterobacteriaceae*, and *Corynebacterineae* --- comprise 17%, 20%, and 17.5% of all bacteria in samples of indoor air, floor dust, and ventilation duct supply air, respectively. The HVAC filter dust sample demonstrated significant differences from all other samples, being strongly dominated by the *Streptophyta* phylum (chloroplast 16 S rRNA encoding gene from plants) with only minor (3%) representation from the five human-associated taxa described above. {#pone-0034867-g004} To quantitatively compare populations, the similarities and differences between the sample bacterial community structures are presented in relation to principal coordinate analysis (see [Figure 5A](#pone-0034867-g005){ref-type="fig"}) on a weighted-UniFrac basis. Stemming from *p*-test significance evaluation using the Bonferroni correction for multiple comparisons, the bacterial communities characterized in indoor air and duct air during human occupancy were significantly different from the communities collected on the HVAC filter dust sample (*p*\<0.001). Differences were not statistically significant between indoor air and ventilation supply duct air bacterial communities (*p*\>0.1), or between ventilation supply duct air and floor dust communities (*p*\>0.1). Indoor air bacterial communities reveal almost significant differences compared to those of floor dust (0.05\<*p*\<0.1). ![Comparison of indoor bacterial populations.\ (**A**) Weighted UniFrac-based bacterial diversity principal coordinate analysis of indoor air (yellow outlined squares), ventilation duct supply air (orange outlined squares), floor dust (outlined circles) samples, and HVAC filter dust samples (outlined triangles) from this study. (**B**) UniFrac-based bacterial diversity principal coordinate analysis displaying the two coordinates that explain most of the variation between samples from this study (open squares, circles, and triangles) and the bacterial ecology of human skin samples (filled diamonds) from Costello et al. [@pone.0034867-Costello1], outdoor air samples (filled triangles) from Bowers et al. [@pone.0034867-Bowers1], and floor dust samples from Täubel et al. [@pone.0034867-Taubel1] and Rintala et al. [@pone.0034867-Rintala1] (filled circles).](pone.0034867.g005){#pone-0034867-g005} [Figure 5B](#pone-0034867-g005){ref-type="fig"} displays the indoor air, ventilation supply duct air, HVAC filter dust, and floor dust samples in this study along with other samples from published studies on the microbial diversity of potential sources including floor dust, human skin, and outdoor air. In all, 104 samples were evaluated for this comparison: 16 samples from the present study; 12 floor dust samples from nursing homes and private residences [@pone.0034867-Taubel1], [@pone.0034867-Rintala1]; 15 outdoor air samples taken in areas with varying land use types including urban, rural, and agricultural sites [@pone.0034867-Bowers1]; and 61 human skin samples from two female and two male individuals sampled at different times including left and right palms, index fingers and forearms [@pone.0034867-Costello1]. The weighted UniFrac analysis, which encompasses several different environments, demonstrates distinct groupings for aerosol samples, for human skin samples, and for floor dusts samples. The data show broad similarities among outdoor and indoor aerosol bacterial ecology, likely owing to the presence of many environmentally associated organisms in the indoor air samples taken in this study ([Figure 4](#pone-0034867-g004){ref-type="fig"}). Larger differences are observed in floor dust samples across studies, with the floor dust measured here (open blue circles) residing more closely to aerosol samples, and floor dust from nursing homes and private residences (closed blue circles) [@pone.0034867-Taubel1], [@pone.0034867-Rintala1] clustering more closely to human skin samples than to aerosol samples. Discussion {#s3} ========== This study advances knowledge about the sources, origins, and character of bacterial aerosols in indoor settings through two main findings. First, human occupancy produces a marked concentration increase of respirable particulate matter and bacterial genomes. Second, bacteria from human skin and from other environmental sources significantly contribute to indoor air bacterial populations. Biological and PM~10~ aerosol concentrations during occupancy {#s3a} ------------------------------------------------------------- Box and whisker plots of bacterial GCN and total particulate matter concentrations for occupied and vacant cases ([Figure 1A--B](#pone-0034867-g001){ref-type="fig"}) demonstrate that human occupancy produces a 15× increase in PM~10~ mass and a 66× increase in PM~10~ airborne bacterial genomes when compared to the vacant room case. These numbers relate to a PM~10~ increase of 75 µg m^−3^ during occupancy, compared to the average background outdoor concentration of 15 µg m^−3^, and an addition of 65,000 GCN m^−3^ compared to an average background outdoor concentration of 4,600 GCN m^−3^. Analogous increases, but with smaller proportionality factors, were also observed in PM~2.5~ fractions. Here, the increase during occupancy represented 10 µg m^−3^ (against a 12 µg m^−3^ average outdoor background) and 5,600 GCN m^−3^ (against a 1,600 GCN m^−3^ average outdoor background). These trends extend the findings from previous indoor particle resuspension studies, in which a strong direct dependence of resuspension rate on the size of abiotic particles has been reported [@pone.0034867-Thatcher1]. Our findings also reinforce and extend prior observations about the contribution of occupancy to increases in coarse-particle bacterial marker concentrations and the association of bacteria with coarse particles emitted from desquamated human skin [@pone.0034867-Mackintosh1], [@pone.0034867-Fox2]. The increases in airborne total particle mass concentrations during occupancy, above both indoor vacant and outdoor airborne concentrations, are also consistent with a broad range of particulate matter studies in diverse indoor environments. These studies suggest that, in the absence of smoking or cooking, resuspension is a dominant source of airborne particulate matter in occupied indoor environments [@pone.0034867-zkaynak1], [@pone.0034867-Chen1], [@pone.0034867-Qian2]. Three additional lines of evidence reinforce the importance of resuspension in shaping the bacterial populations suspended in indoor air. First, bacterial mass per mass of particles was enriched in floor dust by an order of magnitude ([Figure 3](#pone-0034867-g003){ref-type="fig"}) compared to the bacterial mass percentage in particles collected from indoor air, from outdoor air, or from the ventilation duct supply air. Thus, emissions from this floor represented an enriched source of suspended bacteria per particle mass. Such enrichment also supports our observation that occupancy generates a 66× increase in bacterial GCN in indoor air, greater than the 12× increase in total airborne particle mass. A second line of evidence supporting floor dust as a source of airborne bacteria derives from quantitative comparisons of bacterial population structure in indoor air with potential sources ([Figure 5A](#pone-0034867-g005){ref-type="fig"}): the floor dust bacterial population in the test environment was similar to the ecology of indoor air. Finally, ventilation duct air that supplied the room showed higher concentrations of PM~10~ total mass and bacterial GCN during occupancy than during vacant periods. Such increases indicate that activity throughout the building during human occupancy results in a greater concentration of particulate matter and bacterial aerosols in other rooms and increased concentrations in the return air component of the ventilation system that supplied the study room. Human and environmental origins of bacteria suspended in indoor air {#s3b} ------------------------------------------------------------------- While the qPCR data demonstrate an increase of airborne bacteria due to occupancy and the principal coordinate population-based comparisons point to the importance of the resuspension of floor dust, neither approach fully elucidates the fundamental origin of bacteria in indoor air. Insight into the origin of indoor air bacteria can be gained by considering the most abundant taxa contained in the potential sources. For indoor air, ventilation duct supply air, floor dust, and HVAC filter dust, 17%, 17.5%, 20%, and 3% respectively, of the total bacterial abundance was comprised of human associated taxa --- *Propionibacterineae*, *Staphylococcus*, *Streptococcus*, *Enterobacteriaceae*, and *Corynebacterineae* [@pone.0034867-Roth1], [@pone.0034867-Fierer1], [@pone.0034867-Grice1], [@pone.0034867-Costello1], [@pone.0034867-Frank1], [@pone.0034867-Hamady1]. Unique indicators of the human oral cavity and saliva, including *Fusobacterium* and *Veillonella* [@pone.0034867-Costello1], [@pone.0034867-Lazarevic1], [@pone.0034867-Nasidze1], [@pone.0034867-Cephas1], [@pone.0034867-Nasidze2], [@pone.0034867-Zaura1], were also found in the indoor air and floor dust samples, although at very low abundances of 0.02% and 0.1%, respectively. These oral cavity and saliva-associated taxa were neither found in the ventilation duct supply air nor in the HVAC filter dust samples. The evidence suggests that, although emissions from the oral cavity are present, they were less important contributors to overall indoor airborne bacterial loads than emissions from human skin. Other recent investigations of floor dust have demonstrated the presence of skin-associated taxa [@pone.0034867-Taubel1], [@pone.0034867-Rintala1]. The present study leverages new methods in high-throughput DNA sequencing to extend these ecologies to indoor air as well as to identify potential sources of human-associated bacteria including floor dust and ventilation duct supply air. Results from previous studies and the data collected here demonstrate that, during occupancy, resuspension and direct shedding of microorganisms from humans are potential sources of bacterial aerosol particles. The origin of many of the airborne bacteria is from human skin, hair, nostrils, and the oral cavity. It has been estimated that humans shed roughly a billion skin cells daily [@pone.0034867-Milstone1], with each square centimeter of skin per human hand having a concentration between 10^2^ to 10^7^ bacteria [@pone.0034867-Leyden1]. Desquamated human skin cells are an important contributor to particles in indoor air, and there is strong evidence that bacteria are associated with these skin cells [@pone.0034867-Milstone1]. Skin shedding may influence indoor air concentrations both through skin cells and their fragments directly becoming airborne, and also by deposition of cells onto floors and other surfaces followed by fragmentation and resuspension. [Figure 2](#pone-0034867-g002){ref-type="fig"} demonstrates that room occupancy with the resuspension mechanisms inhibited (through the use of plastic sheeting on the floor) still yields particle number concentrations that are significantly greater than outdoor levels. The indoor air, ventilation supply duct air, HVAC filter dust, and floor dust include taxa of environmental origin such as *Sphingomonadaceae*, *Rhodobacteraceae*, and *Streptophyta* (chloroplasts from land plants). Although environmental organisms are not as clearly defined as organisms of human origin, each of the three listed above have been previously reported in outdoor bioaerosol microbial ecology investigations [@pone.0034867-Bowers1], [@pone.0034867-Brodie1], [@pone.0034867-Lighthart1], [@pone.0034867-Bowers2]. The presence of these environmental taxa in indoor air illustrates the potential importance of outdoor air particles conveyed through infiltration or ventilation and/or the tracked-in contribution of outdoor material to floor dust that is subsequently resuspended [@pone.0034867-Koistinen1]. *Streptophyta* were found in high abundance in the HVAC filter dust samples (∼45%) and at lower abundance (2--4%) in the indoor air, ventilation duct supply air, and floor dust samples. The larger plant particles in outdoor air would be captured efficiently on the HVAC filter (capture efficiency is high for particle sizes larger than 3 µm, [Figure S3](#pone.0034867.s003){ref-type="supplementary-material"}), while individual bacteria may pass through the ventilation filter. The outdoor environment near the building was highly vegetated, being situated on a tree-lined street with maintained lawns and flower gardens, and there were no green plants in the room, nor were they common throughout the building. Sampling occurred in the fall during foliage change. Thus, decomposed material would be widely present outdoors. The finding of *Streptophyta* in indoor air and floor dust likely resulted from some combination of outdoor air infiltration into the building, tracked in dust from outdoors, or tracked in particles on the clothes, skin, and hair of people entering the building. *Streptophyta* abundance of the HVAC filter dust from this study is comparable with previously reported *Streptophyta* enrichments in alpine air (44%) [@pone.0034867-Leaderer1], and urban air (19.9%) [@pone.0034867-Bowers1]. Finally, we note that this study was designed as an in-depth investigation of a single environment characterized by air-exchange rates, occupancy levels, and flooring types that are typical of buildings in industrialized countries. While this design allows for a more mechanistic understanding of the sources and origins of bacteria in indoor air, it does so with the limitation that variation among buildings (and across seasons) was not considered. Future investigations should extend this line of inquiry to multiple environments and should also consider variable occupancy levels and flooring types. Additional chamber-based studies that isolate humans during prescribed activities will be required to determine the range of emission rates and ecologies of directly shed bacteria. The limitation of only one environment sampled was partly ameliorated by considering external samples in the weighted UniFrac-based principal coordinate analysis in [Figure 5B](#pone-0034867-g005){ref-type="fig"}. Phylogenetic data from outdoor air in rural, agricultural, and urban settings and from floor dust in private residences and nursing home facilities were considered along with phylogenetic data from the human skin microbiome. The pooling of studies suggests that while bacterial ecologies present in indoor air or floor dust have consistent contributions of human- and environmentally associated bacteria, a significant amount of variability in these relative contributions occurs from building to building. These differences correspond to known relative abundances of human microbiota in floor dust. Specifically, the floor dust measured herein (open blue circles in [Figure 5B](#pone-0034867-g005){ref-type="fig"}) had only 15 to 20% content of human microbiome taxa and clustered with the aerosol samples, whereas the referenced floor dust studies from commercial buildings and private residences (closed blue circles) had \>75% content of human taxa in each case [@pone.0034867-Taubel1], [@pone.0034867-Rintala1], and resided more closely to human skin samples than to aerosol samples. Overall, these data suggest (as one might have anticipated) that the relative contribution of human-associated bacteria is variable and environment-dependent. Characterizing the influence of design, operational and occupancy differences among buildings that account for these indoor air ecology differences is identified as an important future area of study. Conclusion {#s3c} ---------- The integration of aerosol science with modern microbial ecology has revealed new insights into the sources and origins of airborne bacteria in indoor environments. Quantitative monitoring of indoor and outdoor air revealed that human occupancy is a dominant factor that contributes to the concentration of indoor airborne bacterial genomes. During occupancy, it appears that both resuspension from carpet and direct human shedding contributed to significantly elevate respirable particulate matter and bacterial concentrations above background concentrations. Similarities between indoor air populations and bacteria associated with the human skin microbiome point to the important contribution of human microflora. This work extends previous microbial ecology-based observations of human microflora from floor dust into indoor air, where exposure occurs. An important public health consequence of these results is that, through direct inhalation of resuspended or shed organisms, there is potential for current or previous occupants of a room to contribute substantially to inhalation exposure to bioaerosols. Methods {#s4} ======= Aerosol, floor, and HVAC sampling {#s4a} --------------------------------- The study site was a 90-m^3^ (L = 5.9 m, W = 4.9 m, H = 3.1 m) room whose floor was covered with lightly worn, commercial, medium pile, level-loop carpet. This classroom was located on the first floor of a five-story building on a university campus in the northeastern United States. One classroom wall bordered the outside of the building and the opposing wall contained a doorway opening to a hall. There was no visible water damage or known history of water damage in the building. The location adheres to a continuous cleaning schedule that includes vacuuming every second day and semiannual wet carpet cleaning. The room was mechanically ventilated and students and teachers were asked not to open windows and doors during the sampling campaign. Air movement followed a mixed ventilation configuration, with the ventilation supply air register located near the ceiling and outlets located on the floor at the wall opposite the supply air register. Outdoor samplers were located on the outside window ledge approximately 20 cm away from the window and 1.5 meters from the ground. Indoor air samplers were located near the middle of the room at a height of 1.5 m, and ventilation duct air samplers were placed in the supply duct air discharge register located above the room door. Based on carbon dioxide release and decay experiments measured with a LI-COR 820 CO~2~ analyzer (Licor Environmental, Lincoln, NE) the room air-exchange rate (AER) was determined to be 5.5±1.3 h^−1^ (mean ± standard deviation). Sampling was conducted on three occupied days, and on four vacant days during the fall of 2009. For the three occupied sampling days, the average human occupancy during the cumulative 22.2 hours of sampling was 4.7 persons. All necessary permits were obtained for the described field studies. Experiments were cleared through the university\'s environmental health and safety office and permission was granted from all classroom instructors. [Table 1](#pone-0034867-t001){ref-type="table"} summarizes all particle and dust samples acquired during this study and provides information on the analysis for each sample type. Aerosol samples were collected in duplicate on each of the seven sampling days for both PM~10~ and PM~2.5~ in indoor air. One PM~10~ and one PM~2.5~ sampler each collected outdoor air and ventilation duct supply air. For each of the occupied days, samplers were started when occupancy occurred and then stopped one hour after occupancy ended, typically 5 to 8 hours. For each of the vacant days, the room was sampled for 8 hours during normal classroom hours. Indoor, ventilation duct supply air and outdoor sampler filters were collected and changed after each occupied and vacant day." All PM samples were analyzed for mass and bacterial concentrations, and a subgroup of indoor air, ventilation duct supply air, and HVAC filter dust samples were used for phylogenetic library production in accordance with the schedule reported in [Table 1](#pone-0034867-t001){ref-type="table"}. Each PM sample was collected on a 0.8-µm pore-sized, 37-mm diameter sterile polycarbonate track-etched (PCTE) filter that was loaded into commercial PM~10~ and PM~2.5~ samplers (Personal Exposure Monitors, SKC, Eighty Four, PA, USA) operated at 10±0.5 liters per minute. Occupied samplers were operated for a cumulative 22.2 hours during the three occupied sampling days, with sampling started at the onset of classes and stopped approximately one hour after the end of the last class each day. Unoccupied samplers were operated on weekends during typical classroom hours for approximately nine hours per day. Floor dust was collected using a high-volume vacuum sampler (Eureka MightyMite Canister Vacuum, Eureka Company, Bloomington, IL, USA) fitted with a Mitest adapter and dust filter (Indoor Biotechnologies, Charlottesville, VA, USA) [@pone.0034867-Leaderer1]. Floor dust samples were collected each day for occupied and vacant conditions and each sample was a composite of five randomly selected 20 cm×30 cm portions of flooring in commonly trafficked areas of the room. Prior to analysis, floor dust was processed by means of sieving to produce a 37 µm and smaller size fraction. Sieving results in a more homogenous mixture and selects for the smaller size range of dust that can potentially become aerosolized. A portion of this size fraction was also resuspended in a 0.66 m^3^ chamber and collected onto SKC Personal Exposure Monitoring PM~10~ and PM~2.5~ filters in accordance with the method described by Viau et al. [@pone.0034867-Viau1]. Dust samples from the filter in the HVAC system that processed a blend of outdoor supply air and recirculated inside air were also collected; these samples are referred to as "HVAC filter dust" in this study. As described in the results section, the HVAC filter dust represents an aggregate sample of particles collected over the filter operation period of 6 months (August 2008 to February 2009). Representative portions of the filter material from the top, bottom, left, and right side of the used unit were removed and filter dust collected for subsequent processing and analysis. HVAC filters have previously been used as a sampling mechanism for indoor bioaerosols [@pone.0034867-Viau1]. PM~10~ and PM~2.5~ mass analysis {#s4b} -------------------------------- To determine particulate matter mass concentrations, filters were weighed before and after sampling. Weighing was performed using a precision balance (Mettler Toledo type XP6, Columbus, OH, USA). Static electricity was removed with a polonium α-particle source (Staticmaster static eliminator, NRD, Grand Island, NY, USA) and prior to weighing, filters were equilibrated at constant temperature and humidity (30°C±0.5°C, 31%±2% RH) for at least 24 hours. DNA extraction and quantitative PCR {#s4c} ----------------------------------- The quantification of bacterial genomes from samples of indoor air, outdoor air, ventilation duct supply air, and resuspended floor dust was achieved using TaqMan real-time PCR. A three-stage DNA extraction method specifically developed for low concentration aerosol samples was utilized [@pone.0034867-Boreson1]. Briefly, cells on one half of the PCTE filter were lysed by enzymatic treatment and physical disruption through bead beating. Next, phenol/chloroform isoamyl alcohol extractions were used to isolate nucleic acids. Finally, DNA purification and concentration was conducted using spin columns and reagents from the Mobio PowerMax Soil DNA extraction kit (Mobio, Carlsbad, CA, USA). Exceptions to the cited method included proteinase K incubation at 54°C instead of 37°C, omitting the freeze-thawing cycle during DNA extraction, and omitting the 1-hour, 65°C incubation step prior to bead beating. The spin column was eluted two times in 100 µl of 10 µM Tris buffer (pH = 8). This sample was freeze-dried and re-eluted in 40 µl of 10 µM Tris buffer before amplification. Quantitative PCR was performed using an ABI 7500 fast real-time PCR system (Applied Biosystems, Carlsbad, CA, USA). Universal bacterial primers and TaqMan**®** probes [@pone.0034867-Nadkarni1] targeted the 331 to 797 *E. coli* numbering region of the 16 S rDNA with forward primer 5′-TCCTACGGGAGGCAGCAGT-3′, reverse primer 5′-GGACTACCAGGGTATCTAATCCTGTT-3′, and the probe, (6-FAM)-5′-CGTATTACCGCGGCTGCTGGCAC-3′-(BHQ1). For this assay, 20 µl qPCR mixtures were prepared including 10 µl of 2× TaqMan**®** Universal PCR master mix with 6-carboxy-X-rhodamine (ROX) passive reference dye (Roche Diagnostics, Indianapolis, IN), 2 µl of 0.4 µg ml^−1^ bovine serum albumin, 0.4 µl of each 10 µM primer, 0.8 µl of 5 µM probe, and 5 µl of DNA template. Thermocycler conditions were 2 minutes at 95°C for initial denaturation and 45 subsequent cycles of 15 seconds at 95°C, 45 seconds at 56°C, and 60 seconds at 72°C. Real-time PCR standard curves of genome quantity versus cycle threshold number for bacteria were developed using known amounts of *Bacillus atrophaeus* (ATCC 49337) genomic DNA. To produce standard curves, five independent dilution series were produced corresponding to 10^1^ to 10^6^ genome copies. For presenting bacterial genome quantities, cycle threshold values were calibrated versus total bacterial genomes. The calibration accounted for the ten rRNA operon copies in *B. atrophaeus* and the average of four rRNA operon copies per genome for all bacteria [@pone.0034867-Lee1]. To test for PCR inhibition, standard curves for spiked standard *B. atrophaeus* DNA were produced in aerosol filter and sieved floor dust extracts. No inhibitory effects were observed. DNA extracted from filter field blanks was amplified along with the samples, and, if positive, was subtracted from the values obtained for aerosol samples. Generation of phylogenetic libraries and data analysis {#s4d} ------------------------------------------------------ Phylogenetic library preparation was conducted for sixteen samples and included five indoor occupied PM~10~ filters each from a different sampling day and including one replicate, four occupied duct supply air PM~10~ filters representing each sampling day and a sequencing duplicate, three HVAC filter dust PM~10~ filters each from a separate portion of the filter, three sieved 37-µm floor dust samples each taken on an independent occupied sampling day, and one PM~10~ fraction of resuspended floor dust from one of the 37-µm sieved samples. [Table 1](#pone-0034867-t001){ref-type="table"} provides additional information on the samples used in phylogenetic analysis. Ribosomal RNA encoding genes were amplified using the 343F and 926R primers [@pone.0034867-Liu1], [@pone.0034867-Wang1], which also included sequencing adaptors, keys, and multiplex identifiers. Prior to PCR, primer-dimer formation was screened for each set of primers, barcodes, keys, and sequencing adaptors using OligoAnalyzer 3.1 software. Each PCR reaction was of 25 µl volume and included 1×PCR master mix (Roche Applied Science, Indianapolis, IN), 0.4 µg ml^−1^ of bovine serum albumin, 0.3 µM of each primer, and 3 µl of DNA template. PCR was performed at the following cycling conditions: initial denaturation at 94°C for 5 minutes, and 25 to 35 cycles of 95°C dissociation for 30 seconds, annealing at 47°C for 30 seconds, and extension for 1 minute at 72°C, followed by a final extension at 72°C for 8 minutes. Four PCR reactions were conducted for each sample and amplicons were combined before removing salts and unincorporated primers using a Qiagen MinElute PCR purification kit (Qiagen Inc., Valencia, CA, USA) [@pone.0034867-Bibby1]. Amplicons were visualized on a 1.2% agarose gel and, if necessary, extracted using a Qiagen MinElute gel extraction kit (Qiagen Inc., Valencia, CA, USA). DNA extracts from blank filters were used as negative controls and, as they did not result in amplicons, were not further considered for sequencing preparation. Sequencing was performed at the Yale Center for Genome Analysis. Raw data were subjected to quality control at the machine; keypass, dots, and mixed filters were utilized to assess the quality of the whole read, whereas the quality of read ends was checked by signal intensity and primer filters. Libraries were produced using Roche 454 pyrosequencing and incorporated the GS FLX sequencer and Titanium series chemistry. Quantitative sequence analysis was performed using tools in the Quantitative Insights Into Microbial Ecology (QIIME) package [@pone.0034867-Caporaso1]. Denoising was conducted using Titanium Pyronoise [@pone.0034867-Caporaso1], [@pone.0034867-Quince1], [@pone.0034867-Quince2]. Sequences were removed from the analysis if they were shorter than 200 base pairs (bp) in length, did not contain the barcode or primer sequence, had any ambiguous nucleotides, produced less than 100 reads per sample, or had a machine-based quality score below 25. Sequences were clustered into OTUs at a minimum identity of 97% and representative sequences were aligned using PyNAST against the greengenes core set from May 2009 [@pone.0034867-DeSantis1]. Phylogenetic assignments were made using the naïve Bayesian classifier in the Ribosomal Dataset Project [@pone.0034867-Wang2]. Fast UniFrac [@pone.0034867-Hamady1] was utilized to produce principal coordinate analyses (PCoA) for comparing the phylogenetic distances between pairs of the 16 samples. Phylotype assignment was made using the RDP classifier and greengenes core set as described above. The resulting PCoA axes were exported and used to produce a graph summarizing the analysis. For PCoA analysis that used previously published datasets including human microflora [@pone.0034867-Costello1], outdoor air [@pone.0034867-Bowers1], and house dust [@pone.0034867-Taubel1], [@pone.0034867-Rintala1] data, the sequences were assigned to phylotypes by BLASTing against the greengenes database to identify their closest matching sequences [@pone.0034867-Hamady1]. To accomplish this, the greengenes database was formatted into the BLAST database using formatdb and the resulting tree was used to assess the phylogenetic relationships between all examined samples. Each sequence was assigned to its closest BLAST hit in the formatted greengenes database and clustered into phylotypes at 97% sequence similarity for input into Fast UniFrac. Tag information and the unprocessed DNA sequences obtained in this study have been deposited in the MG-RAST archive under accession numbers 40389, 40390, 403991. Statistical Analysis {#s4e} -------------------- Comparisons between two sets of data were made using an unpaired homoscedastic t-test. Comparisons between more than two values were conducted post-ANOVA analysis via Tukey\'s test using the Matlab statistical toolbox on Matlab (software version R2010b). Unifrac P-test analysis was performed with the FastUnifrac user interface on the University of Boulder Colorado web interface on <http://bmf2.colorado.edu/fastunifrac/index.psp>. Supporting Information {#s5} ====================== ###### **HVAC filtration efficiency.** Filtration efficiency was estimated at the HVAC filter---through which indoor return air and outdoor air passes---by placing optical particle counters (size ranges 0.3--0.5 µm, 0.5--1 µm, 1--2.5 µm, 2.5--5 µm, 5--10 µm and \>10 µm) before and after the filter. Submicron size particles are inefficiently removed whereas particles bigger than 2.5 µm are removed at 75--90%. The inset is a graphical representation of the air handling unit setup. Dampers were temperature controlled and regulated the relative flow of outdoor air and indoor return air. (TIF) ###### Click here for additional data file. ###### **Rarefaction curves for samples of indoor air, ventilation duct supply air, HVAC filter dust, and floor dust.** Curves are based on samples that contained more than 300 sequences to avoid diversity estimate biases. The inset shows the same plot for one floor dust and one indoor air sample, each containing more than 1350 sequences. Error bars represent one standard error using observed species values for independent samples. Chao1 diversity indexes were calculated to be 3720, 1259, 2988, and 637 for floor dust, HVAC filter dust, indoor air, and ventilation duct supply air, respectively. (TIF) ###### Click here for additional data file. ###### **Abundance of dominant (A) and rare (B) bacterial phyla from indoor air (Indoor10), ventilation duct supply air (Duct10), HVAC filter dust (HVAC10), and floor dust (Floor10/37).** The dominant phyla represent 93%--98.5% of the sequences recovered. The Cyanobacteria are dominated by chloroplast sequences from plant (*Streptophyta*) material. The number after the samples indicates whether it is a sieved (37, PM~37~) or respirable size fraction sample (10, PM~10~). (TIF) ###### Click here for additional data file. ###### **Sequencing summary table describing the sample type and listing corresponding sample multiplex identifiers (MIDs).** MIDs were contained on the forward primer. The key adaptor on both primers was TCAG. (DOC) ###### Click here for additional data file. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The research was funded by the Alfred P. Sloan Foundation "Microbiology for the Built Environment Program". The funders had no role in study design, data collection, and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: DH WWN JP. Performed the experiments: DH JQ NY HR-Y. Analyzed the data: DH WWN KB JP. Contributed reagents/materials/analysis tools: KB. Wrote the paper: DH WWN JP. [^2]: Current address: Department of Civil and Environmental Engineering, Clarkson University, Potsdam, New York, United States of America [^3]: Current address: Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States of America

INTRODUCTION ============ Multiple sclerosis (MS) is the most-common immune-mediated demyelinating disease of the central nervous system, with a wide range and heterogeneous clinical presentations with no definitive symptoms, but with certain highly characteristic symptoms and signs such as unilateral subacute visual loss, bilateral internuclear ophthalmoplegia, subcute spinal-cord syndromes, fatigue, and heat sensitivity. Common symptoms and signs of MS in decreasing order of presentation frequency are sensory symptoms in the limbs or face, unilateral visual loss, subacute motor weakness, diplopia, balance problems with gait disturbances, vertigo, limb ataxia, bladder symptoms, and pain.[@B1] The main MS phenotypes are relapsing disease and progressive disease, which are further classified into several clinical subtypes depending on the disease activity (with relapses or new MRI lesions forming over time).[@B2] Typically 85--90% of patients have relapsing-remitting multiple sclerosis (RRMS), with clinically isolated syndrome (CIS) presenting as the first manifestations in the form of optic neuritis, pyramidal-tract symptoms, brainstem, and spinal-cord symptoms.[@B3][@B4] Typically 5--10% of patients have primary progressive multiple sclerosis (PPMS), with the most-common presentation of a spinal-cord syndrome without a sensory level.[@B5] Secondary progressive multiple sclerosis (SPMS) is seen at 12--19 years after the RRMS onset, when there is gradual worsening of disability with or without occasional relapses and minor remissions.[@B6][@B7][@B8] Recent advances in neuroimaging and biomarker research have resulted in SPMS being further defined for better understanding of the exact phenotype and its course. The epidemiology of MS has been studied extensively and reported mainly for Western countries due to high prevalence of MS in Caucasians.[@B9][@B10] The variation in prevalence is interrelated with geographical location and genetic factors. While the prevalence of MS is higher in Northern Europe, America, and South Australia than in Asia and Africa, there have been many reports of MS increasing in parts of the world with a low prevalence.[@B11][@B12][@B13] The purpose of the present study was to determine the prevalence and clinical presentation of MS patients in the Saudi population in order to obtain a better understanding among physicians of the disease activity, prognosis, and decision-making. This will open a different horizon for future research and management decisions in these patients. METHODS ======= This retrospective cross-sectional observational study was carried out in the Department of Neurology, King Fahd Hospital of University Alkhobar in the Kingdom of Saudi Arabia (KSA) from April to October 2017 after Institutional Review Board approval (IRB No. 2017-01-027) had been received from the university. Data were obtained for 224 male and female patients aged \>14 years who presented to the hospital with a suspected demyelinating disorder and who were investigated with brain and spine MRI with gadolinium enhancement from January to December 2016. Some of the initially enrolled patients out to have MS-mimicking conditions such as neuromyelitis optica, acute disseminated encephalomyelitis, and vasculitis, and they were excluded from study. Only 190 patients who fulfilled the McDonald criteria for MS who had regular findings in subsequent follow-up MRI and patients with CIS were included in this study. The different phenotypic patterns of disease were defined as follows: CIS --- The presence of a first clinical attack with objective evidence of neurological changes and a lesion compatible with MS in MRI studies. RRMS ---- A history of at least two clinical attacks or at least one clinical attack with new documented MRI lesions over time. Relapse was defined as an acute deterioration of neurological function lasting for at least 24 hours followed by a period of total or partial recovery. SPMS ---- Evolving in patients previously with a history of clinical attack (RRMS) who demonstrated gradual and progressive disability with or without periods of relapse. PPMS ---- Gradual and progressive worsening of neurological function from the beginning of the disease with or without subsequent relapses observed at least 1 year. The collected data were entered into and analyzed on the Statistical Package for Social Sciences (version 20.0, IBM Corp., Armonk, NY, USA). The results were computed as frequency and percentage values for age, sex, race, age at onset, disease duration, first clinical presentation, other established clinical symptoms, disease phenotype, Expanded Disability Status Scale (EDSS) score, and sites of MRI lesions, gadolinium-enhanced lesions, and treatment group. Mean±SD values for age, age at disease onset, and duration of disease were calculated. Subgroup analyses were performed for each disease phenotype. The chi-square test was used to compare categorical variables and a probability value of \<0.05 was taken as indicative of statistical significance. RESULTS ======= The demographic and clinical characteristics of the 190 patients with MS included in this study are presented in [Table 1](#T1){ref-type="table"} and [2](#T2){ref-type="table"}. About two-thirds of the participants were female (*n*=160, 62.6%), giving a male-to-female ratio of 1:1.6. The cases were aged 32.6±8.8 years, and their age at disease onset was 26.27±8.2 years, while the disease duration was 6.38±5.10 years. Most (*n*=179, 94.2%) of the cases were of Saudi origin. The 190 cases comprised 144 (75.8%) of RRMS, 23 (12.1%) of SPMS, 18 (9.5%) of CIS, and 5 (2.6%) of PPMS. Age, sex, disease duration, and the number of MRI gadolinium-enhanced lesions in each phenotype are presented in [Table 3](#T3){ref-type="table"}. Optic neuritis and myelitis were the most-frequent first clinical presentations of the disease, comprising 53 (27.9%) of cases, while brainstem and pyramidal-tract manifestations comprised 42 (22.1%) of the cases of first clinical attack. Other subsequent clinical manifestations are presented in [Fig. 1](#F1){ref-type="fig"}. The frequency and distribution of clinical symptoms in each phenotype are further presented in [Table 4](#T4){ref-type="table"}. The frequencies of T2-weighted MRI lesions in the cases are presented in [Fig. 2](#F2){ref-type="fig"}. In more than two-thirds of the cases (*n*=156, 82.1%) the EDSS scores were within the range of 1.0--5.5, while in 27 (14.2%) cases the EDSS scores were within the range of 6.0--6.5, and only 7 (3.7%) cases had EDSS scores within the range of 7.0--7.5. No case had an EDSS score of 7.5 in this study. More than half of the 128 (67.4%) patients were taking interferon β-1a (IFNβ-1a), and 18 (9.5%) were taking IFNβ-1b while only 12 (6.3%) were receiving a different oral disease-modifying therapy (DMT), 4 (2.1%) were taking natalizumab and 32 (16.8%) were not being treated. DISCUSSION ========== The prevalence of MS is changing rapidly in different parts of the world.[@B14][@B15][@B16] There are many factors underlying these changes, such as lifestyle modifications, changes in certain environmental factors, or simply an increased awareness and better diagnosis of the disease.[@B17][@B18][@B19][@B20][@B21] Similar studies from the Middle East and other parts of the Arab world have shown variable prevalence rates of MS.[@B22] The prevalence and pattern of clinical presentation have been studied previously in the KSA.[@B23][@B24] The demographic data obtained this study are similar to those in other local studies, with a few notable differences. This study found that MS is more common in females than males, with a male-to-female ratio of 1:1.6. Daif et al.[@B24] reported similar observations for a hospital-based study in KSA,[@B24] while Yaqub reported that this ratio was 4.3 in another study from KSA.[@B23] As with other autoimmune disorders, there is a slight female dominance worldwide, with a small amount of variability in the ratio in different parts of world, such as 3.0 in East Asia and 2.6 in the USA. There was a slightly higher number of male patients with MS in KSA that resulted in this ratio being different from that in the Yaqub study.[@B23] The age at onset in our study was 26.27±8.2 years, which is almost same as that of 27.7±7.8 years in the study of Daif et al.[@B24] The overall mean age at onset in the Middle East is 28.54 years, which is similar to that in Western countries and other parts of the world. The youngest age at presentation in our study was 9 years, for which the diagnosis was of pediatric MS, and this patient later presented in an adult neurology clinic at 14 years of age. The oldest age at presentation was 58 years. The youngest age at presentation in the literature is before 10 years, and MS is rarely seen in the eighth decade of life.[@B25][@B26] The duration of disease in the present study was 6.38±5.11 years, ranging between 1 and 20 years, whereas other studies from the Middle East have found disease durations of 3.9±9.3 and 5.5±4.7 years.[@B13][@B27] In this study optic neuritis and myelitis were the most-frequent first clinical presentations of disease, while brainstem and pyramidal-tract findings as well as ataxia, vertigo, and hemisensory loss were less-frequent first clinical presentations. These findings are consistent with other studies from the region.[@B20][@B28][@B29] Optic-nerve and spinal-cord involvement, with predominant visual impairment in the beginning of the course, is more commonly reported in Asian countries. Among other symptoms, fatigue was reported in one-third of our cases. A study from Pakistan found fatigue in 25% of cases.[@B30] Dysarthria, psychiatric, and memory disturbances were also seen in up to 10% of the cases in our study. Other cranial-nerve symptoms and signs including facial weakness, hearing loss, and trigeminal neuralgia were seen in up to 6% of cases. Seizures were seen only in two patients (1%). The clinical phenotype was RRMS in 75% of our cases and 60.7% in the study of Daif et al.[@B24] A higher frequency is probably attributable to early diagnoses resulting from the better availability of services and imaging facilities along with modifications and revisions of diagnostic criteria for MS. An earlier diagnosis can lead to treatment starting earlier and reducing the probability of RRMS converting to SPMS over time. The frequency of SPMS was 11.1% in our study, and only 3.7% of the cases had PPMS; in contrast Daif et al.[@B24] observed no cases of SPMS, while 19.1% of their cases were PPMS and 20% were progressive-relapsing multiple sclerosis (PRMS). These findings need to be reconsidered since PRMS has been eliminated and categorized as PPMS with activity and CIS is now considered part of the spectrum of MS after the 2013 revisions by the international advisory committee on clinical trials of MS.[@B8] The clinical phenotype findings in our study were very similar to those found in recent studies from the United Arab Emirates and Kuwait.[@B31][@B32] Supratentorial T2-weighted white-matter lesion lesions and deep-gray-matter or juxtacortical lesions were the most-frequent MRI lesions, comprising 28% and 23.7% of all MRI lesions observed in 93.6% and 79.4% of the cases, respectively. Optic-nerve and spinal-cord lesions were seen in 12.1% and 45.7% of the cases, respectively; despite their high presentation frequency clinically, cord lesions are small and peripheral and so they can be missed in conventional MRI. They can be detected more easily in short-time inversion recovery MRI sequences, and these are not obtained in all cases. Cerebellar and brainstem involvement were seen in 9--10% of the MRI lesions, which is similar to the rate seen in Western countries.[@B33][@B34] Gadolinium-enhanced lesions comprised 11.62% of all lesions, and it was observed that their frequency was higher (57.1%) in SPMS cases than in RRMS (38.9%) and CIS (22.1%) cases. Two (28.6%) of the seven PPMS cases had gadolinium-enhanced lesions in this study, suggesting that a continuous inflammatory process was ongoing in cases of SPMS compared to other subtypes and that PPMS tends to be a more-progressive disease process that often presents with less-prominent gadolinium enhancement.[@B35][@B36] About two-thirds of our patients (82.1%) had EDSS scores within the range of 1.0--5.5, reflecting functional mobility without requiring any walking aid. This indicates either a relatively benign course of disease or treatment-responsive disease in this population. However, studies from different parts of the world have also found mean EDSS scores of around 2 and 3. A Brazil study found that the EDSS score was 3.50±1.9,[@B37] while a study of newly diagnosed MS from Qatar found a mean EDSS score of 2.4 (median=2), reflecting a lower level of clinical progression despite a higher radiological burden,[@B38] and a study of 200 MS cases from Iran found that the EDSS score was 2.1±1.9.[@B13] None of our patients had an EDSS score higher than 7.5 because patients with higher disability scores are bedridden, receives supportive treatment only. Regarding the distribution of DMTs in this case series, 66.8% of the 127 patients were taking IFNβ-1a, including both subcutaneous and intramuscular on alternate days, and once-weekly regimens depending upon the physician decision and patient acceptance, tolerability, and drug availability. There were 18 (9.5%) patients taking IFNβ-1b, while only 12 (6.3%) of cases were receiving a different oral DMT at other centers, including teriflunomide (*n*=4), fingolimod (*n*=2) dimethylefumarate (*n*=2) and ocrelizumab (*n*=1) while 4 cases were started on natalizumab therapy due to treatment failure with IFNβ, with a more-severe and aggressive course of disease. Thirty-two (16.8%) patients were not receiving any treatment due to an advanced disease stage or patient refusal. The present study had some limitations. Firstly, since this was a hospital-based retrospective cross-sectional study, the exact number of relapses and transitions of the disease spectrum from one clinical subtype to another could not be assessed. Secondly, we did not address other paraclinical parameters such as cerebrospinal fluid (CSF) analysis, oligoclonal bands, and evoked potentials, since these parameters have only appeared in new MS criteria.[@B39] In particular, the presence of oligoclonal bands in the CSF with the first symptom is proposed as an alternative for waiting for another symptom or lesion for the early diagnosis of MS. Lastly, we could not determine the exact duration and any change in the treatment regimen according to a transition of disease spectrum due the cross-sectional nature of our study. In conclusion, this study has shown that MS presenting in the hospital setting is more common in KSA then reported previously, and that the number of diagnosed cases is increasing. It is therefore an emerging and disabling neurological illness in KSA with clinical characteristics not dissimilar to those in other Middle Eastern countries. A decrease in the frequency of SPMS patients indicates that more new cases of RRMS are being diagnosed or that adequate treatments of RRMS are preventing the evolution to SPMS. Further larger and population-wide epidemiological and clinical studies with the long-term follow-up of MS patients are required to better assess the clinical spectrum of MS in KSA. **Conflicts of Interest:** The authors have no financial conflicts of interest. {#F1} {#F2} ###### Demographic characteristics of cases with multiple sclerosis (*n*=190)  -------------------------- ------------------- Gender, n (%)  Male 72 (37.9)  Female 118 (62.1) Ethnicity, n (%)  Saudi 179 (94.2)  Non-Saudi 11 (5.8) Age (years) 32.6±8.8 (15--58) Age at onset (years) 26.2±8.2 (9--54) Disease duration (years) 6.38±5.10 (1--20) -------------------------- ------------------- Data are *n* (%) or mean±SD (range) values. ###### Clinical characteristics of cases with multiple sclerosis (*n*=190)  Characteristic *n* (%) ------------------------------------------ ------------ First clinical presentation/relapse  Optic neuritis 53 (27.9)  Myelitis 53 (27.9)  Brainstem 42 (22.1)  Pyramidal tract 42 (22.1) Disability score (EDSS)  1.0--5.5 156 (82.1)  6.0--6.5 27 (14.2)  7.0--7.5 7 (3.7)  8.0--8.5 0 (0.0) Distribution of lesions in MRI  One system/site 18 (9.5)  Two systems/sites 47 (24.7)  Three systems/sites 64 (33.7)  Four systems/sites 44 (23.2)  Five systems/sites 17 (8.9) Topography of T2-weighted lesions in MRI  Optic nerve 23 (3.6)  Supra tentorial cerebral white matter 178 (27.9)  Deep gray matter 151 (23.7)  Cerebellum 58 (9.1)  Brainstem 66 (10.4)  Spinal cord 87 (13.7) DMT  Patients not on any DMT treatment 32 (16.8)  IFNβ-1a 128 (67.4)  IFNβ-1b 18 (9.5)  Other/oral 12 (6.3) Data are *n* (%) values. DMT: disease-modifying therapy, EDSS: Expanded Disability Status Scale, IFNβ: interferon β. ###### Age, gender, age at onset, disease duration, and gadolinium-enhanced lesions in each multiple sclerosis phenotype (*n*=190)  Clinical type Total (%) Females (%) Males (%) Age (years) Onset age (years) Disease duration (years) Patients with gadolinium-enhanced lesion (%) ------------------------------------------ ------------ ------------- ----------- ------------- ------------------- -------------------------- ---------------------------------------------- Clinically isolated syndrome 18 (9.5) 10 (55.6) 8 (44.4) 32.7±10.5 28.9±9.0 3.72±4.30 4 (22.1) Relapsing-remitting multiple sclerosis 114 (75.8) 91 (63.2) 53 (36.8) 32.0±8.3 25.9±7.8 5.94±4.60 56 (38.9) Secondary progressive multiple sclerosis 21 (11.1) 15 (71.4) 6 (28.6) 35.6±9.9 24.8±9.7 10.8±5.2 12 (57.1) Primary progressive multiple sclerosis 7 (3.7) 2 (28.6) 5 (71.4) 39.7±7.4 31.0±7.9 8.7±7.1 2 (28.6) Data are *n* (%) or mean±SD values. ###### Frequency and distribution of clinical symptoms in each multiple sclerosis phenotype  Clinical symptom Clinical subtype (%) ---------------------------- --------------------------------------- --------------------------------------------------- ---------------------------------------------------- ------------------------------------------------ Clinically isolated syndrome 18 (9.5) Relapsing-remitting multiple sclerosis 114 (75.8) Secondary progressive multiple sclerosis 21 (11.1) Primary progressive multiple sclerosis 7 (3.7) Motor 6 (33.3) 85 (59.3) 21 (91.3) 4 (66.7) Sensory 9 (50) 104 (72.2) 22 (95.7) 6 (100) Sphincter disturbances 2 (11.1) 30 (20.8) 11 (47.8) 2 (33.3) Ataxia 2 (11.1) 52 (36.1) 16 (69.6) 3 (50) Dysarthria 2 (11.1) 12 (8.3) 4 (17.4) 1 (16.7) Vertigo 1 (5.6) 37 (25.7) 5 (21.7) 1 (16.7) Visuo-ocular disturbances 7 (38.9) 89 (61.8) 14 (60.9) 2 (33.3) Trigeminal neuralgia \- 3 (2.1) \- \- Facial weakness 1 (5.6) 8 (5.6) 4 (17.4) \- Dysphagia 1 (5.6) 3 (2.1) 1 (4.3) Hearing loss 3 (2.1) 2 (8.7) \- Memory disturbances \- 6 (4.2) 5 (21.7) 1 (16.7) Psychiatric manifestations 10 (6.6) 6 (26.1) 1 (16.7) Fatigue \- 45 (31.3) 14 (60.9) 1 (16.7) Cramps \- 12 (8.3) 5 (21.7) 1 (16.7) Seizures 1 (5.6) 1 (0.7) \- \- Data are *n* (%) values.

Erratum {#Sec1} ======= After the publication of the article \[[@CR1]\], the authors realised it contained an error with the Methods section. The following paragraph should be deleted from the original article: Cell proliferation assay {#Sec2} ------------------------ Proliferation was measured using Delfia cell proliferation kit (Perkin-Elmer, Waltham, MA, USA) according to manufacturer protocol. BrdU incorporation was measured by time--resolved fluorescence 48 h after transfection using VictorX4 multimode plate reader (Perkin-Elmer, Waltham, MA, USA). All experiments were performed in triplicates and repeated tree times. The online version of the original article can be found under doi:10.1186/s12885-016-2867-z.

1. Introduction {#sec1} =============== Diabetes mellitus represents an expanding pandemic that contributes markedly to worldwide morbidity and mortality. The world prevalence of diabetes among adults (aged 20--79 years) will be 6.4%, affecting 285 million adults, in 2010 and will increase to 7.7% and 439 million adults by 2030 \[[@B1]\]. There is a strong relationship between obesity and type 2 diabetes mellitus (T2DM) \[[@B2]\]. In a large USA population study, the prevalence of diabetes increases with increasing weight classes according to body mass index (BMI). Approximately half of those diagnosed with T2DM are obese \[[@B3]\]. Weight control is the key to successful T2DM management. Weight loss achieved by lifestyle interventions has been shown to be effective in preventing and treating T2DM \[[@B4]--[@B7]\]. However, conventional treatment, such as, lifestyle modification and pharmacotherapy has produced small improvements in weight \[[@B7]--[@B10]\]. By contrast, bariatric surgery has been shown to effectively provide durable weight loss \[[@B11]\]. Currently, bariatric surgery is now considered appropriate for T2DM patient with BMI ≧ 35 kg/m^2^. Bariatric surgery leads to remission of T2DM in the majority of patients and improvement in the rest \[[@B12]\]. The American Society of Bariatric and Metabolic surgery estimates that the number of bariatric procedures increased from about 16,000 in the early 1990s to more than 103,000 in 2003 and 220,000 people in the United States had bariatric surgery in 2008. Growing evidence from clinical and animal studies indicates that bariatric/metabolic surgery may be beneficial for T2DM in nonseverely obese or even nonobese patients (BMI \< 35 kg/m^2^) \[[@B13], [@B14]\]. Recently, International Diabetes Federation has released its position statement \[[@B15]\]: surgery should be an accepted option in people who have T2DM and BMI of 35 more. Surgery should be considered as an alternative treatment option in persons with BMI 30 to 35 when diabetes cannot be adequately controlled by optimal medical regimen, especially in the presence of other major cardiovascular disease risk factors. The surgical approach is now being extended to overweight and mild to moderate obese (BMI \< 35 kg/m^2^) patients with T2DM. Herein, to evaluate current evidence of metabolic surgery for treatment of T2DM in patients with a BMI \< 35 kg/m^2^, we conducted a review to date of available clinical studies. 2. Materials and Methods {#sec2} ======================== 2.1. Search Strategy and Study Inclusion Criteria {#sec2.1} ------------------------------------------------- We conducted a comprehensive review of all studies published containing data on weight loss and T2DM-related outcomes of patients treated with any form of bariatric/metabolic surgery where the mean study BMI was \<35 kg/m^2^. Only reports published in English were included for review. Studies whose inclusion criteria indicated bariatric or metabolic surgery for low BMI patients were excluded if their mean BMI was ≧35 kg/m^2^, all participants were not T2DM patients, diabetic participants had gastric surgery with anatomical similarities to RYGB because of gastric cancer and ulcer, or they did not report diabetes-related outcomes such as fasting plasma glucose (FPG), glycated hemoglobin (HbA1c), and postoperative clinical status. Pubmed was searched from January 1, 1980, to November 1, 2011, for citations using the following keywords: "metabolic surgery", "bariatric surgery", "diabetes surgery", "T2DM", "type 2 diabetes", "obesity", "BMI \< 35 kg/m^2^", "mild to moderate obesity", and "low-BMI". To supplement the electronic search, manual reference checks were performed in the identified studies. 2.2. Data Reporting {#sec2.2} ------------------- Study authors, country, year of publication, surgical procedure, and study design were summarized. Characteristics of the study groups, BMI, FPG, and HbA1c were recorded. Diabetes-related clinical outcomes were collected as % meds resolved (the percentage of the patients who discontinued antidiabetic medication postoperatively) and remission rate (the percentage of the patients who had remission or resolution of T2DM according to the varied definition in the studies included). These two parameters were calculated using available values if not specifically stated. Then, the studies were divided into 2 groups according to mean duration of T2DM prior to surgery. The percentage of insulin using patients prior to surgery and T2DM remission rate was compared. Regarding safety of metabolic surgery, all major and minor complications were counted because they were variably reported. Mortality was also checked. 2.3. Statistical Analysis {#sec2.3} ------------------------- Weighted means for ages and duration of T2DM were calculated. Pre- and Postoperative mean values (and corresponding 95% confidence intervals) for BMI, FPG, and HbA1c were summarized using a random effects model to account for the variability between the different studies. Only studies which provided both a pre- and post-measurement were included in the mean estimation for each of the parameters considered. Comparisons of 2 groups with a short (\<8 years) and long history (\>8 years) of diabetes with respect to insulin using patients and diabetes remission were performed by the chi-square test. The statistics was performed using the software package PASW, version 18.0, for Windows (SAS Institute, Cary, NC, USA). 3. Results {#sec3} ========== 3.1. Study Characteristics {#sec3.1} -------------------------- A total of 18 studies met inclusion criteria as identified by literature search and reference checks \[[@B16]--[@B33]\] Of the 18 studies, 17 (94%) were performed outside of the United States, in Brazil (7, 39%), Italy (4, 22%), Taiwan (4, 22%), Chile (1, 6%), and India (1, 6%). 13 studies (72%) have been published in the last 3 years from 2009 to 2011. The studies were performed prospectively (16, 89%) or retrospectively (2, 11%). Surgical procedures performed in this paper included Roux-en-Y gastric bypass (RYGB) in 6 (33%) studies, duodenal-jejunal bypass (DJB) in 4 (22%), biliopancreatic diversion (BPD) in 3 (17%), minigastric bypass (MGB) in 2 (11%), ileal interposition with sleeve or diverted sleeve gastrectomy (II-SG or II-DSG) in 2 (11%), sleeve gastrectomy (SG) in 1 (6%), and stomach- and pylorus-preserving BPD (BPD-SPP) in 1 (6%). The data are summarized in [Table 1](#tab1){ref-type="table"}. 3.2. Patient Characteristics {#sec3.2} ---------------------------- Of the 18 studies, total 477 patients underwent metabolic/bariatric surgery. 16 studies reported the patient gender, and 53% of the total study population was female. The mean age of the patients ranged from 34 to 56 and its weighted mean was 47. The followup period ranged from 6 months to 18 years, and its weighted mean was 22 months. 2 studies reported the results of a longer than 5-year followup. The duration of diabetes prior to surgery ranged from 6 months to 28 years, and its weighted mean was 8.2 years. Majority of patients were taking oral antidiabetic medication and/or insulin. The percentage of Insulin using patients was 30.1% in 16 studies. The data are summarized in [Table 1](#tab1){ref-type="table"}. 3.3. BMI {#sec3.3} -------- 14 studies were included in the mean estimation. The BMI decreased from 30.4 (95% CI 28.4--32.3) to 24.8 (95% CI 24.1--25.5) kg/m^2^. There were only two studies reporting that one of the total 15 patients was in the mildly undernourished range (BMI 17--18.5 kg/m^2^) after RYGB without any evidence of malnutrition \[[@B21]\] and 12 patients (17.4%) after II-DSG were underweight (BMI \< 20 kg/m^2^) without lowering serum albumin value \[[@B24]\]. Overall, the risk of excessive weight loss after metabolic surgery was 2.7% (13 patients). The data are summarized in Tables 2 and [3](#tab3){ref-type="table"}. 3.4. T2DM Outcomes {#sec3.4} ------------------ 12 studies were included in the mean estimation of FPG. It decreased from 203.5 (95% CI 187.4--219.6) to 112.5 (95% CI 103.9--121.1) mg/dL. 10 studies were included in the mean estimation of HbA1c. It decreased from 9.0 (95% CI 8.6--9.5) to 6.3 (95% CI 6.1--6.6) %. Regarding clinical outcomes of diabetes, 86.8% of the patients stopped taking antidiabetic medication after surgery (% meds off). The definition of resolution or remission of T2DM varied. When it is defined as FPG \< 126 mg/dL and/or HbA1c \< 6.5% without the use of antidiabetic medication at the time of evaluation, 64.7% of the patients met the criteria. The data are summarized in Tables [2](#tab2){ref-type="table"} and [3](#tab3){ref-type="table"}. 3.5. Clinical Outcomes of Diabetes According to Duration of T2DM prior to Surgery {#sec3.5} --------------------------------------------------------------------------------- When the studies were stratified by mean duration of T2DM (5 studies, ≦8 years, and 7 studies, \>8 years), the percentage of insulin-using patients prior to surgery was 18.2% and 45.9% (*P* \< 0.01). Remission of T2DM was achieved in 66.0% of the patients with a short history (≦8 years) of T2DM and 52.9% of those with a long history (\>8 years) of T2DM (*P* = 0.03). The data are shown in [Figure 1](#fig1){ref-type="fig"}. 3.6. Complications and Mortality {#sec3.6} -------------------------------- Overall, complication rate was 10.3% (range 4.5--33.3%) in 16 studies. The types of complication varied and were dependent on follow-up period and surgical procedures. Therefore, we included all major and minor complications. The mortality rate was 0% in 17 studies. The data are summarized in [Table 2](#tab2){ref-type="table"}. 4. Discussion {#sec4} ============= The concept of metabolic surgery was defined by Buchwald and Varco in 1978 in their book "*Metabolic Surgery* as the operative manipulation of a normal organ or organ system to achieve a biological result for a potential health gain" \[[@B34]\]. Now, metabolic surgery is defined as any modification of the gastrointestinal (GI) tract, where rerouting the food passage seems to improve T2DM, based on mechanisms that are weight loss independent. This new frontier of bariatric/metabolic surgery includes the application of conventional bariatric procedures (RYGB, BPD, SG, MGB) and the introduction of new procedures (DJB, II-SG, II-DSG, BPD-SPP) designed with the specific aim of having metabolic effects irrespective of causing massive weight loss. There is strong evidence that bariatric surgery for severely obese patients (BMI ≧ 35 kg/m^2^) provides exceptional sustained weight loss and 50--85% remission of T2DM \[[@B35]\]. In view of growing enthusiasm for surgical interventions to treat T2DM, the 1st diabetes surgery summit was held in Rome in March 2007 to develop guidelines for the use of GI surgery to treat T2DM. The recommendations were made by a multidisciplinary group of 50 voting delegates \[[@B36]\]. Accordingly, the "Standards of Medical Care in Diabetes" published yearly by the American Diabetes Association, for the first time, mentions surgical therapy in 2009 \[[@B37]\]. Recently, International Diabetes Federation has released its position statement \[[@B20]\]. These statements have mentioned that bariatric surgery for T2DM patients with a BMI ≧ 35 kg/m^2^ is considered an accepted option as with standard medical therapy and metabolic surgery might, moreover, be considered a reasonable therapeutic alternative for low BMI (\<35 kg/m^2^) patients with T2DM who do not respond to standard medical therapy. The aim of this paper was to explore the current evidence with a view to evaluate the potential of metabolic surgery for T2DM with a BMI \< 35 kg/m^2^. Metabolic surgery to treat T2DM in patients with low BMI provided desirable results regarding weight loss. The estimated mean BMI categorized as class I obesity prior to surgery reached normal weight range after surgery. Importantly, only 13 patients (2.7%) following RYGB or II-DSG in 2 studies reported excessive weight loss, and they did not show any evidence of malnutrition \[[@B21], [@B24]\]. Even the procedures that typically produce the greatest reduction in BMI and excess weight in morbidly obese patients did not affect a similarly dramatic BMI reduction in the low-BMI patients \[[@B38]\]. Scopinaro and so forth reported that BPD does not entail risk of excessive or undue weight loss because there is a maximum energy absorption capacity after the operation, which corresponds to a weight of stabilization of low BMI patients \[[@B39]\]. The similar homeostatic mechanism may explain weight stabilization without causing undesirable weight loss after surgical procedures including intestinal bypass. In this paper, diabetic status was significantly improved after metabolic surgery in the majority of studies. Discontinuation of antidiabetic medication and remission of T2DM after metabolic surgery were achieved in 86.8% and 64.7% of the patients with FPG and HbA1c approaching slightly above normal range. Moreover, metabolic surgery provided adequate glycemic control for 30.1% of the patients using insulin prior to surgery. It has been described that malabsorptive bariatric procedures have higher diabetes remission rates than restrictive ones \[[@B12], [@B40]\]. T2DM typically resolves within a few days to weeks following malabsorptive procedures such as RYGB and BPD before significant weight loss is achieved. Although the exact mechanism is not yet fully understood, growing evidence shows that malabsorptive procedures involving rerouting of food might improve T2DM by enhancing insulin sensitivity and/or by improving *β*-cell function that is additive to weight loss and reduced caloric intake \[[@B16], [@B41], [@B42]\]. The recent studies have described that acute insulin response to intravenous glucose and early phase insulin response to oral glucose load improved significantly within a month following GI bypass surgery \[[@B16], [@B20]\]. The mechanism for these changes could be due to a dramatic decrease of insulin resistance and an increase in postprandial plasma levels of glucagon-like peptide-1 (GLP-1) early after surgery. Currently, two hypotheses (hindgut and foregut theory) have been proposed to explain T2DM remission after metabolic surgery in addition to decreased calorie intake after surgery and surgical-induced weight loss which might contribute to improving insulin sensitivity. The former states that surgical rerouting of nutrients to the distal part of the small intestine results in increased secretion and concomitant glucose-lowering effects of GLP-1, and the latter emphasizes that surgical bypass of the foregut prevents the release of a hitherto unidentified nutrient-induced diabetogenic signal in susceptible individuals \[[@B43]\]. The novel surgical procedures such as DJB, BPD-SPP, II-SG, and II-DG were designed to apply hindgut or/and foregut hypotheses without massive weight loss and achieved 56% of T2DM remission and 84% of diabetes meds off in this paper. The weight loss effect of metabolic surgery on T2DM in low BMI patients might be lower than that of bariatric surgery on T2DM in high BMI patients. Understanding and enhancing the abovementioned mechanism are the key to success in metabolic surgery. There is no strong evidence describing the durability of metabolic surgery in long-term followup. In this paper, 2 studies showed durable diabetes remission of T2DM during 5--18 years period after MGB and BPD \[[@B29], [@B30]\]. By contrast, the recent studies of bariatric surgery for T2DM patients with severe obesity showed that 24%--43% of the patients with initial remission or improvement of their T2DM subsequently developed T2DM recurrence or worsening during the mid- to long-term followup period \[[@B44], [@B45]\]. A low preoperative BMI and severe T2DM status were associated with failure of consistent durable remission of diabetes. The common causes for failure of diabetes remission after bariatric surgery are known as inadequate weight loss or regain of weight, longstanding poorly controlled or aggressive T2DM, lower preoperative BMI, and latent autoimmune diabetes in adults (LADA) \[[@B46]\]. LADA comprises 10% of diabetic age 30--55 and is more prevalent in low BMI individuals \[[@B47], [@B48]\]. Most clinical guidelines and statements have followed the BMI-based criteria established by 1991 National Institutes of Health Consensus Conference Guidelines \[[@B49]\]. Although BMI is convenient to classify the grade of obesity, it does not seem to be appropriate in selecting the suitable T2DM candidates for metabolic surgery. For instance, the risk of diabetes and/or metabolic syndrome is determined by ethnicity, waist circumference, fat distribution, body composition, and intrahepatic fat \[[@B50], [@B51]\]. South Asian and Chinese individuals have distributions of elevated glucose and lipid levels similar to Europeans at significantly lower BMI values \[[@B52]\]. The natural history of type 2 diabetes is also important to consider in determining the timing of intervention. As the diabetes state progresses, there is continued beta-cell deterioration together with a decline in insulin secretion within 6--10 years of T2DM diagnosis \[[@B53], [@B54]\]. Schauer et al. showed that a shorter history of diabetes and milder disease according to preoperative medication status were associated with an increased likelihood of remission after RYGB \[[@B55]\]. Dixon and O\'Brien reported that a shorter history of diabetes and greater weight loss were positive predictive factors for remission \[[@B56]\]. This paper was consistent with this. A shorter history of diabetes with less number of insulin using patients prior to metabolic surgery resulted in greater remission rate of diabetes. Metabolic surgery should be considered early in the diabetic stage before irreparable beta-cell damage occurs. BMI alone is not an adequate measure to define the overall risk of morbidity and mortality in patients with established diabetes \[[@B50]\]. The clinical status of T2DM should be taken into account to select the suitable candidates for metabolic surgery. The goals of treatment of T2DM are not only glycemic control but also prevention of diabetes-related complications such as macro- and microvascular diseases. The target blood pressure of \<130/80 mmHg, the target cholesterol level of \<200 mg/dL, and HbA1C level \<7% should be achieved in diabetic patients. It has been reported that only 7.3% of adults with diabetes achieved all three recommended goals with conventional medical treatment \[[@B57]\]. In contrast, bariatric surgery improved hyperlipidemia in 70% or more of patients and resolved or improved hypertension in 78.5% of patients \[[@B58]\]. A systematic review to evaluate the effect of bariatric surgery on cardiovascular risk profile demonstrated that on average, hypertension, diabetes, and dyslipidemia resolved in 68%, 75%, and 71% and a 40% relative risk reduction for 10-year coronary heart disease risk was observed after bariatric surgery, as determined by the Framingham risk score \[[@B59]\]. In this paper, Shah et al. reported that RYGB for T2DM in low BMI patients reduced the predicted 10-year cardiovascular disease risk substantially for fatal and nonfatal coronary heart disease and stroke \[[@B21]\]. The mortality rates from bariatric operations (0.28--0.35%) \[[@B60]\] are compared favorably with those of other commonly performed operations, including laparoscopic cholecystectomy, whose mortality in USA ranges between 0.35 and 0.60% \[[@B61]\]. In this paper, no mortality was observed. Major and minor complication rate was also low (10.3%). Huang et al., and so forth reported that the operating time and duration of hospitalization of LRYGB for low BMI patients were lower than those for morbidly obese patients because of lower BMI \[[@B19]\]. T2DM-related additional risk should be counted, but safety of metabolic surgery for low BMI patients seems to be higher or at least similar, compared to bariatric surgery for severe obesity. Metabolic surgery for T2DM, although not the current standard care for the disease, may be coming closer to the mainstream. The ponderable statement has suggested that metabolic surgery might be considered a reasonable therapeutic alternative for low BMI (\<35 kg/m^2^) patients with T2DM who do not respond to standard medical therapy. The data from the studies included in this paper are encouraging. Although large randomized clinical trials against best medical care and assessment of the long-term efficacy and safety should be prioritized to define the role of metabolic surgery, it is clear that a high proportion of low BMI patients with T2DM will derive substantial benefit from metabolic surgery. 5. Conclusions {#sec5} ============== In this paper, including 18 studies and 477 patients, T2DM patients with a BMI \< 35 kg/m^2^ derived benefit from metabolic surgery. The weight loss effect was reasonable without any serious excessive weight loss. The antidiabetic effect was also considered excellent. Remission of T2DM and % meds off were achieved in 64.7% and 86.8% of the patients. T2DM clinical status is important to select the eligible candidates for metabolic surgery besides current BMI criteria for bariatric surgery. Metabolic surgery can be performed safely with acceptably low complication rate and mortality. Although several concerns need to be addressed, metabolic surgery for low BMI patients is coming closer to the mainstream of diabetes treatment. The authors declare that they have no relevant competing interests. The authors would like to thank K. Wolski at Cleveland Clinic, OH, USA, for statistical consultation during paper development. {#fig1} ###### Baseline characteristics of T2DM patients with BMI \< 35 kg/m^2^. Author Year Procedure *N* Female/male Mean age (range) Mean followup (range) Mean duration of diabetes Insulin user ---- --------------- ------ --------------- ------------ ------------- ------------------ ----------------------- --------------------------- ------------------ 1 Lee 2011 MGB, RYGB 62 38/24 43.1 24 months 5.4 years 23% (*n* = 14) 2 Boza 2011 RYGB 30 17/13 48.0 (28--65) 24 months (*n* = 20) 4 years 3.3% (*n* = 1) 3 de Sa 2011 RYGB 27 --- 50.3 20 (4--86) months 8.8 years 22% (*n* = 6) 4 Huang 2011 RYGB 22 20/2 47.4 (28--63) 12 months 6.6 years 18.2% (*n* = 4) 5 Scopinaro 2011 BPD 30 11/19 56.4 (43--69) 12 months 11.2 years 40% (*n* = 12) 6 Shah 2010 RYGB 15 7/8 45.6 9 months 8.7 years 80% (*n* = 12) 7 Lee 2010 SG 20 14/6 46.3 12 months \>6 months 20% (*n* = 4) 8 DePaula 2009 II-SG, II-DSG 58 (30,28) 18/40 51.4 (40--66) 19.2 (14--28) months 9.6 years 37.9% (*n* = 22) 9 DePaula 2009 II-DSG 69 22/47 51.4 (41--63) 21.7 (7--42) months 11 years 44% (*n* = 30) 10 Ramos 2009 DJB 20 9/11 43.0 (29--60) 6 months 5.3 years 0% (*n* = 0) 11 Ferzli 2009 DJB 7 --- 43.3 (33--52) 12 months 10.7 years 85.7% (*n* = 6) 12 Geloneze 2009 DJB 12 3/9 50.0 6 months 9 years 100% (*n* = 12) 13 Chiellini 2009 BPD 5 2/3 48.0 18 months 3--15 years --- 14 Lee 2008 MGB 44 38/6 39.0 1--5 years --- --- 15 Scopinaro 2007 BPD 7 2/5 49.0 (39--60) 13 (10--18) years 4.1 years 0% (*n* = 0) 16 Cohen 2007 DJB 2 0/2 47.0 (43--51) 9 months 4.5 years 100% (*n* = 2) 17 Cohen 2006 RYGB 37 30/7 34.0 (28--45) 20 (6--48) months --- 0% (*n* = 0) 18 Noya 1998 BPD-SPP 10 5/5 52.1 (40--62) 7 (2--18) months --- 40% (*n* = 4) Total --- --- 477 236/207 --- --- --- 30.1% (129/428) Weighted mean --- --- --- --- 47 22 months 8.2 years --- RYGB: Roux-en-Y gastric bypass, MGB: mini-gastric bypass, BPD: biliopancreatic diversion, SG: sleeve gastrectomy, II-SG: ileal interposition with sleeve gastrectomy, II-DSG: ileal interposition with diverted sleeve gastrectomy, DJB: duodenal jejunal bypass, BPD-SPP: stomach- and pylorus-preserving BPD. ###### Outcomes of metabolic surgery: changes in BMI, fasting plasma glucose (FPG), and glycated hemoglobin (HbA1c), clinical outcomes of diabetes (% meds resolved and remission rate), and safety of metabolic surgery (complication and mortality). BMI Fasting plasma glucose HbA1c (%) ---- ----------- --------------------------- ----------------------------- ---------------------------- --------------------------- -------------------- ------------------ --------------------- --------------------- ------------------ ----- 1 Lee 30.1 kg/m^2^ 23 kg/m^2^ 195.8 mg/dL 106.3 mg/dL 9.7% 5.9% 90% (*n* = 18/20) 55% (*n* = 11/20) 11.3% (*n* = 7) 0 % 2 Boza 33.7 (30.4--35) kg/m^2^ 23.9 kg/m^2^ 145 mg/dL 109.9 mg/dL 8.1% 6.5% --- 83.3% (*n* = 25/30) 33.3% (*n* = 10) 0 % 3 de Sa 33.6 kg/m^2^ 25.7 kg/m^2^ 176.1 mg/dL 93.9 mg/dL 8.4% 6.0% 74.1% (*n* = 20/27) 48.1% (*n* = 13/27) 25.9% (*n* = 7) 0 % 4 Huang 30.8 (25.0--34.8) kg/m^2^ 23.7 kg/m^2^ 204.2 mg/dL 103.5 mg/dL 9.2% 5.9% 90.9% (*n* = 20/22) 63.6% (*n* = 14/22) 9.1% (*n* = 2) 0 % 5 Scopinaro 30.6 (25.3--34.9) kg/m^2^ 25.3 kg/m^2^ 220 mg/dL 149 mg/dL 9.3 (7.5--12.9) % 6.5% 83.3% (*n* = 25/30) 30% (*n* = 9/30) 16.7% (*n* = 5) 0 % 6 Shah 28.9 kg/m^2^ 23 kg/m^2^ 233 mg/dL 89.2 mg/dL 10.1% 6.1% 100% (*n* = 15/15) 100% (*n* = 15/15) 0% (*n* = 0) 0 % 7 Lee 31 kg/m^2^ 24.6 kg/m^2^ 240.1 mg/dL 132.9 mg/dL 10.1% 7.1% --- 50% (*n* = 10/20) 0% (*n* = 0) 0 % 8 DePauLa 28.2 (20--34.8) kg/m^2^ --- 215.3 mg/dL 105.4 mg/dL 8.9 (7.5--12.8) % --- 91.2% (*n* = 53/58) 63.7% (*n* = 37/58) 10.3% (*n* = 6) 0 % 9 DePaula 25.7 kg/m^2^ 21.8 (17.7--25.8) kg/m^2^ 218.1 (90--334) mg/dL 102.0 (73--161) mg/dL 8.7 (7.5--13.7) % 5.9 (4.8--8.5) % 95.7% (*n* = 66/69) 65.2% (*n* = 45/69) 7.3% (*n* = 5) 0 % 10 Ramos 27.1 (25--30) kg/m^2^ 24.4 (20.2--28.3) kg/m^2^ 171.3 (127.0--242.0) mg/dL 96.3 (78.0--118.0) mg/dL 8.8 (7.5--10.2) % 6.8 (5.8--7.9) % 90% (*n* = 18/20) --- 0% (*n* = 0) 0 % 11 Ferzli 27.5 (21.7--33.0) kg/m^2^ 27.3 (23--33) kg/m^2^ 208.9 (112.0--286.0) mg/dL 154.9 (63.0--315.0) mg/dL 9.4 ( 6.6--11.8) % 8.5 (6.3--12) % 14% (*n* = 1/7) 14 % (*n* = 1/7) 0% (*n* = 0) 0 % 12 Geloneze 26.1 (1.7) kg/m^2^ 25.6 (1.2) kg/m^2^ 183.8 mg/dL 156.8 mg/dL 8.9% 7.8% 0% (*n* = 0/12) 0% (*n* = 0/12) 16.7% (*n* = 2) 0 % 13 Chiellini 30.9 kg/m^2^ 25.1 kg/m^2^ --- --- 8.5% 5.7% 100% (*n* = 5/5) --- --- 0 % 14 Lee 31.7 kg/m^2^ 23.2 kg/m^2^ 168.7 mg/dL 88.6 mg/dL 7.3% 5.6% --- 89.5% (*n* = 40/44) 4.5% (*n* = 2) 0 % 15 Scopinaro 33.4 (32.0--34.6) kg/m^2^ 27.1 (22.0--31.2) kg/m^2^ 252.7 (131--400) mg/dL 121.0 (68--146) mg/dL --- --- 100% (*n* = 7/7) --- --- 0 % 16 Cohen 29.6 (29.0--30.3) kg/m^2^ 28.3 (27--29.5) kg/m^2^ --- 83.0 (77--89) mg/dL --- 5.4 (5.0--5.7) % 100% (*n* = 2/2) 100% (*n* = 2/2) 0% (*n* = 0) 0 % 17 Cohen 32.5 (32.0--34.90 kg/m^2^ --- 146.0 (126--242)  mg/dL 88.0 (60--94) mg/dL --- \<6.0% 100% (*n* = 37/37) 100% (*n* = 37/37) 0% (*n* = 0) 0 % 18 Noya 33.2 (24.0--38.9) kg/m^2^ 27.6 (20.46--32.4)  kg/m^2^ --- --- --- --- 90% (*n* = 9/10) 90% (*n* = 9/10) 20% (*n* = 2) 0 % ###### Mean estimation of BMI, FPG, and HbA1c before and after metabolic surgery. Combined data from 18 existing bariatric studies. Variable (\# studies) Pre Means ± SE 95% CI Post Means ± SE 95% CI ------------------------- ---------------- ---------------- ----------------- ---------------- BMI (*n* = 14), kg/m^2^ 30.4 ± 0.98 (28.4, 32.3) 24.8 ± 0.33 (24.1, 25.5) FPG (*n* = 12), mg/dL 203.5 ± 8.2 (187.4, 219.6) 112.5 ± 4.4 (103.9, 121.1) HbA1c (*n* = 10), % 9.01 ± 0.22 (8.6, 9.5) 6.3 ± 0.14 (6.1, 6.6) SE: standard error. [^1]: Academic Editor: R. Prager

Introduction {#Sec1} ============ While online price comparison websites have burgeoned, there is scant understanding of how they influence online consumer (eSwitching) behavior. This study addresses this gap in the literature by investigating the influence of price comparison websites on online switching behavior. While Osakwe and Chovancová ([@CR31]) examined female shoppers' perceptions towards the use of price comparison websites, Jung et al. ([@CR18]) examined work shoppers' response to price comparison websites, and Pourabedin *et al.* ([@CR36]) investigated the role of customer value and attitudes in online channel switching behavior, there are no studies to the authors' knowledge that investigates the relationship between eSwitching behavior and price comparison websites. The emerging question, therefore, is whether the use of price comparison websites can lead to eSwitching behavior. This study by addressing this question seeks to close the void in our understanding concerning the relationship between price comparison websites and eSwitching behavior. This study also suggests some additional factors that may be considered when looking at this relationship. Accordingly, and based on a survey of existing research (Broniarczyk and Griffin [@CR6]; Hudders et al. [@CR17]; Osakwe and Chovancova [@CR31]; Xu et al. [@CR40]; Zhang et al. [@CR42]), it is argued that shoppers' innovativeness, their perceived usefulness of online ads, and their customer service experience consciousness are important factors to consider when evaluating the impact of price comparison websites on eSwitching behavior and this forms the overarching objective of our paper. We also argue that the most appropriate theoretical lens through which to investigate this relationship is that of the consumer empowerment paradigm (e.g. Broniarczyk and Griffin [@CR6]; Camacho et al. [@CR7]; Kucuk [@CR23]). In other words, by employing the consumer empowerment paradigm, this paper's objective is to investigate the relationship between (the use of) price comparison websites and eSwitching behavior in addition to the investigation of the determinants behind these phenomena, which we have identified above. Altogether, this allows us to contribute to the literature on online shopping in at least two important ways. First, is that we contribute to the online shopping literature particularly concerning the presentation of evidence that the use of price comparison websites, especially with regard to young and existing online shoppers, significantly empowers shoppers to engage in eSwitching behavior. The second contribution pertains to the investigation of the determinants behind the perceived use of price comparison websites in addition to the investigation of the empirical question on the determinants of eSwitching behavior. In particular, we find that shoppers' traits primarily customer experience consciousness and innovativeness are enablers to eSwitching behavior, while positive perceptions towards the use of online ads as well as customer experience consciousness engender the use of price comparison websites. Because it is commonly known that young adults, especially student population, are usually heavy users of online services and in general more receptive to emerging technologies (see also Yoo and Donthu [@CR41]), we chose to test the study hypotheses using a (university) student sample; which is also consistent with studies on consumer switching in the technology context (e.g. Bhattacherjee et al. [@CR3]). At the same time, it is worth noting that our sample is primarily made up of experienced internet users and who are familiar with online shopping. Taken together, this empirical context is appropriate for our investigation. The following literature review addresses each of these constructs in turn, before suggesting the theoretical model that was tested. This is followed by a brief overview of the methodology, and findings of an exploratory study to test the model. We conclude with a short discussion of these findings. Literature Informing Hypotheses Development {#Sec2} =========================================== Consumer Empowerment Paradigm {#Sec3} ----------------------------- The consumer empowerment paradigm in marketing is an important extension of the psychological empowerment construct, long studied, in the psychology literature (cf. Zimmerman [@CR43]; Cattaneo and Chapman [@CR8]). Camacho et al. ([@CR7]:294), while citing previous research, describe empowerment as "strategies or mechanisms that equip people with sufficient knowledge and autonomy to allow them to exert control over a certain decision". Similarly, it has been noted in the marketing literature that "empowerment requires mechanisms for individuals to gain control over issues that concern them, including opportunities to develop and practice skills necessary to exert control over their decision making" (Pires et al. [@CR35], p. 938). In light of previous discussions and among them Broniarczyk and Griffin ([@CR6]), Kucuk ([@CR23]), and Pires et al. ([@CR35]), the current investigation argues that this paradigm can help us gain understanding into the perceived use of price comparison websites and eSwitching behaviour in particular among existing shoppers. In general, the internet equips users with the tool to gain access to credible and quality information and this by implication confers more power and freedom of choice to users and in this case online shoppers. Moreover, since it is known that customers often rely on expert advice (in this instance, price comparison websites) when making purchase decisions (Camacho et al. [@CR7]), such advice - well-intentioned -- can dramatically reduce switching cost. One of the benefits of comparison tools and in this case price comparison websites is that they offer shoppers a range of choices (Broniarczyk and Griffin [@CR6]), thus allowing shoppers to shop across stores. Online switching behaviour (or eSwitching) is considered to be a possible outcome of this process. Similarly, since it is has been suggested in the literature that digital ads may be an important source of consumer empowerment today (Hudders et al. [@CR17]), it is considered therefore to play an influential role in shoppers' behavioural tendencies to use price comparison websites and to consequently eSwitch. Consequently, the following hypothesis is proposed:*H1: Price comparison websites' perceptions positively influence eSwitching behaviour.* At the same time, it is known that intrapersonal factors, described as "how people think about themselves and includes domain-specific perceived control and self-efficacy, motivation to control, perceived competence, and mastery" (Zimmerman [@CR43], p. 588), is a key aspect of empowerment. Accordingly, we consider intrinsic factors, or more precisely traits, like the customer service experience and consumer/shopper innovativeness, to be strong propelling force for price comparison websites use and eSwitching behaviour. Before discussing each of these in turn, we turn to eSwitching behaviour. eSwitching Behaviour {#Sec4} -------------------- The reasons why customers engage in switching intentions and/or behaviour remains an important topic in the literature to this day (Bhattacherjee et al. [@CR3]; Fan and Suh [@CR12]; Gopta et al. [@CR14]; Malhotra and Malhotra [@CR26]; Mosavi et al. [@CR30]; Pourabedin et al. [@CR36]). Multiple reasons exist for this kind of behaviour (Chuang and Tai [@CR9]; Keaveney [@CR19]; Malhotra and Malhotra [@CR26]; Mosavi et al. [@CR30]; Zhang et al. [@CR42]) but a factor often identified as a key determinant is the attractiveness of alternatives (Chuang and Tai [@CR9]; Xu et al. [@CR40]; Zhang et al. [@CR42]). In the context of this study, consumers might easily switch between online stores when it is believed they can get better price deals elsewhere. Price comparison websites lower the barriers to switching by offering such alternatives in one central place, with links that could easily navigate to said alternatives. Because price comparison websites are price aggregators, it offers the opportunity for shoppers to easily know the prices of each competing online stores and this consequently reduces switching cost. This study hypothesized (see H1) that these price comparison websites directly influence eSwitching behavior. The following section describes the former. Price Comparison Websites {#Sec5} ------------------------- Comparison shopping can be defined as "the practice of comparing the prices of items from different sources to find the best deal" (Hajaj et al. [@CR15]: 563). Price comparison websites provide the online alternative to this and we, therefore, define online price comparisons as *the online tool that allows for the comparison of item prices from different sources to find the best deal.* The use of price comparison websites has been acknowledged as an important search information tool, which has strong potential to alter shopping behaviour both in the online and offline environment (Bodur et al. [@CR4]; Broeckelmann and Groeppel-Klein [@CR5], Osakwe and Chovancová [@CR31]; Passyn et al. [@CR33]). Osakwe and Chovancová ([@CR31]:597) describe price comparison websites as "a near-frictionless marketing intermediary that can drive down online shoppers' search costs". Because shoppers can easily compare prices of similar firms and/or brands in a matter of seconds, the active use of price comparison websites not only reduces search costs but also empowers shoppers to buy from firms offering the best deal. This study argues that the greater the perceived usefulness of price comparison websites, the more it is expected that shoppers will use the information available on these sites in their pre- and post-purchase decisions. It, therefore, acts as a key mediator in the relationship between key influencing factors and online switching behavior. From a consumer empowerment perspective, it is pertinent to consider how these websites influence the relationship between eSwitching and shopper innovativeness, perceived usefulness of an ad and the consumers' service experience consciousness. Each of these now discussed in turn. Shopper Innovativeness {#Sec6} ---------------------- Consumer/shopper innovativeness, or what some scholars may well refer to as variety seeking propensity, is a well-researched concept in the literature (e.g. Agarwal and Prasad [@CR1]; Bhattacherjee et al. [@CR3]; Xu et al. [@CR40]; Mishra [@CR29]). In this instance and following prior literature (Agarwal and Prasad [@CR1]; Bhattacherjee et al. [@CR3]), we define shopper innovativeness as *a trait reflecting the willingness on the part of the shopper to experiment and/or try out any new products or range of services*. Moreover, because consumer innovativeness has been implied in the internet browser context to positively relate to switching intentions among a similar sample of respondents like this study (Bhattacherjee et al. [@CR3]), it stands to reason, that both the relationship between shopper innovativeness and eSwitching behavior, and its relationship to price comparison website perceptions begs investigation. Hence, we proposed that:*H2: Shopper innovativeness positively influences eSwitching behavior.H3: Shopper innovativeness is positively related to price comparison website perceptions.* Customer Service Experience Consciousness {#Sec7} ----------------------------------------- The role of customer service experience is well researched in academic and business literature (Berry et al. [@CR2]; Khan et al. [@CR20]; Meyer and Schwager [@CR28]; Osakwe and Chovancová [@CR31]). However, its relationship to online switching behavior and price comparison websites remains underdeveloped in empirical research. Service experience consciousness reflects a trait among shoppers who are highly demanding in service encounters, and thus tend to exhibit a higher dissatisfaction threshold. This trait, for example, has been reported to be an important enabler to customer-perceived use of price comparison websites (Osakwe and Chovancová [@CR31]), and is therefore included in the proposed conceptual model:*H4: Customer service experience consciousness is positively related to price comparison website perceptions.* We extend this research by arguing that since this set of shoppers invests a significant amount of cognitive, emotional and even intellectual resources into service interactions/encounters with the firm, they are therefore more demanding, difficult to please and more likely to move from one store to another in search for better services at all times. Customer service experience consciousness has already been identified as a trait reflecting the likelihood of a customer to become highly dissatisfied with service encounters and since customer dissatisfaction has been linked to switching intentions in prior research (e.g. Fan and Suh [@CR12]), it can be posited therefore that customer service experience consciousness and eSwitching behavior are related. Moreover, following previous research (Keaveney [@CR19]; Liang et al. [@CR25]), we argue that service inconvenience, impolite behavior on the part of service personnel, and occurrence of core service and service encounter failures will be less tolerated by those scoring high in customer service experience consciousness and thus in this case impacting eSwitching behavior. Consequently, the following hypothesis has been developed:*H5: Customer service experience consciousness is positively related to eSwitching behavior.* The final construct, from an empowered customer perspective, we felt pertinent to include in the study is that of the perceived usefulness of online ads. Perceived Usefulness of Online Ads {#Sec8} ---------------------------------- Consumers' attitudinal response to advertisements in general and online ads, in particular, is mixed (Le and Vo [@CR24]; Shavitt et al. [@CR38]; Schlosser et al. [@CR37]), yet it is considered an important contributing factor when investigating online consumer behavior (Ducoffe and Carlo [@CR11]; Mehta [@CR27]; Osakwe and Chovanocva [@CR31]; Paliwoda et al. [@CR34]). Some consumers have ill-feelings about online ads, others may be indifferent to online ads, while some have positive perceptions about online ads. In theory, however, ads are often meant to inform consumers and offer choices which they can easily choose from. Online ads, therefore, empower consumers concerning his/her buying decisions (cf. Ducoffe and Curlo [@CR11]). Therefore, in this instance, online ads can confer significant power to consumers (Hudders et al. [@CR17]), especially when it is perceived to be informative and valuable. In an online services context, strong perceptions towards online ads provide fertile ground for shoppers to become increasingly price-sensitive (Osakwe and Chovancová [@CR31]). This may be particularly true with regard to the use of price comparison websites. This study, therefore, extends this line of the suggestion by including eSwitching behaviour as an alternative outcome of online ads' perceived usefulness. Consequently, the following hypotheses were developed:*H6: Customers' perceived usefulness of ads positively influences their perception of price comparison websites.H7: Customers' perceived usefulness of ads positively influences their eSwitching behavior.* The above literature review and proposed hypotheses can be summarized in the theoretical model proposed in Fig. [1](#Fig1){ref-type="fig"}. This model extends prior research (e.g. Osakwe and Chovancová [@CR31]) by our assessment of enablers to the perceived use of price comparison websites and eSwitching behavior, especially concerning young and existing online shoppers.Fig. 1.Empirical model The methodology used to test the proposed model is outlined in the following section. Empirical Study {#Sec9} =============== Survey Data and Method {#Sec10} ---------------------- More specifically, student sample was used in the study because this is an important consumer segment for studying online behaviour and has also been extensively employed in the literature (Fan and Suh [@CR12]; Hong [@CR16]; Ozok and Wei [@CR32]; Wu et al. [@CR39]). This study recruited participants from one of the state universities in the Czech Republic using a convenience-based sampling approach which we consider to be most practical in this case. This study uses both online and self-administered surveys, nevertheless, most of the completed responses were from the self-administered questionnaire. Because we wanted to ensure that those who participated in the study have fairly good internet experience with online purchases, in the end -- particularly after deleting responses from six non-online shoppers - we had in total 345 valid responses. Therefore, the empirical focus is on existing online shoppers. The majority of sample respondents were female (59%), aged between 17--24 (80%), and undergraduates (66%). In this study, statistical analyses were performed using both IBM SPSS and WarpPLS (Kock [@CR22]). Finally, the research constructs - except for demographics - were measured using a five-point scale (ranging from completely disagree to completely agree). Construct Measurement Validation {#Sec11} -------------------------------- In order to improve the face and construct validity of the research constructs, constructs were adapted constructs from the literature. In particular, the measures for customer service experience consciousness, perceptions regarding the use of online ads and price comparison websites were based on Osakwe and Chovancová ([@CR31]), while the measure for consumer innovativeness was based on Daghfous et al. ([@CR10]) and finally the measure for eSwitching was modified from Kim et al. ([@CR21]) in addition to reading from the broader literature. The research hypotheses were tested by using the PLS-path modeling technique and precisely using mode A algorithm. The inspected composite reliability scores were as follows: 0.85 (online ads perceptions/OAD), 0.84 (price comparison websites use/PCWs), 0.80 (shopper innovativeness/INNOV), 0.71 (customer service experience consciousness/CSEC), and 0.74 (eSwitching behavior/eSWITCH). At the same time, all indicator loadings and weights were statistically significant at *p *\< 0.01, besides all the indicator loadings but two exceeded the 0.6 scores required for this kind of exploratory work. In terms of convergent validity, average variance extracted (AVE) scores range from 0.59 (OAD), 0.58 (PCW), 0.58 (INNOV), 0.46(CSEC), to 0.49 (eSWITCH). Although not reported here, following Fornell and Larcker ([@CR13]) discriminant validity was established for the constructs. Structural Model {#Sec12} ================ Model fit and quality criteria were inspected based on SRMR and R-squared contribution ratio (RSCR). We obtained 0.09 (SRMR value and thus acceptable since it is less than 0.1) and RSCR scores of 0.99 (which approximates to the ideal cut-off value of 1) (Kock [@CR22]). Regarding the hypothesized relationships, there is evidence for all but two (see Table [1](#Tab1){ref-type="table"} for details). Notably, the control variable i.e. gender neither statistically impacted price comparison websites use nor eSwitching. Finally, model predictive power concerning shoppers' use of PCW was 14%, whereas for eSwitching it was 30%; meaning that the empirical model explains about 14% and 30% variations in the use of price comparison websites and eSwitching respectively.Table 1.Structural Model statisticsHypothesisβ (t-ratios)*p*-*values*Significant/NotH1.19 (3.26)\<*.01*YesH2.36 (6.31)\<*.01*YesH3.04 (.51).30NotH4.29 (4.58)\<*.01*YesH5.26 (4.88)\<*.01*YesH6.20 (4.36)\<*.01*YesH7.03 (.49).31Not Short Discussion and Conclusion {#Sec13} =============================== This study has been able to identify antecedent factors leading to switching behaviour in the context of service and in particular in online stores beyond the usual suspects in the literature, for instance, attitudes towards switching (cf. Pourabedin et al. [@CR36]). Through the consumer empowerment paradigm, we find, not surprisingly, the perceived use of price comparison websites relates strongly with eSwitching. This novel finding in some ways mirrors the conclusion in past research about the role that search-intentions play in customers' channel switching (Gopta et al. [@CR14]). The point is that shoppers who use price comparison websites mainly use it for bargain hunting. This suggests that to reduce this positive effect on online switching behaviour, online retail merchants, particularly with a focus on young shoppers, will need to do more in the area of sales promotion and loyalty coupons as this might be one of the most effective ways to reduce the incidence of eSwitching and even customer churn. Also, we find that a higher possession of the following traits in shoppers namely innovativeness and their perception of the customer service experience empowers these shoppers to engage substantively in online switching behavior. Shopper innovativeness had the greatest impact on eSwitching. Since it has been suggested that individuals who are more likely to experiment with new ideas and/or products are more prone to switching (Bhattacherjee et al. [@CR3]; Xu et al. [@CR40]) and perceived innovativeness of the services provider inhibits customer switching intentions (Malhotra and Malhotra [@CR26]), the study's finding, therefore, is a reinforcement to extant research. Because shoppers who possess a higher level of service experience conscientiousness than others may be more prone to service dissatisfaction, the finding, therefore, mirrors previous discussions about the role of customer dissatisfaction in customer switching behaviour (cf. Chuang and Tai [@CR9]; Fan and Suh [@CR12]). Moreover, as predicted and consistent with previous research (Osakwe and Chovancová [@CR31]), this study finds that positive perceptions concerning online ad usefulness, in addition to service expectations, increasingly empower shoppers to use price comparison websites. Although this study initially proposed that consumer innovativeness and perceived use of price comparison websites are strongly related, evidence, however, shows it to be marginal, at best. Therefore, further research is needed to explore not only this insignificant finding but even further the supported research evidence reported in this paper. In other words, there is a need for more analysis on the research issues discussed in this work because until they are reassessed our findings are at best preliminary and limited to a specific sample. Furthermore, since the strength of relationships was never hypothesized, it is important therefore for future analysis to validate the assumption that shopper innovativeness, compared to others, has the strongest impact on online switching behaviour. Meanwhile, an important limitation of this analysis is that it was conducted using a student population and so makes it difficult to generalize beyond the target population. This consequently reinforces our call for further research on this topic. Finally, this study despite its limitations has added to the customer switching behaviour literature in addition to the heavily under-researched research stream of price comparison websites through the demonstration of the antecedents for eSwitching and shoppers use of price comparison websites based on the consumer empowerment paradigm. This work was supported by the Internal Grant Agency of FaME through TBU in Zlín No. IGA/FaME/2020/002; and further by the financial support of research project NPU I no. MSMT-7778/2020 RVO - Digital Transformation and its Impact on Customer Behaviour and Business Processes in Traditional and Online markets.

###### Supplementary figures, legends, and table ###### Click here for additional data file. ###### Countries in which eligible trials had been done ###### Click here for additional data file. Introduction ============ The predominant share of the global burden of disease is concentrated in less developed countries.[@ref1] However, until recently relative few trials were being conducted in these nations.[@ref2] [@ref3] Evidence on the management of many diseases affecting less developed countries often had to be tentatively extrapolated from studies performed in more developed countries with a longer standing tradition of conducting clinical research. This situation is currently changing. As participation rates in clinical trials decrease and cost increases in Western countries,[@ref4] [@ref5] [@ref6] many organisations providing contracts for research are focusing on eastern Europe,[@ref7] Asia,[@ref8] and South America,[@ref9] where the cost of recruiting participants is low. By 2015, for example, 15% of clinical trials are expected to be conducted in India.[@ref8] [@ref10] [@ref11] With the globalisation of international health,[@ref12] results of studies done in countries without a longstanding tradition of clinical research are becoming important to clinical practice in more developed nations.[@ref2] Trials carried out in less developed countries may differ in important aspects from those done in countries with stronger traditions in clinical research.[@ref13] Firstly, publication dynamics and biases may differ.[@ref14] Investigators in less developed countries may face a higher barrier against publication of "negative" results. Trials may remain unpublished or appear in domestic or local journals,[@ref15] [@ref16] and language biases may exist.[@ref17] [@ref18] [@ref19] Some national literatures on specific disciplines contain only significant results (for example, Chinese and Russian studies on acupuncture).[@ref20] Secondly, treatment effects may genuinely vary between countries owing to differences in study populations, baseline risk, concomitant diseases, background management, and clinical settings. We performed a large scale assessment of meta-analyses on topics with randomised evidence from more developed and less developed countries. We assessed how often randomised trials performed in these countries with different levels of development and different traditions in clinical research give different results, whether treatment effects are systematically larger in one setting than the other, and whether discordant effects are the result of bias or genuine differences. Methods ======= Definition of countries ----------------------- There are varying definitions and no perfect consensus on what countries should be included in the lists of "more developed" and "less developed," so their separation is not absolute.[@ref21] [@ref22] These categorisations try to take into account the per capita income but also other factors such as the composite human development index. For research purposes it is also helpful to know whether a country has a longstanding tradition in modern clinical research, uses critical scientific thinking, and in general applies empirical modern methods. We considered more developed countries to be those with both longstanding established market economies and longstanding traditions in clinical research.[@ref22] Such countries included the United States, Canada, Australia, New Zealand, Israel, and Japan, and western European countries. All other countries except for those in eastern Europe were considered as less developed. Our definition is consistent with the list of less developed countries of the International Monetary Fund, except that Israel (considered less developed until 2001 according to the International Monetary Fund) is classified among the more developed countries, given its strong longstanding research tradition, and eastern European countries (less developed according to the International Monetary Fund, except for Slovenia after 2007, the Czech Republic and Slovakia after 2009, and Estonia after 2011) are excluded. Eastern European countries were excluded from our analyses as they may have unique differences[@ref7] and are considered to be a separate group of countries in transition. We also performed sensitivity analyses where we excluded four Asian "tigers" (Hong Kong, Taiwan, Singapore, and South Korea) that evolved from less developed countries into advanced economies according to the International Monetary Fund in 1997, although their tradition of clinical research is not as longstanding as the main more developed countries. Sensitivity definitions excluding nations with a high per capita income but no tradition of clinical research (for example, several Arabian nations) yielded similar results, since few trials were identified that had been done in these countries. Eligible meta-analyses ---------------------- We identified meta-analyses that included data from one or more randomised trials conducted in a less developed country and one or more randomised trials conducted in a more developed country. Trials were classified on the basis of the countries in which participants were recruited; countries were considered with the names (for example, United Kingdom) or geographical indicators (for example, North America) as these were reported in the eligible Cochrane review. For consistency we focused on mortality, the most serious outcome. We searched the Cochrane database of systematic reviews (last update 27 August 2012) using terms for mortality (death OR mortality OR survival) in the title, abstract, or keywords. The reviews included in this database are considered to be thorough in searching for eligible studies.[@ref23] We excluded protocols and reviews that had been withdrawn, had no statistical synthesis on mortality, and had no country information of individual trials. Only reviews including randomised and pseudorandomised trials were eligible. Whenever a systematic review contained two or more different pertinent intervention comparisons we considered these separately. We also excluded reviews in which all the randomised trials from less developed or more developed countries had zero deaths. Multicentre international trials were eligible if all the countries were either less developed or more developed. Data extraction --------------- From each eligible trial we extracted the publication year, country of origin, number of participants, and number of deaths in each trial arm. We selected deaths from all causes; if, however, there were no data for all cause mortality we used cause specific mortality. Whenever there were several forest plots on mortality, we selected the one that reported overall data rather than subgroups. Whenever separate forest plots pertained to non-overlapping events for the same participants (for example, stillbirths and neonatal deaths), we selected whichever analysis included more deaths. For many conditions deaths are uncommon, conferring low power to show differences in effect sizes for mortality. Therefore, for each eligible topic we also examined separately the meta-analysis on the primary binary outcome (mortality or other). Whenever several eligible binary outcomes existed, we selected whichever had the largest number of studies regardless of whether this was mortality. Only meta-analyses with data from one or more trials from a less developed country and one or more trials from a more developed country were eligible. For each eligible trial we recorded the publication year, country, characteristics of participants, and events per arm. Two investigators (OAP, DGCI) independently extracted data. Any disagreements were resolved after discussion with the third investigator (JPAI). Statistical analysis -------------------- As a metric of relative risk, we used the odds ratio for our analyses when this could be estimated from available 2×2 tabular data for each trial. When this estimation was not possible we used the hazard ratio or risk ratio estimates as provided in the forest plots. Effect estimates for each trial were coined consistently to represent the odds of death or unfavourable primary outcome for the experimental (newer) intervention versus control. When survival or favourable primary outcomes were reported, we took the complementary mortality or unfavourable primary outcome event counts. Whenever two or more trials per country group were included in a forest plot, we synthesised them by fixed effects and random effects models.[@ref24] Fixed effects assume a common effect across the combined studies, whereas random effects assume that the true treatment effect may differ in each trial, and the summary aims at identifying an average treatment effect. Subsequently, for each topic we calculated the relative relative risk, with corresponding 95% confidence intervals, by dividing the summary relative risk from trials in more developed countries by that in less developed countries on the same topic. A relative relative risk \>1.00 means that the experimental intervention has more favourable outcomes in trials from less developed countries versus more developed countries. Furthermore, we calculated the summary relative relative risk across all topics of more developed versus less developed countries, by synthesising the relative relative risks of more developed versus less developed countries from each individual topic using a random effects model.[@ref25] Heterogeneity was probed using the Q statistic and I^2^ metric with corresponding 95% confidence intervals.[@ref26] [@ref27] In each eligible forest plot we also examined whether estimated intervention effects in smaller studies differed beyond chance from those estimated in larger studies (small study effects) using the Harbord's test[@ref28] when 2×2 data were available and the Egger's test[@ref29] otherwise; both test are considered to be significant for P\<0.10. For each topic where the results of trials from less developed countries differed beyond chance from trials from more developed countries, we examined whether there was evidence for small study effects. We also evaluated the constituent trials to examine whether there was any reason for anticipating ceiling effects related to the standard of care and the mode that an intervention was implemented in less developed or more developed countries---for example, whether an intervention was difficult to apply or required other concomitant interventions or background care to be effective. Moreover, we examined for each topic whether the baseline event risk in the control arms differed significantly between the two country groups by synthesising the baseline risks per country group by random effects using the Freeman-Tukey arcsin transformation.[@ref30] Such differences may mean that patients in these two settings vary in severity of disease, concomitant care, or other risk factors that can influence outcomes. Finally, for all topics where there were significant differences in treatment effects in more developed versus less developed countries, we also examined the risk of bias in the reported study design (mode of randomisation, allocation concealment, blinding, intention to treat, losses to follow-up). Quality deficits for these characteristics are associated with potential inflation in treatment effects in randomised controlled trials.[@ref31] In estimating the summary relative relative risk, sensitivity analyses were performed limited to topics where a nominally significant treatment effect for mortality had been found, when all trials were combined. Moreover, old trials may have different characteristics and less relevance to current practice.[@ref32] Therefore, we also performed sensitivity analyses excluding meta-analyses with any trials published before 1970. Finally, we performed sensitivity analyses excluding trials from countries that evolved from less developed to more developed countries. All analyses were done in Stata version 11.2. P values are two sided. Results ======= Eligible meta-analyses for mortality outcomes --------------------------------------------- The electronic search identified 2025 reviews. After exclusions (see supplementary fig A1), 131 eligible systematic reviews with 139 meta-analyses for mortality outcomes were considered (see supplementary table A1). The 139 meta-analyses included 1297 eligible trials (312 were conducted in less developed countries and 985 in more developed countries, see supplementary fig A2). The median publication year of eligible trials was 1997 (interquartile range 1990-2002). Each meta-analysis included a median of 13 trials (interquartile range 8-19) and 2856 participants (interquartile range 1355-11593). Trials from more developed countries did not have substantially larger sample sizes (median 117; interquartile range 54 to 319) than trials from less developed countries (105; 54 to 365): P=0.93, Mann-Whitney U-test. By fixed effects synthesis, 31 meta-analyses favoured (P\<0.05) the experimental or new intervention, five the control, and 103 showed no statistically significant difference. By random effects synthesis, the respective numbers were 27, 1, and 111. Significant evidence was found for small study effects in a total of 16/139 (12%) meta-analyses: 15/133 (11%) meta-analyses with Harbord's test and 1/6 meta-analyses with Egger's test. Significant differences in effect sizes for mortality ----------------------------------------------------- By using fixed effects to combine the relative risks from individual trials within the same country group, we identified 11 topics where the treatment effects between trials from more developed and less developed countries differed beyond chance (95% confidence intervals for relative relative risk excluded 1.00). For all these topics the experimental intervention had significantly less favourable results in the more than less developed countries (relative relative risk \>1.00, table 1[](#tbl1){ref-type="table"}). Antenatal corticosteroids[@ref33] noticeably reduced fetal and neonatal deaths when given to women at risk of preterm birth in trials conducted in less developed countries, but had a modest, non-nominally significant benefit in trials conducted in more developed countries (fig 1[](#fig1){ref-type="fig"}). A similar pattern was observed for corticosteroids in the treatment of sepsis or septic shock (fig 2[](#fig2){ref-type="fig"}),[@ref34] systemic antifungals in non-neutropenic critically ill patients (fig 3[](#fig3){ref-type="fig"}),[@ref35] calcium antagonists in aneurysmal subarachnoid haemorrhage (fig 4[](#fig4){ref-type="fig"}),[@ref36] intravenous immunoglobulin for preventing infection in preterm or low birthweight infants (fig 5[](#fig5){ref-type="fig"}),[@ref37] and transarterial embolisation in unresectable hepatocellular carcinoma (fig 6[](#fig6){ref-type="fig"}).[@ref38] Moreover, antioxidants, given for diverse conditions (fig 7[](#fig7){ref-type="fig"}),[@ref39] or specifically for prevention of gastrointestinal cancers (fig 8[](#fig8){ref-type="fig"}),[@ref40] and postoperative radiotherapy for non-small cell lung cancer (fig 9[](#fig9){ref-type="fig"})[@ref41] conferred a significantly increased risk of mortality in trials from more developed countries but not in trials from less developed countries (fig 1). Additionally, admission to hospital for bed rest for women with multiple pregnancy (fig 10[](#fig10){ref-type="fig"})[@ref42] tended to increase mortality in trials from more developed countries and decrease mortality in trials from less developed countries. Finally, altered fractionation radiotherapy compared with conventional radiotherapy resulted in nominally significant decreases in total mortality from oral cavity and oropharyngeal cancer in both less and more developed countries, although this was larger in trials from less developed countries (fig 11[](#fig11){ref-type="fig"}).[@ref43] {#fig1} {#fig2} {#fig3} {#fig4} {#fig5} {#fig6} {#fig7} {#fig8} {#fig9} {#fig10} {#fig11} ######  Statistically significant differences in treatment effects in less developed countries versus more developed countries Topic Experimental intervention\* Outcome Summary relative risks (95% CI) Relative relative risk (95% CI) for more *v* less developed countries P values -------------------------------------------------------------------- --------------------------------------------------------------- ------------------------------------------ --------------------------------- ----------------------------------------------------------------------- ------------------------ ---------- --------- Antenatal prevention in preterm birth^33^ Corticosteroids Fetal and neonatal deaths 0.83 (0.68 to 1.02) 0.40 (0.26 to 0.61) 2.08 (1.30 to 3.33) 0.103 \<0.001 Antioxidant supplements for prevention^39^ Antioxidants Mortality‡ 1.06 (1.03 to 1.10) 0.94 (0.85 to 1.05) 1.13 (1.01 to 1.27) 0.434 \<0.001 Multiple pregnancy^42^ Admission to hospital for bed rest Perinatal death 3.15 (0.88 to 11.23) 0.71 (0.35 to 1.46) 4.42 (1.03 to 18.99) 0.951 0.21 Treatment of sepsis and septic shock^34^ Corticosteroids All cause mortality at 28 days‡ 0.89 (0.74 to 1.08) 0.35 (0.14 to 0.87) 2.58 (1.01 to 6.63) 0.03 \<0.001 Prevention of gastrointestinal cancers^40^ Antioxidants Mortality 1.07 (1.04 to 1.11) 0.93 (0.84 to 1.04) 1.15 (1.03 to 1.29) 0.152 \<0.001 Non-neutropenic critically ill patients^35^ Systemic antifungals Mortality‡ 0.76 (0.58 to 1.01) 0.24 (0.08 to 0.68) 3.18 (1.08 to 9.40) 0.019 \<0.001 Treatment of non-small cell lung cancer^41^ Postoperative radiotherapy Mortality‡ 1.37 (1.12 to 1.68) 0.85 (0.57 to 1.28) 1.61 (1.03 to 2.53) 0.758 0.028 Aneurysmal subarachnoid haemorrhage^36^ Calcium antagonists alone Case fatality 0.86 (0.64 to 1.13) 0.15 (0.03 to 0.76) 5.73 (1.13 to 28.3) 0.302 0.99 Prevention of infection in preterm or low birthweight infants^37^ Intravenous immunoglobulin All cause mortality 0.95 (0.77 to 1.17) 0.49 (0.27 to 0.91) 1.93 (1.01 to 3.66) 0.137 \<0.001 Unresectable hepatocellular carcinoma^38^ Transarterial chemoembolisation or transarterial embolisation All cause mortality‡ 0.88 (0.72 to 1.08) 0.50 (0.31 to 0.81) 1.76 (1.05 to 2.97) 0.683 NP Oral cavity and oropharyngeal cancer^43^ Altered fractionation radiotherapy Total mortality‡ 0.91 (0.85 to 0.98) 0.57 (0.37 to 0.88) 1.60 (1.03 to 2.48) 0.790 NP Cirrhosis with upper gastrointestinal bleeding^44^ Antibiotics Bacterial infections‡ 0.30 (0.21 to 0.42) 0.13 (0.07 to 0.25) 2.30 (1.10 to 4.82) 0.088 \<0.001 Peripheral arterial disease of leg^45^ Lipid lowering regimens Cardiovascular events‡ 0.77 (0.70 to 0.85) 0.06 (0.01 to 0.46) 13.28 (1.68 to 104.97) 0.411 0.081 Pneumococcal infection in adults^46^ Vaccination All cause pneumonia‡ 0.85 (0.77 to 0.95) 0.52 (0.43 to 0.63) 1.64 (1.32 to 2.03) 0.864 \<0.001 Prevention of pre-eclampsia^47^ Antioxidants Pre-eclampsia‡ 0.96 (0.80 to 1.16) 0.38 (0.20 to 0.73) 2.54 (1.29 to 5.01) 0.071 0.60 Vitamin E supplementation in pregnancy^48^ Vitamin E Clinical pre-eclampsia‡ 0.55 (0.30 to 1.01) 0.06 (0.01 to 0.45) 9.52 (1.11 to 81.74) 0.858 0.48 Prevention of gastrointestinal cancers^40^ Antioxidants Incidence of gastrointestinal cancers‡ 1.05 (0.97 to 1.14) 0.86 (0.74 to 0.99) 1.23 (1.04 to 1.45) 0.004 \<0.001 Hypertension during pregnancy^49^ β blockers Caesarean section‡ 1.03 (0.80 to 1.32) 0.36 (0.17 to 0.77) 2.83 (1.28 to 6.29) 0.302 0.37 Hepatic encephalopathy^52^ Probiotics No recovery from hepatic encephalopathy‡ 0.03 (0.01 to 0.53) 0.61 (0.32 to 1.18) 0.04 (0.01 to 0.93) 0.429 0.09 Prevention of pre-eclampsia^51^ Antiplatelet agents Proteinuric pre-eclampsia‡ 0.74 (0.65 to 0.86) 0.93 (0.79 to 1.09) 0.80 (0.65 to 0.99) \<0.0001 0.13 Tuberculosis prevention in non-HIV infected people^53^ Isoniazid Active tuberculosis‡ 0.34 (0.28 to 0.42) 0.59 (0.40 to 0.86) 0.58 (0.38 to 0.90) 0.811 \<0.001 Prophylactic antifungal agents in very low birthweight infants^54^ Fluconazole Invasive fungal infection‡ 0.23 (0.11 to 0.49) 1.09 (0.48 to 2.48) 0.21 (0.07 to 0.64) 0.833 0.07 Prevention of infection in preterm or low birthweight infants^37^ Intravenous immunoglobulin Sepsis‡ 0.89 (0.74 to 1.06) 0.31 (0.16 to 0.62) 2.84 (1.40 to 5.79) 0.059 0.99 Antenatal care for low risk pregnancy^50^ Reduced number of visits or goal oriented visits Preterm birth‡ 1.24 (1.01 to 1.52) 0.99 (0.91 to 1.08) 1.26 (1.01 to 1.57) 0.787 NP Diarrhoea prevention^55^ Rotavirus vaccine Episodes of rotavirus diarrhoea‡ 0.48 (0.42 to 0.54) 0.61 (0.52 to 0.72) 0.78 (0.63 to 0.96) 0.611 0.006 NP=not pertinent because number of events in each arm was not available. \*In all cases, the experimental intervention has been compared with placebo or no treatment, except for surgery plus postoperative radiotherapy, which was compared against surgery alone; altered fractionation which was compared with conventional radiotherapy; reduced number of antenatal care visits or goal oriented visits, which were compared with standard care visits. †Individual trial relative risks from trials within the same country group were combined with a fixed effect model. ‡The primary binary outcomes of the respective systematic reviews. Evidence for small study effects was strong in the meta-analyses of steroids and antifungals and possibly also antenatal corticosteroids. The interventions were simple and easy to administer in diverse settings, regardless of the availability of other concomitant interventions and standards of care. The one possible exception was postoperative radiotherapy, where better outcomes might be expected in countries with higher standards of technology, although, if anything, the opposite was seen. The baseline risk of death was significantly higher in less developed countries in the meta-analyses of antenatal corticosteroids (33% *v* 16%), corticosteroids for sepsis (76% *v* 34%), systemic antifungals for non-neutropenic critically ill patients (54% *v* 28%), and intravenous immunoglobulin in preterm infants (19% *v* 13%), whereas it was significantly higher in more developed countries in meta-analyses of preventive antioxidants for various conditions (8% *v* 4%) or for gastrointestinal cancers (12% *v* 5%), and postoperative radiotherapy for non-small cell lung cancer (39% *v* 31%) (see supplementary table A2). Table 2[](#tbl2){ref-type="table"} shows the number of trials from the countries that had unclear or high risk of bias for randomisation sequence generation, allocation concealment, and blinding for the topics where significant differences in the treatment effects for mortality were documented. It is difficult to make comparisons within single topics, given the limited number of trials. However, summing the data across all trials, the proportion of trials with an unclear or high risk of bias was not significantly different in trials from more developed versus less developed countries for sequence generation (27% *v* 33%, P=0.45), allocation concealment (27% *v* 30%, P=0.68), or blinding (39% *v* 33%, P=0.53). ######  Risk of bias in studies from more developed versus less developed countries based on reported features Topic Experimental intervention Outcome Randomisation sequence generation Allocation concealment Blinding -------------------------------------------------------------------- --------------------------------------------------------------- ------------------------------------------ ----------------------------------- ------------------------ ---------- ------ ------- ------ Antenatal prevention in preterm birth^33^ Corticosteroids Fetal and neonatal deaths 7/10 0/3 7/10 1/3 5/10 1/3 Antioxidant supplements for prevention^39^ Antioxidants Mortality† 14/58 2/7 12/58 1/7 6/58 0/7 Multiple pregnancy^42^ Admission to hospital for bed rest Perinatal death 1/3 0/4 1/3 1/4 3/3 0/4 Treatment of sepsis and septic shock^34^ Corticosteroids All cause mortality at 28 days† 4/16 0/3 4/16 0/3 5/16 0/3 Prevention of gastrointestinal cancers^40^ Antioxidants Mortality 0/8 0/5 0/8 0/5 0/8 0/5 Non-neutropenic critically ill patients^35^ Systemic antifungals Mortality† 4/9 1/1 3/9 0/1 1/9 0/1 Treatment of non-small cell lung cancer^41^ Postoperative radiotherapy Mortality† 0/8 0/2 5/8 1/2 8/8 2/2 Aneurysmal subarachnoid haemorrhage^36^ Calcium antagonists alone Case fatality 0/10 1/1 3/10 1/1 5/10 1/1 Prevention of infection in preterm or low birthweight infants^37^ Intravenous immunoglobulin All cause mortality 7/10 5/5 1/10 4/5 6/10 5/5 Unresectable hepatocellular carcinoma^38^ Transarterial chemoembolisation or transarterial embolisation All cause mortality† 1/8 1/1 2/8 0/1 8/8 1/1 Oral cavity and oropharyngeal cancer^43^ Altered fractionation radiotherapy Total mortality† 3/13 1/1 3/13 1/1 13/13 1/1 Cirrhosis with upper gastrointestinal bleeding^44^ Antibiotics Bacterial infections† 1/7 3/5 3/7 4/5 7/7 5/5 Peripheral arterial disease of leg^45^ Lipid lowering regimens Cardiovascular events† 5/6 0/2 4/6 0/2 0/6 0/2 Pneumococcal infection in adults^46^ Vaccination All cause pneumonia† 4/9 2/4 7/9 3/4 5/9 3/4 Prevention of pre-eclampsia^47^ Antioxidants Pre-eclampsia† 1/4 2/4 1/4 2/4 0/4 1/4 Vitamin E supplementation in pregnancy^48^ Vitamin E Clinical pre-eclampsia† 1/2 1/1 1/2 1/1 0/2 0/1 Prevention of gastrointestinal cancers^40^ Antioxidants Incidence of gastrointestinal cancers† 0/6 5/11 0/6 5/11 0/6 0/11 Hypertension during pregnancy^49^ β blockers Caesarean section† 7/11 1/1 7/11 0/1 7/11 0/1 Hepatic encephalopathy^52^ Probiotics No recovery from hepatic encephalopathy† 0/1 1/3 1/1 3/3 1/1 3/3 Prevention of pre-eclampsia^51^ Antiplatelet agents Proteinuric pre-eclampsia† 13/30 5/10 20/30 7/10 26/30 8/10 Tuberculosis prevention in non-HIV infected people^53^ Isoniazid Active tuberculosis† 0/6 0/4 1/6 1/4 0/6 0/4 Prophylactic antifungal agents in very low birthweight infants^54^ Fluconazole Invasive fungal infection† 1/4 0/1 1/4 0/1 0/4 0/1 Prevention of infection in preterm or low birthweight infants^37^ Intravenous immunoglobulin Sepsis† 5/7 3/3 0/7 2/3 3/7 3/3 Antenatal care for low risk pregnancy^50^ Reduced number of visits or goal oriented visits Preterm birth† 2/4 2/3 2/4 1/3 4/4 1/3 Diarrhoea prevention^55^ Rotavirus vaccine Episodes of rotavirus diarrhoea† 7/14 1/6 10/14 1/6 4/14 0/6 \*Numbers represent trials with unclear or high risk of bias/total number of trials. †The primary binary outcomes of the respective systematic reviews. Summary of comparisons for mortality ------------------------------------ When summary relative risks from trials within each country group were synthesised by fixed effects, on average the results from more developed countries were significantly less favourable than those from less developed countries, with a summary relative relative risk of 1.12 (95% confidence interval 1.06 to 1.18, P\<0.001, I^2^=0%, 95% confidence interval 0% to 21%, Q statistic P=0.709). When data were synthesised within each country group by random effects, inferences were similar (1.08, 1.02 to 1.14, P=0.005, I^2^=0%, 0% to 21%, Q statistic P=0.922), but confidence intervals were wider and only the differences for antenatal corticosteroids, systematic antifungals, calcium antagonists, transarterial embolisation, and altered fractionation radiotherapy were beyond chance. Additionally, summary relative relative risks per fixed effects were 1.10 (1.04 to 1.18, P=0.002, I^2^=10.5%, 10% to 42%, Q statistic P=0.303) per fixed effects and 1.11 (1.01 to 1.21, P=0.023) per random effects for more developed countries versus China, and 1.21 (1.13 to 1.30, P=0.003, I^2^=92.3%, 90% to 94%, Q statistic P\<0.001) per fixed effects and 1.26 (0.91 to 1.74, P=0.158) per random effects for more developed countries versus India (the two less developed countries with the largest number of trials). Results were similar when analyses were limited to the 36 meta-analyses that had found nominally significant mortality effects overall at 1.15 (1.08 to 1.23, P\<0.001) per fixed effects and 1.17 (1.06 to 1.30, P=0.002) per random effects, I^2^=17%, 0% to 45%, Q statistic P=0.19, fig 12[](#fig12){ref-type="fig"}), the 124 meta-analyses where all trials had been published after 1970 (1.14, 1.08 to 1.21, P\<0.001, I^2^=0%, 0% to 22%, Q statistic P=0.81), and when excluding from calculations the four Asian countries (Hong Kong, Taiwan, Singapore, and South Korea) evolving into more developed countries (1.12, 1.06 to 1.18, P\<0.001, I^2^=0%, 0% to 22%, Q statistic P=0.624). {#fig12} Eligible meta-analyses for primary binary outcomes -------------------------------------------------- Overall, 127 meta-analyses had primary binary outcomes and available data from at least one more developed country and at least one less developed country (see supplementary table A3); for 58 of those the primary binary outcome was mortality. These 127 meta-analyses included a total of 1312 trials; 319 conducted in less developed countries (median sample size 121, interquartile range 58-318) and 993 in more developed countries (114, 53-310). The median number of trials per meta-analysis overall was 14 (interquartile range 9-21). By fixed effect synthesis 55 meta-analyses were overall in favour (P\<0.05) of the experimental intervention when all trials in the respective forest plot were considered, seven favoured the control, and 65 showed non-significant differences. By random effects, the respective numbers were 46, 5, and 76. In 26/127 (20%) meta-analyses evidence for small study effects was significant: 24/122 with Harbord's test and 2/5 with Egger's test. Significant differences in effect sizes for any primary outcome --------------------------------------------------------------- Combining by fixed effects model, treatment effects for the experimental intervention in trials from more developed and less developed countries, respectively, the relative differences were beyond chance (relative relative risk and 95% confidence intervals excluding 1.00) in 20 cases, of which six pertained to mortality and 14 to non-mortality outcomes (table 1; also see supplementary fig A3). In 15 of the 20 cases, results were more favourable (or less unfavourable) in trials from less developed countries (relative relative risk \>1.00). Antibiotic prophylaxis for bacterial infections in cirrhotic patients with upper gastrointestinal bleeding,[@ref44] lipid lowering regimens for peripheral arterial disease of the leg,[@ref45] vaccination for pneumococcal infection in adults,[@ref46] and intravenous immunoglobulin for sepsis in preterm or low birthweight infants[@ref37] had beneficial effects on the respective outcomes in trials from both less developed and more developed countries, although the benefit was considerably larger in less developed countries. Antioxidants[@ref47] and vitamin E[@ref48] prevented pre-eclampsia, antioxidants prevented gastrointestinal cancers,[@ref40] and β blockers decreased the rate of caesarean sections[@ref49] in trials from less developed countries but not in trials from more developed countries. Finally, the reduced number of antenatal visits or goal oriented visits only marginally increased the risk of preterm birth in more developed countries, but showed a small non-significant reduction in trials from less developed countries.[@ref50] Evidence of small study effects was strong for the meta-analyses of antioxidants for pre-eclampsia, antioxidants for gastrointestinal cancers, and intravenous immunoglobulin for sepsis. All interventions were simple and easy to apply in diverse settings and background standards of care. The baseline risk of the outcome was significantly higher in less developed countries in the meta-analyses of corticosteroids for treating sepsis (76% *v* 34%) and systemic antifungals for non-neutropenic critically ill patients (54% *v* 28%), and it was significantly higher in more developed countries in the meta-analyses of preventive antioxidants in various conditions (8% *v* 4%) (see supplementary table A2). For the remaining five non-mortality related primary outcomes, the treatment effects were more beneficial in trials from more developed countries. These included antiplatelet agents to prevent proteinuric pre-eclampsia,[@ref51] probiotics in hepatic encephalopathy,[@ref52] isoniazid prophylaxis against active tuberculosis in people not infected with HIV,[@ref53] prophylactic fluconazole for invasive fungal infections in very low birthweight infants,[@ref54] and rotavirus vaccine for the prevention of diarrhoea.[@ref55] For the meta-analysis of antiplatelets, evidence for the presence of small study effects was strong and the larger trials showed no clear benefit in either country group. All four interventions were simple and easy to administer in diverse settings. The risks in the control groups at baseline were significantly lower in more developed countries for active tuberculosis (2% *v* 15%), gastrointestinal cancer (2% *v* 3%), and rotavirus diarrhoea, whereas the risk was higher in more developed countries for bacterial infections in cirrhotic patients with upper gastrointestinal bleeding (43% *v* 28%) (see supplementary table A2). The proportion of trials with an unclear or high risk of bias was not significantly different in trials from more developed countries for sequence generation (42% *v* 45%, P=0.76), allocation concealment (52% *v* 52%, P=0.95), or blinding (51% *v* 41%, P=0.22, table 2). When the relative relative risks were synthesised across all 127 topics, there was some between topic heterogeneity with fixed effects summary relative odds ratio 1.07 (95% confidence interval 1.02 to 1.12, P=0.009) and random effects summary relative relative risk 1.09 (1.01 to 1.18, P=0.034, I^2^=37%, 95% confidence interval 22% to 49%, Q statistic P\<0.001). Discussion ========== We evaluated 139 meta-analyses with mortality outcomes, which included trials performed in less developed and more developed countries. In 11 cases, experimental interventions had significantly more favourable results in less developed countries than in more developed ones, whereas the opposite was never seen. On average, trials conducted in less developed countries had 1.12-fold more favourable effect sizes than trials done in more developed countries. When focusing only on interventions with an overall statistically significant impact on mortality the difference was 1.15-fold. Given that even effective interventions rarely achieve more than 1.15-fold to 1.20-fold reductions in the relative risk of mortality,[@ref56] relative differences of 1.10-fold to 1.15-fold may confound the presence or not of a genuine effect of many interventions. When we considered any primary binary outcome, in 20 topics treatment effects varied significantly based on the country group of included trials, and in 15/20 results were more favourable in less developed countries. Totally ineffective treatments may spuriously seem effective based on research published from less developed countries. As an increasingly larger share of clinical research is being done in less developed countries without strong research traditions, this may create a flood of spurious evidence. Possible explanations --------------------- Given the systematic preponderance of more favourable results in trials from less developed countries, one potential explanation is that the available randomised evidence from developed countries is more biased. An empirical evaluation of 307 published randomised trials from China, 117 from India, and 304 from Western countries showed that Indian and Chinese trials were of much lower methodological quality.[@ref57] Another empirical study highlighted that authors of Chinese trials often mislabelled basic study designs. Among 3137 studies indexed in the China national knowledge infrastructure database and claimed by their authors to be randomised, only 207 were indeed randomised.[@ref58] Most Chinese trials do not adhere to the CONSORT guidelines for reporting[@ref59] and many trials from less developed countries are not registered in ClinicalTrials.gov or even the composite World Health Organization trials registry.[@ref60] Trials from less developed countries tend to report on average more significant results.[@ref20] [@ref57] For example, such publication bias has been previously seen for Chinese (but not Indian) trials.[@ref20] [@ref57] [@ref61] Publication bias or selective analysis and outcome reporting biases[@ref62] [@ref63] may be influential in shaping this picture. A higher barrier to publication for authors from less developed countries that do not have a longstanding tradition in clinical research may further boost selective reporting.[@ref17] Of course mortality is a hard endpoint and more difficult to manipulate than other endpoints, but even for mortality, selective analysis may achieve inflated effects, as recently shown by corticosteroid trials.[@ref64] The presence of patterns showing small study effects is also suggestive (not conclusive) of selective reporting biases.[@ref65] Small study effects may also influence the literature in nations with strong traditions of running clinical trials. For example, this may be the case for antiplatelet agents to prevent proteinuric pre-eclampsia,[@ref51] where small trials suggest substantial benefits (more so in more developed than less developed countries), but the largest trials[@ref66] [@ref67] in both more and less developed countries have shown no benefits. Large, well conducted trials are needed to probe the claims for country specific major benefits and they may demonstrate that many of these claims are spurious. For example, after the publication of the examined Cochrane reviews, a recent large trial[@ref68] conclusively found no benefit from antioxidants in the prevention of pre-eclampsia (odds ratio 1.00), as opposed to the extremely large benefit that previous small trials had suggested (0.38). Additionally, in the topics where results between more and less developed countries differed, there was no overall pattern of having a higher or lower proportion of trials with unclear or high risk of bias in sequence generation, allocation concealment, or blinding, when research was compared between more developed and less developed countries. There is some evidence that these quality deficits are associated with inflated treatment effects, although the impact is lesser when the outcome is mortality.[@ref31] Although we cannot exclude the possibility that these quality aspects may have played a role in explaining the difference in some specific topics, they do not seem to be the main answer for the discrepancies overall. It should also be acknowledged that reported quality may not necessarily reflect the true quality of trials.[@ref69] Differences in treatment effects in less developed versus more developed countries may also be due to genuine differences rather than to biases. Low income and middle income countries face substantial financial barriers to the total healthcare budget,[@ref70] which may limit the implementation of expensive interventions.[@ref71] This might hold especially true for trials designed by the same body (institute, industry, etc) and conducted in more and less developing countries, as study quality and biases are usually not expected to be different, except maybe for selective reporting of negative results. However, we did not identify any discrepancies where the implicated intervention was expensive or difficult to administer and its efficacy may have depended largely on sophisticated background standards of care. The one exception was postoperative radiotherapy, but then the observed benefit was larger in less developed countries, a paradox that suggests that bias is a more likely explanation than differences in standards of care and technological ability. Trials with different results sometimes studied populations with different baseline risks. For example, mortality in newborns is on average higher in less developed countries[@ref72] and we cannot exclude the possibility that corticosteroids may result in a larger benefit in these locations. Similarly, the larger benefit of isoniazid prophylaxis in more developed countries may be explained by an increased burden of multidrug resistant tuberculosis,[@ref73] lower rates of treatment compliance,[@ref74] and limited access to healthcare[@ref74] in less developed countries. Potential limitations of the study ---------------------------------- Some caveats should be acknowledged. Firstly, we worked with available meta-analyses that may have already removed some biases from the primary literature. Meta-analysts may have contacted the authors of primary trials and obtained outcome information not reported in published reports, or they may have standardised outcomes, for example, to include all cause mortality and all available follow-up, whereas primary papers may have focused on subset analyses or other secondary analyses such as cause specific deaths.[@ref64] Thus bias may be larger in the reports of primary trials than that seen in meta-analysis based data. Secondly, agreement in treatment effects between more developed and less developed countries does not necessarily mean that both estimates are correct; occasionally both may be equally biased. Some recorded treatment effects in meta-analyses may simply reflect bias.[@ref75] For example, some investigators have argued convincingly that pneumococcal vaccination is ineffective in adults and that the apparent benefits in preventing pneumonia in the respective meta-analysis (seemingly larger in less developed countries) are entirely spurious.[@ref76] Thirdly, the endpoints of primary trials may not be the same as the respective primary endpoints in meta-analyses, and mortality might not be the primary endpoint for several considered trials. Even so, mortality is a major outcome and trials with favourable mortality results should attract attention regardless of whether this was a primary or secondary endpoint. In fact, an unexpectedly large number of small trials claim significant differences in mortality.[@ref77] Fourthly, not all organisations agree on what countries are less developed, and the status of countries has changed over time, with several previously less developed countries adopting market economies. However, these countries still can be separated from countries with longstanding traditions in clinical research, and sensitivity analyses using different definitions yielded similar results. We cannot separate whether per capita income or tradition in clinical trials research is the decisive factor that makes the difference, since few trials were done in countries without strong longstanding traditions in clinical research, where per capita income has increased dramatically in the past decades. Finally, some trials performed in less developed countries may be designed and coordinated by investigators in more developed countries. If anything this would tend to diminish differences between the two groups. Nevertheless, none of the trials implicated in the seven topics with statistically significant differences in mortality had such collaborative patterns. Moreover, it is possible that industry sponsorship may also affect the results of trials, in particular for expensive interventions where large markets are at stake. However, most of the interventions where discrepancies were identified were not expensive and sponsors would not have major invested interests. Conclusions and implications for future research ------------------------------------------------ Overall, in a globalised world, evidence from less developed countries will increasingly influence decisions in more developed countries and vice versa. It is important to generate randomised evidence in diverse settings including populations with differences in baseline risk, comorbidities, and access to healthcare. It is also important to improve the quality and minimise the biases of randomised trials around the world. Biases could be reduced through more thorough registration of trials from less developed countries, strengthening ethical standards,[@ref78] and a global view in the design and interpretation of the overall clinical research agenda.[@ref79] Meta-analyses of the available evidence can routinely explore differences and potential explanations thereof for trials performed in countries with different economies and traditions of clinical research. This information should be taken into consideration in guidelines and in the adoption of these interventions. ### What is already known on this topic 1. An increasing number of trials are performed in less developed countries with no longstanding tradition in clinical research 2. It is unclear whether the results of trials in more developed versus less developed countries are similar for the same intervention and conditions ### What this study adds 1. Randomised trials from less developed countries in a few cases show significantly different treatment effects from randomised trials in more developed countries 2. The reported treatment effects are larger in the less developed countries 3. These discrepancies may often reflect biases as well as genuine differences and should be taken into account when generalising evidence across different settings Contributors: JPAI conceived the original idea. OAP, DGC-I, and JPAI designed the study. OAP and DGC-I identified the eligible reviews and performed the data extraction. OAP and JPAI performed the statistical analyses. OAP, DGC-I, and JPAI interpreted the data and wrote the manuscript. All authors have critically commented on and approved the final version of the manuscript. JPAI is the guarantor. All authors, external and internal, had full access to all of the data (including statistical reports and tables) in the study and can take responsibility for the integrity of the data and the accuracy of the data analysis. Funding: This study received no funding. Competing interests: All authors have completed the ICMJE uniform disclosure form at [www.icmje.org/coi_disclosure.pdf](http://www.icmje.org/coi_disclosure.pdf) (available on request from the corresponding author) and declare: no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years, no other relationships or activities that could appear to have influenced the submitted work. Ethical approval: Not required. Data sharing: The statistical code and datasets are available from the corresponding author at [jioannid\@stanford.edu]([email protected]). Cite this as: *BMJ* 2013;346:f707

Correction to: *Scientific Reports* 10.1038/s41598-017-06666-2, published online 26 July 2017 This Article contains errors in Table [1](#Tab1){ref-type="table"} where the RT-qPCR Primer Sequences for the following genes,Table 1RT-qPCR Primer Sequences.GeneAccession \#ForwardReverse*ACAN*M_55172AGGGCGAGTGGAATGATGTTGGTGGCTGTGCCCTTTTTAC*COL1A2*NM_000089TTG CCC AAA GTT GTC CTC TTC TAGC TTC TGT GGA ACC ATG GAA*COL2A1*NM_033150CTG CAA AAT AAA ATC TCG GTG TTC TGGG CAT TTG ACT CAC ACC AGT*COL10A1*X60382GAAGTTATAATTTACACTGAGGGTTTCAAAGAGGCACAGCTTAAAAGTTTTAAACA*SOX9*Z46629CTTTGGTTTGTGTTCGTGTTTTGAGAGAAAGAAAAAGGGAAAGGTAAGTTT*YWHAZ*NM_003406TCTGTCTTGTCACCAACCATTCTTTCATGCGGCCTTTTTCCA*HOXC4*NM_014620.5ACCGTCGCATGAAATGGAATGCTGACCTGACTTTGGTGTTG*HOXC5*NM_018953.3GGTGCAGGCATCCAGGTACTCGGTGGGAAAGTGATGCTTAA*HOXC*8NM_022658.3AACCCGTGCTCGCTTAGCTGCCTCGTAGCCATAGAATTTGG*HOXD3*NM_006898.4GGAGCTTCCTGAGTGCACAATCCTCCAAACAGTCCTGGGTTT*HOXD8*NM_019558.3GGAATTTCTTTTTAACCCCTATCTGAGCTAGGGCGTGGGAAACC*FGFR1*NM_023110.2AACCTGACCACAGAATTGGAGGCTATGCTGCCGTACTCATTCTCCACA*FGFR2*NM_000141.4TGCACAAGCTGACCAAACGTCTGGACTCAGCCGAAACTGTT*FGFR3*NM_000142.4GCGCCTGAGGCCTTGTTTCCAAAGGACCAGACGTCACTCT*FGFR4*NM_002011.4ACCCCACGCCCACCATTGCGGTTCTCCCCATGAA *FGFR2, FGFR3 & FGFR4* are incorrect. The table with the corrected gene sequences is included below.

Introduction ============ Mostly, the amino acid sequence of a protein is conserved in order to maintain its function and structure. However, the conservation may also be caused by the selection at the nucleic acid level due to essential cis-acting sequences located in the protein-coding region. Thus, certain regions in a protein-coding sequence might encode specific amino acids not because of the selective pressure to the amino acid sequence, but because of the conservation at the nucleic acid level in DNA or RNA. There can be multiple reasons: the existence of nucleic acid secondary structures, splice sites, binding sites for proteins (e.g. transcription factors) or short RNAs, internal promoters, ribosome frameshifting signals, subgenomic promoters in RNA viruses, viral packaging signals and other regulatory elements. Additionally, the conservation at nucleic acid level might exist due to overlapping reading frames, which are common in viruses but also occur in cellular organisms ([@ref-31]; [@ref-41]; [@ref-48]; [@ref-2]; [@ref-36]; [@ref-6]; [@ref-13]). Computational annotation is extremely important when the experimental annotation is impracticable, for example, in case of organism or hosts which are uncultivable. However, thanks to the massive deployment of second-generation sequencing, the number of complete or near-complete genomes of previously unknown viruses has increased tremendously. This is one of the reasons why comparative analysis and computational annotation is needed to get some insight into the molecular biology of these viruses. Additionally, it has been shown that a large number of these new viruses will most likely constitute a new viral genus or even a family ([@ref-22]; [@ref-57]; [@ref-12]; [@ref-37]; [@ref-54]; [@ref-55]; [@ref-9]; [@ref-21]; [@ref-42]). Sometimes these novel viral species are too different from the existing species in the database; therefore, the homology-based methods are unable to detect any similarities to previously characterised sequences or cis-elements. However, thanks to the current advances in sequencing, the number of different relatives of the same virus can be quite high. Thus, a lot of evolutionary information is available. Proper analysis of these sequences can uncover at least some of the embedded functional elements and give us a better understanding of a virus ([@ref-17]; [@ref-13]; [@ref-40]). Several studies have used synonymous substitution restriction to identify overlapping or embedded functional elements in coding sequences of viruses ([@ref-43]; [@ref-17]; [@ref-25]; [@ref-13]; [@ref-40]). However, most of these methods detect overlapping or embedded elements only at a low resolution (over several codons) and often lack available implementation and/or a web interface. Here, we introduce the cRegions, which identifies regions within diverged protein-coding sequences where the distribution of observed nucleotides is significantly different from the expected distribution which is based on the amino acid composition and codon usage. Therefore, cRegions does not identify regions of excess synonymous constraint strictly, but rather compares observed codon usage to predicted codon usage at a single-nucleotide resolution. This allows cRegions to identify potential embedded functional cis-elements in coding sequences regardless of their nature. To demonstrate the capabilities of the cRegions web tool, we used the non-structural and structural polyprotein of alphaviruses as an example. Implementation -------------- The overall principle of the cRegions tool is to compare observed nucleotide frequencies to expected probability distribution and calculate appropriate metrics (described below) to detect regions where the coding sequence is more conserved than expected. Scripts used in the cRegions web tool are available in GitHub repository at <https://github.com/bioinfo-ut/cRegions>. The workflow of cRegions is as follows: Two inputs are required: a protein multiple sequence alignment (MSA) and nucleic acid sequences containing coding sequences (CDS) of respective proteins. Both inputs have to be in FASTA format. mRNA or the full genome can be used instead of the exact CDS. However, the coding sequence must not contain introns.Protein alignment is converted into a corresponding codon alignment using respective coding sequences with PAL2NAL ([@ref-46]).Henikoff position-based sequence weights are calculated using the codon alignment ([@ref-18]).Codon usage bias is calculated from the codon alignment. Calculated proportions are adjusted by Henikoff position-based sequence weights to account for non-uniform phylogenetic coverage ([Fig. S1](#supp-2){ref-type="supplementary-material"}). Codon preference for serine in the TCN block and in the AG\[A/G\] are calculated separately.Expected nucleotide proportions are calculated for each position in the codon alignment based on the amino acid sequence and the codon usage bias (calculated or provided by the user). Henikoff position-based sequence weights are used to adjust expected proportions ([Fig. S2](#supp-3){ref-type="supplementary-material"}).The observed nucleotide frequencies are compared to the expected probability distribution of nucleotides in each position. The comparison is made only for positions with amino acids having more than one codon. Three different metrics are used for this: The algorithm uses R ([@ref-34]) to calculate *p*-values of Chi-square goodness of fit test (chisq.test) for each column in the codon alignment. The test allows us to see whether the observed distribution of nucleotides is significantly different from expected distribution. We use the negative logarithm of the *p*-value of the Chi-square goodness of fit test as the metric. Bonferroni correction is used to show the threshold with significance level α = 0.05. $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$$p{\rm{ - value}} = {\rm{chisq}}.{\rm{test}}({\rm{c}}({{\rm{A}}_{{\rm{obs}}}},{{\rm{C}}_{{\rm{obs}}}},{{\rm{G}}_{{\rm{obs}}}},{{\rm{T}}_{{\rm{obs}}}}),\quad {\rm{p}} = {\rm{c}}({{\rm{A}}_{{\rm{exp}}}},{{\rm{C}}_{{\rm{exp}}}},{{\rm{G}}_{{\rm{exp}}}},{{\rm{T}}_{{\rm{exp}}}}))$$\end{document}$$ The subscript 'obs' indicates observed frequencies, the subscript 'exp' indicates expected proportions.The second metric is the root-mean-square deviation (RMSD). Only nucleotides which have a predicted probability over zero are included in the RMSD calculation. $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$${\rm{RMSD}} = \sqrt {{1 \over 4}\left[ {{{\left( {{A_{{\rm{obs}}}} - {A_{{\rm{exp}}}}} \right)}^2} + {{\left( {{C_{{\rm{obs}}}} - {C_{{\rm{exp}}}}} \right)}^2} + {{\left( {{G_{{\rm{obs}}}} - {G_{{\rm{exp}}}}} \right)}^2} + {{\left( {{T_{{\rm{obs}}}} - {T_{{\rm{exp}}}}} \right)}^2}} \right]}$$\end{document}$$ The subscript 'exp' indicates expected frequencies.The third metric is the maximum difference (MAXDIF), which selects only a single nucleotide from each column having the highest absolute difference between predicted and observed values. $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$${\rm{MAXDIF}} = \max \left( {\left| {{A_{{\rm{obs}}}} - {A_{{\rm{exp}}}}\left| , \right|{C_{{\rm{obs}}}} - {C_{{\rm{exp}}}}\left| , \right|{G_{{\rm{obs}}}} - {G_{{\rm{exp}}}}\left| , \right|{T_{{\rm{obs}}}} - {T_{{\rm{exp}}}}} \right|} \right)$$\end{document}$$The subscript 'exp' indicates expected frequencies. In all cases, the larger numerical value of a metric indicates higher conservation at the nucleic acid level. Additionally, if a position in the codon alignment has more than 20% of gaps, the metric is not calculated for that position ([Fig. S3](#supp-4){ref-type="supplementary-material"}). Material and Methods ==================== Alphavirus dataset ------------------ In the present work, we used the non-structural and structural polyprotein of alphaviruses as an example. Sequences were downloaded from the NCBI viral genome database (non-redundant dataset, 16 April 2018). The first dataset consists of 24 known alphaviruses ([Table S1](#supp-1){ref-type="supplementary-material"}). The non-structural polyprotein dataset was further divided into two sub-datasets. The first subset (SFV dataset) consists of seven viruses from genus Alphavirus, all belonging to the 'SFV Complex' ([Table S1](#supp-1){ref-type="supplementary-material"}) ([@ref-16]). The second subset was formed by nine 'New World' alphaviruses ([Table S1](#supp-1){ref-type="supplementary-material"}). cRegions and synplot2 --------------------- Protein sequences were aligned with MAFFT ([@ref-19]) using the default settings at <http://www.ebi.ac.uk/Tools/msa/mafft/>. Graphs were created with default settings using a sliding window with size one unless stated otherwise. Codon alignments for the synplot2 ([@ref-13]) were created with PAL2NAL ([@ref-46]), thus the input alignments for synplot2 and cRegions are identical. In this study the smallest possible window size was used for synplot2, giving the resolution of three codons (2*n* + 1 codons). In case of synplot2, significant hits were selected at threshold *p* \< 10^−5^. It is less conservative compared to the threshold used in the synplot2 paper ([@ref-13]). Sequence weighting ------------------ Henikoff position-based sequence weights are used to compensate for the over-representation of well-sequenced taxa in the MSA ([@ref-18]). Contrary to the original work of Henikoffs, we applied position-based weights to nucleotide sequences in the codon alignment, not to protein sequences. Thus, including variance at the codon level. Predicted nucleotide proportions for each position in the codon alignment are adjusted with sequence weights. Sliding window mode ------------------- Cis-acting elements may be longer than a single codon, for example, dual-coding regions, thus the possibility to calculate a single metric over consecutive codons may be preferred. The cRegions web tool allows the user to set the window size from one to 1/9 of the length of the codon alignment. By default, the third codon position is used in the sliding window mode, as it is most informative. An additional threshold exists in the sliding window mode. The threshold is for skipping columns instead of terminating the metric calculation for these consecutive positions. For example, if there is an insertion in a single sequence, the position should be skipped and the next codon included in the current window instead. It should be noted that skipping can happen several times in a row. By default, if a position in an MSA has more than 90% of gaps, it is skipped in the sliding window mode. It should be noted that the threshold for skipping gaps and the threshold for metric calculation are different parameters ([Fig. S3](#supp-4){ref-type="supplementary-material"}). In case of RMSD and MAXDIF, an arithmetic mean is calculated over consecutive codons. However, a *p*-value of Chi-square test is calculated based on observed values that are added over all consecutive codons. Again, Bonferroni correction is used to show the threshold with significance level α = 0.05. It should be kept in mind that adjacent positions with low metric values will decrease the value of a single conserved position if the window size is larger than one. Visualisation ------------- The cRegions web tool uses 'highcharts' libraries to visualise results (<http://www.highcharts.com/>). Alignment visualisation is provided by MSAViewer ([@ref-53]). The combination of highcharts and MSAViewer allows the user to pinpoint (by clicking on the point) and navigate directly to a conserved region or nucleotide. In addition to scatter plots of different metrics, cRegions web tool displays an interactive graph of codon usage. Codon usage is calculated over all analysed sequences. A table of codon frequencies and a file with tab-separated values are included in the downloadable zip container. The same codon table can be used as an input for the cRegions algorithm. VEEV sequences -------------- The Venezuelan equine encephalitis virus (VEEV) neighbour sequences were downloaded from the NCBI viral genomes database (<https://www.ncbi.nlm.nih.gov/genomes/GenomesGroup.cgi?taxid=11018>, 18 May 2018). A total of 11 entries were removed as they did not have annotated full-length non-structural polyprotein. Identical protein sequences were removed using jalview 'remove redundancy 100' ([@ref-50]). The final dataset contained a total of 94 isolates, including 93 VEEV neighbour sequences and a reference VEEV sequence ([NC_001449](NC_001449)). Randomly mutated protein-coding sequences ----------------------------------------- A random 3,000 nt long protein-coding sequence was created with SMS v2 tool (<http://www.bioinformatics.org/sms2/random_coding_dna.html>). A different number of random mutations (25--1,600) were introduced into that sequence with the SMS2 mutate tool (<http://www.bioinformatics.org/sms2/mutate_dna.html>) ([@ref-44]). In total, we generated 3 × 8 datasets. Each set consisted of one original randomly generated protein-coding sequence and six, nine or 14 randomly mutated sequences with a different number of mutations per bp. For the set with 15 sequences we generated different initial protein-coding sequence to remove a bias which could occur if we only use one protein-coding sequence as a seed. These sequences do not need aligning, because homologous nucleotides are already aligned. Threshold correction for nearly identical sequences --------------------------------------------------- cRegion algorithm assumes that, in general, the distribution of nucleotides at each position in the MSA correspond to an average codon usage of the same sequences under analysis. The assumption is reasonable if the sequences under analysis have diverged. However, in the case of nearly identical sequences, the distribution of nucleotides in each position in the MSA is more similar to observed nucleotide proportions rather than to average codon usage. Therefore, the expected proportions are more similar to observed proportions. Thus, even if using Bonferroni correction, the Chi-square test may still give many potentially false positive signals. Therefore, we need to adjust the threshold such a way that in the case of nearly identical sequences the threshold is stricter. For that, we also include the average pairwise identity of nucleotide sequences into calculations of expected nucleotide proportions at each position. The expected nucleotide proportions are adjusted by the observed proportions which depend on the average pairwise identity of nucleotide sequences. The exponent value was found empirically. $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$${\left( {{A_{{\rm{exp}}}},{C_{{\rm{exp}}}},{G_{{\rm{exp}}}},{T_{{\rm{exp}}}}} \right)_{{\rm{adj}}}} = \left( {{A_{{\rm{exp}}}},{C_{{\rm{exp}}}},{G_{{\rm{exp}}}},{T_{{\rm{exp}}}}} \right) + {i_n}\hskip-3.7pt^7*\left( {{A_{{\rm{obs}}}},{C_{{\rm{obs}}}},{G_{{\rm{obs}}}},{T_{{\rm{obs}}}}} \right)$$\end{document}$$ *i~n~* = the average pairwise identity of nucleotide sequences In addition to the previous adjustment of expected values, also the threshold itself is adjusted. The threshold correction depends on the ratio between the average pairwise identity of nucleotide and protein sequences and the number of sequences in the MSA.  \\documentclass\[12pt\]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{upgreek} \\usepackage{mathrsfs} \\setlength{\\oddsidemargin}{-69pt} \\begin{document} }{}\$\${{t\_{{\\rm{corrected}}}} = {t\_{{\\rm{current}}}}\*{{{i_n}} \\over {{i_p}}}\*\\left({1 + {i_n}\\hskip-3.7pt\^d} \\right)}\$\$\\end{document} t corrected = t current \* i n i p \* ( 1 \+ i n d ) 1. *t* = threshold − log10(*p*--value) 2. *i~n~* = the average pairwise identity of nucleotide sequences 3. *i~p~* = the average pairwise identity of protein sequences 4. *d* = the number of sequences in the multiple sequence alignment Results ======= Alphaviruses ------------ First, we applied cRegions and synplot2 on all 24 non-structural polyproteins of alphaviruses (see also 'alphavirus dataset' example on the cRegions homepage). We detected a total of six significant signals with cRegions ([Fig. 1A](#fig-1){ref-type="fig"}) and three significant signals with the synplot2 ([Figs. S4A](#supp-5){ref-type="supplementary-material"} and [S5A](#supp-6){ref-type="supplementary-material"}). The first signal from the 5′ end was recognised by both programs ([Fig. 1A](#fig-1){ref-type="fig"}) and spanned from positions 138 to 174 in the codon alignment ([Table 1](#table-1){ref-type="table"}). It is a conserved sequence element (CSE) called '51 nt CSE', which acts as an enhancer for the RNA synthesis, affecting viral replication. This CSE forms two stem-loops and is located at positions 155--205 in the Sindbis virus (SINV) genome ([@ref-29]). Thus, the detected signal lies exactly in the region ([Table 1](#table-1){ref-type="table"}). ![cRegions analysis of non-structural polyproteins of alphaviruses using Chi-square goodness of fit test.\ On each graph, the *y*-axis shows the negative logarithm of the Chi-square goodness of fit test *p*-value and the *x*-axis shows the position in the codon alignment. The red line represents the significance threshold (α = 0.05 with Bonferroni correction). (A) Non-structural polyprotein alignment of all 24 Alphaviruses. (B) Non-structural polyproteins of 'New World' Alphaviruses. (C) Non-structural polyproteins of 'SFV Complex' Alphaviruses. Non-structural polyprotein sequences were aligned with MAFFT version 7 using the default settings at <http://www.ebi.ac.uk/Tools/msa/mafft/> ([@ref-19]). Graphs were generated with sliding window mode (window size = 1). On the panel title, the number of analysed sequences is shown in parentheses.](peerj-07-6176-g001){#fig-1} 10.7717/peerj.6176/table-1 ###### Detected signals in the codon alignment in different datasets and respective positions in SFV and SINV genome.  Signal Description Dataset Position on the codon alignment SFV[\*](#table-1fn2){ref-type="fn"} SINV[\*](#table-1fn2){ref-type="fn"} ------------------------------------------------- ---------------------------------------------- ----------------------------------------- ----------------------------------------- --------------------------------- ------------------------------------- -------------------------------------- Non-structural polyprotein 1 51 nt CSE All ([Fig. 1A](#fig-1){ref-type="fig"}) 138--174 184--220 161--197 2 Signal adjacent to capsid binding region All ([Fig. 1A](#fig-1){ref-type="fig"}) 1,149 NA 1,148 New world ([Fig. 1B](#fig-1){ref-type="fig"}) 1,086 and 1,092 NA 1,142 and 1,148 3 Packaging signal of SFV Complex alphaviruses All ([Fig. 1A](#fig-1){ref-type="fig"}) 2,835 2,812 2,804 SFV Complex ([Fig. 1C](#fig-1){ref-type="fig"}) 2,730 2,812 2,804 4 Signal inside the b region All ([Fig. 1A](#fig-1){ref-type="fig"}) 2,967 2,944 2,936 5 Signal adjacent to leaky stop codon All ([Fig. 1A](#fig-1){ref-type="fig"}) 6,834 5,536 5,768 New world ([Fig. 1B](#fig-1){ref-type="fig"}) 6,159 NA 5,888 6 Subgenomic promoter of alphaviruses All ([Fig. 1A](#fig-1){ref-type="fig"}) 8,658 and 8,664 7,354 and 7,360 7,583 and 7,589 Structural polyprotein 1 UUUUUUA motif All ([Fig. 2](#fig-2){ref-type="fig"}) 2,673--2,679 9,825--9,831 10,022--10,028 **Notes:** NA, not applicable. Shows respective position(s) on the SFV ([NC_003215](NC_003215)) and SINV ([NC_001547](NC_001547)) genome. The second significant hit, a single nucleotide at position 1,149, was detected only by the cRegions algorithm. However, two adjacent positions 1,143 and 1,146 were just below the threshold. In the New World alphavirus dataset, in addition to position 1,149, 1,143 was also significant. The signal is just adjacent to the packaging signal of SINV and New World alphaviruses (see also 'New World alphavirus dataset' example on the cRegions the homepage). It has been shown that a 570 nt fragment positions 684--1,253 from the SINV binds to the viral capsid protein and is required for packaging of SINV. The detected signal lies in this region ([@ref-51]). However, when we analysed VEEVs separately we were able to detect the positions of phylogenetically conserved predicted stem-loops ([Fig. S7](#supp-8){ref-type="supplementary-material"}). The results are similar to the work done by [@ref-20]. The third and the fourth signal are also single nucleotides at positions 2,835 and 2,967, respectively ([Fig. 1A](#fig-1){ref-type="fig"}). Both signals are located inside nsp2 conserved region called region b. This 266-nucleotide region is located from 2,726 to 2,991 in the SFV genome ([@ref-52]). Previous deletion mutation analysis has shown that nucleotides from 2,767 to 2,824 in the b region are required for efficient packaging of SFV genome. ([@ref-52]). The first signal is located in that region. Additionally, analysis of 'SFV Complex' viruses separately led to increased significance of the first signal ([Fig. 1C](#fig-1){ref-type="fig"}) and the same signal became visible with synplot2 ([Figs. S4B](#supp-5){ref-type="supplementary-material"} and [S5B](#supp-6){ref-type="supplementary-material"}). Expectedly, both signals disappeared in the New World dataset, as the packaging signal is in a different location in these viruses ([Fig. 1B](#fig-1){ref-type="fig"}). Therefore, dividing datasets to different subsets may help to detect signals that are only characteristic to smaller subgroups. The fifth significant hit is a single nucleotide at position 6,834. It is downstream of the 'leaky' stop codon (stop codon is at 6,814--6,816 on the codon alignment and in the SINV genome at nt 5,748--5,750). Synplot2 was able to detect a much larger region compared to cRegions in the same area ([Figs. S4A](#supp-5){ref-type="supplementary-material"} and [S5A](#supp-6){ref-type="supplementary-material"}). The detected signal is a 3′ stem-loop RNA secondary structure immediately adjacent to the stop codon (+13 nt downstream of the stop codon in SINV). For many alphaviruses, including VEEV and SINV, it has been reported to influence read-through. In the SINV genome, the double helix part (the stem) of the stem-loop is predicted to form between the two regions: 5,763--5,772 and 5,928--5,939 ([@ref-15]). Therefore, the detected signal at position 6,834 (5,768 in SINV) is inside the first region. However, when we analysed VEEVs separately, we were able to detect multiple significant signals inside this stem-loop region ([Fig. S7](#supp-8){ref-type="supplementary-material"}). The sixth signal consists of two positions 8,658 and 8,664 on the codon alignment ([Fig. 1A](#fig-1){ref-type="fig"}; [Figs. S4A](#supp-5){ref-type="supplementary-material"} and [S5A](#supp-6){ref-type="supplementary-material"}). The signal is located within the subgenomic promoter of alphaviruses ([@ref-35]; [@ref-38]). The cRegions and the synplot2 were also applied to the structural polyproteins of alphaviruses (see also 'alphavirus structural dataset' example on the cRegions homepage). Sliding window size 2 was used with cRegions. A strong signal was detected in positions 2,643--2,649 on the codon alignment, which corresponds to a UUUUUUA motif ([Fig. 2](#fig-2){ref-type="fig"}). The motif is responsible for a frameshift in a structural protein ([@ref-14]; [@ref-7]). The same signal was detected with the synplot2 ([Fig. S6](#supp-7){ref-type="supplementary-material"}). ![cRegions analysis of structural polyproteins of alphaviruses.\ A significant signal was detected in codon alignment positions 2,643--2,649, the region corresponds to a known UUUUUUA motif. The *y*-axis on the plot shows the negative logarithm of the Chi-square goodness of fit test *p*-value and the *x*-axis shows the position on the codon alignment. The red line represents the significance threshold (α = 0.05 with Bonferroni correction). Structural polyprotein sequences were aligned with MAFFT version 7 using the default settings at <http://www.ebi.ac.uk/Tools/msa/mafft/> ([@ref-19]). Sliding window size 2 was used.](peerj-07-6176-g002){#fig-2} Requirements on sequences ------------------------- The method used in cRegions has some limitations and prerequisites ([@ref-33]). First, the sequences under study must have diverged. Second, the embedded functional element must have been under selection. To help users to evaluate their sequences in these aspects, we added an interactive version of [Fig. 3](#fig-3){ref-type="fig"} to the web tool. The plot visualises the sequences under study in comparison to randomly mutated sequences and sequences thoroughly analysed in the previous or current study with respect to divergence and selection. To evaluate divergence and selection we used the relationship between average pairwise nucleotide identity and average pairwise amino acid identity ([Fig. 3](#fig-3){ref-type="fig"}). As shown in [Fig. 3](#fig-3){ref-type="fig"}, randomly mutated simulated sequences form a clear and narrow assembly on the plot. Randomly mutated sequences with a defined extent (N mutations per bp) were used to model neutral evolution and/or non-diverged sequences (more details in 'Materials and Methods'). The naturally occurring sequences used in the previous and in the current study locate clearly away of the simulated sequences. ![The average pairwise identity of nucleotide sequences from codon alignment plotted against the average pairwise identity of protein sequences in respective MSA.\ A different number of random mutations (25--1,600) were introduced into a randomly generated 3,000 nt long protein-coding sequence with the SMS2 mutate tool. On the plot, three different lines represent sets of 7, 10 or 15 sequences. Therefore, each data point on a line consisted of one original protein-coding sequence and 6, 9 or 14 mutated sequences. The number near the point shows mutations per base pair. The real datasets are marked with crosses. Majority of real datasets visualised on the plot are from this paper, others originate from our previous paper ([@ref-33]).](peerj-07-6176-g003){#fig-3} Sequences having low divergence or/and having close to neutral evolution ------------------------------------------------------------------------ As cRegions was designed to work on diverged sequences, the method may give potential false positive signals in low divergence sequences or in sequences locating close to neutrally evolving sequences ([Fig. 4](#fig-4){ref-type="fig"}; [Fig. S8](#supp-9){ref-type="supplementary-material"}). To avoid this, we recommend enabling threshold correction on the web tool. By enabling this option expected values are corrected with observed values and the adjusted threshold is calculated (see 'Materials and Methods'). This removes most of the false positive signal from sequences that are close to randomly mutating sequences ([Fig. 4](#fig-4){ref-type="fig"}; [Fig. S8](#supp-9){ref-type="supplementary-material"}). We would like to note that the correction is needed only in the case of sequences which are close to the neutrally/randomly evolving sequences ([Fig. 3](#fig-3){ref-type="fig"}). Another option is to use synplot2 which uses neutral evolution as its null hypothesis ([@ref-13]) ![The number of signals in randomly mutated sequences.\ A random 3,000 nt long protein-coding sequence was created with SMS v2 tool (<http://www.bioinformatics.org/sms2/random_coding_dna.html>) ([@ref-44]). A different number of random mutations (25--2,000) were introduced into the original randomly generated protein-coding sequence six times with the SMS v2 mutate tool (<http://www.bioinformatics.org/sms2/mutate_dna.html>) ([@ref-44]). Each simulated dataset consisted of seven sequences each having the same number of mutations. Orange circles show the number of signals exceeding the threshold without a threshold correction and using window size 1. Red circles show the number of signals with sliding window size 2. The deep blue-green (teal) shows the number of signals exceeding the threshold when applying threshold correction.](peerj-07-6176-g004){#fig-4} Discussion ========== Given sufficient evolutionary time, the conservation of amino acid residues in different homologous sequences will not necessarily imply the same conservation at nucleotide level due to the redundancy of the genetic code. However, it can be reasoned that functionally important cis-acting elements embedded in protein-coding sequences will be evolutionarily conserved, even if these regions are subject to constant evolutionary pressure both through their translation product (amino acid sequence) and cis-acting functions. Previous studies have produced numerous valuable methods for detecting overlapping or embedded functional elements in coding sequences, mainly by identifying regions of excess synonymous constraints ([@ref-43]; [@ref-17]; [@ref-25]; [@ref-13]; [@ref-40]). Currently, only Firth has created a web interface to the synplot2 and Sealfon et al. have implemented FRESCo as a usable batch script. However, running a script in a Unix terminal might be a daunting task for a biologist, therefore, a web interface is essential for bioinformatics tools to be widely adopted. Additionally, FRESCo needs a phylogenetic tree as an input, which depends highly on the construction method ([@ref-40]). During evolution, homologous protein-coding sequences accumulate random substitutions by switching to different synonymous and non-synonymous codons. In addition, if purifying selection acts on a protein sequence, synonymous substitutions are favoured over non-synonymous substitutions. However, synonymous codons are not used with equal proportions ([@ref-26]; [@ref-10], [@ref-11]; [@ref-4]; [@ref-56]; [@ref-32]; [@ref-1]; [@ref-3]; [@ref-49]). One hypothesis that explains the preferential use of synonymous codons is the adaptation towards translational efficiency and gene expression ([@ref-26]; [@ref-10], [@ref-11]; [@ref-4]; [@ref-32]; [@ref-3]; [@ref-5]). Gene expression and translational efficiency can also be affected by consistent under- or over-representation of certain codon pairs. Multiple works conducted on RNA viruses have shown that altering codon pair frequencies towards those that are disfavoured in their host will reduce virus replication ([@ref-8]; [@ref-24]; [@ref-30]). However, the effect may be an artefact of changes in the CpG and UpA dinucleotide frequencies instead ([@ref-47]). In addition, host adaptation theory (tissue adaptation) may explain the preferential use of synonymous codons. It has been shown that the codon usage is strongly related to the specific host in both bacterial and human viruses. Also, the highest level of adaptation to host codon usage is for proteins which appear abundantly in the virion, meaning that the codon usage of virion and non-virion proteins differ ([@ref-1]). Additionally, [@ref-49] have proved that there is a difference in the codon usage between the host-interacting protein and the rest of structural late phase proteins. Thus, it can be reasoned that each protein-coding gene may have different preferential use of synonymous codons. Therefore, incorporating the preferential use of synonymous codons of the same set of genes into the model is reasonable. However, authors of the synplot2 used neutral evolution as the null model, therefore not including codon usage bias into the analysis. They reasoned that it was not required based on the results and it would be impossible to accurately estimate given the limited genome size of RNA viruses ([@ref-13]). Also, [@ref-40] argued that the genetic region of a typical virus is only about a thousand of codons long, therefore, there may be insufficient information to characterise the codon usage bias. To date, only the method published by [@ref-17] incorporated codon usage bias into their model. Calculating the preferential use of synonymous codons may result in incorrect assessment on many occasions. First, if the analysis is performed on sequences with low divergence, then the estimation of codon usage may be biased. Second, a large overlapping region or the abundance of rare codons in a gene may also affect codon usage estimation. The same conclusion was reached in the synplot2 paper by Firth, where he noted that the divergence parameters of the null model are determined from the full coding region and if the alignment contains extensive overlapping regions then the neutral divergence rates will be underestimated ([@ref-13]). As a remedy, cRegions web tool allows the user to input a custom codon table, which may have been calculated from a larger set of genes and be more suitable in some cases. Additionally, cRegions allows the user to choose between 17 different codon tables that are also implemented in PAL2NAL ([@ref-46]). However, not all synonymous codons should be treated as equals. In case of serine, two mutations are necessary to move from AG\[A/G\] serine codon block to TCN codon block (in the standard codon table). Therefore, usage for codons in the TCN block and in the AG\[A/G\] are calculated separately. For example, if only AG\[A/G\] serine was observed, then only AG\[A/G\] codon proportions were used for predictions and vice versa. This is not implemented in the method published by [@ref-17]. Also, it was not implemented in the original publication of the method ([@ref-33]). As it is impossible to assess the conservation at the nucleic acid level if an amino acid is encoded only by a single codon (e.g. methionine and tryptophan), therefore these amino acids are excluded from the analysis. For these positions, it is unknown if the conservation is due to an amino acid or DNA/RNA constraint or by pure chance. cRegions will not calculate metrics for these positions. In the work done by [@ref-17] these positions were masked to ensure that they were not flagged as low scoring and included them into the moving average. Databases often contain a redundant set of sequences. Therefore, it is very important to include phylogenetic weighting in the analysis. A redundant and biased dataset or a dataset with low variability will affect predictions and therefore may cause false positive signals. Henikoff position-based sequence weighting was used to compensate for the over-representation of similar sequences or taxa in the codon alignment ([@ref-18]). Tree-based weighting methods were excluded due to the uncertain root location, which may give lower weights to sequences close to the root, causing distantly related sequences to be down-weighted. The cRegions algorithm calculates weights for each sequence in the codon alignment, thus including variance at the nucleic acid level. Weighting was not implemented in the original publication of the method ([@ref-33]). Synplot2 also uses sequence weighting. However, it should be noted, phylogenetic weighting only affects the results if the initial set of sequences was biased. Cis-acting elements may be longer than a single codon, for example, dual-coding regions, thus the possibility to calculate a single metric over consecutive codons may be preferred. Sliding window approach is used in synplot2, FRESCo, and method described by [@ref-17]. The method used in [@ref-17] calculated MDP score over a sliding window of 10 codons and minimum window size in the synplot2 web interface is three at (*n* = 1). However, cis-acting sequences shorter than three codons (e.g. canonical splice acceptor site in Mammalia CAG\|G) may be masked by adjacent low scoring areas. Thus, in contrast to them, cRegions allows also single-codon resolution. We would like to note that synplot2 provides single-codon scores in an output text file and these can be used for analysis of regions shorter than three codons. In this study, we applied cRegions to the non-structural and structural polyprotein of alphaviruses as an example. The final dataset contained 24 sequences (see Materials and Methods). The diversity and the number of sequences were sufficient in our analysis to detect the majority of known cis-acting elements in alphaviruses. Several alphaviruses contain an in-frame termination codon and use termination read-through to produce the p1234 non-structural polyprotein ([@ref-45]; [@ref-23]; [@ref-27]). Dataset used in this work includes nine cases of known termination read-throughs: WHAV, AURAV, EILV, BEBV, NDUV, BFV, MADV, VEEV and SESV. However, ONNV, CHIKV, SFV, SPDV and SDV do not have a nonsense codon (at least in reference genomes). Inconsistency between the protein and nucleotide sequences will display warnings, although, the calculation will not be terminated. The presence of a 3′ stem-loop RNA secondary structure immediately adjacent to the stop codon has been reported to influence read-through ([@ref-15]). The leaky stop codons in alphaviruses have the next codon CGG or CTA as expected in type II read-through motif. It has been proposed that in most cases of read-through in this class also involve a 3′ RNA structure---often comprising an extended stem-loop structure beginning around eight nt 3′ of the stop codon ([@ref-28]; [@ref-13]). The cRegions tool was able to detect one significant signal inside the double helix part of the stem-loop and one inside the unpaired loop of the stem-loop. However, synplot2 was able to detect a much larger region of the stem-loop, which shows that synplot2 is more suitable in some situations ([Fig. 1B](#fig-1){ref-type="fig"}; [Figs. S4C](#supp-5){ref-type="supplementary-material"} and [S5C](#supp-6){ref-type="supplementary-material"}). Both programs: cRegions and synplot2 are able to detect a signal, even if the alignment is not perfect. In the codon alignment of the structural polyprotein, UUUUUUA motif was misaligned in two sequences (SPDV and SDV). We have shown that cRegions is capable of detecting different cis-elements. However, the method has multiple prerequisites and limitations: Protein-coding sequences must have diverged. Thus, sufficient evolutionary time is needed for substitutions to occur in homologous genes in different species/isolates.Only those embedded element in a coding-sequence can be detected which are or have been under selection.This method is not applicable to neutrally evolving genes. In the case of neutrally evolving genes, we recommend using synplot2 or other similar solutions.Cis-acting sequences must be conserved in respect to amino acid sequences.It is impossible to assess conservation at the nucleic acid level if an amino acid is encoded by a single codon (e.g. methionine and tryptophan).Long dual-coding areas or abundant rare codons will affect codon usage estimation.A low number of sequences may reduce the signal to noise ratio.Bad alignment quality, especially near large gaps may affect the results. Therefore, usage of different alignment methods or manual correction of the alignment is recommended. We would like to note that cRegions is not restricted to viral genes, but the method can be applied to any set of diverse set of protein-coding sequence if the prerequisites are fulfilled. Depending on the number and divergence of the protein and nucleic acid sequences, the size of the region and other conditions, different approaches like the synplot2, FRESCo or the method developed by [@ref-17] may have different sensitivity. Therefore, we advise using different methods side by side to find all putative cis-elements. Also, it should be noted, that any analysis depends on the quality of the input data and even statistically insignificant signals might be biologically very interesting. Conclusion ========== Evolutionary conserved embedded functional elements within an open reading frame are often overlooked as they are difficult to detect without specialised bioinformatics tools ([@ref-17]; [@ref-39]; [@ref-13]; [@ref-40]). In this work, we described a web tool called the cRegions. It is built for detecting embedded cis-acting elements from diverged protein-coding sequences. The algorithm behind cRegions compares observed nucleotide (codon) frequencies to preferential use of synonymous codons. Observed and predicted values are compared on three different metrics. The results can be displayed at a single-nucleotide resolution. Our method is able to find different cis-acting elements like splice sites, stem-loops, overlapping reading frames, internal promoters and ribosome frameshifting signals in DNA and RNA viruses. Web tools like the cRegions and the synplot2 are important for finding functional embedded elements in coding sequences and are easy to use for non-bioinformaticians. The cRegions web tool is available at <http://bioinfo.ut.ee/cRegions/> and source code is available in GitHub repository at <https://github.com/bioinfo-ut/cRegions>. Supplemental Information ======================== 10.7717/peerj.6176/supp-1 ###### Alphaviruses dataset. 'New World' alphaviruses are marked with asterisk and 'SFV Complex' alphaviruses are written in bold. ###### Click here for additional data file. 10.7717/peerj.6176/supp-2 ###### Codon usage bias is calculated over all positions in all sequences in the codon alignment. \(A\) Henikoff position-based sequence weights are calculated for each sequence based on the codon alignment. (B) Glutamic acid is encoded by two codons, therefore, the observed proportion for GAA is 0.625 and for GAG 0.375. (C) Henikoff position-based sequence weights are used to compensate for the over-representation of well-sequenced taxa in the multiple sequence alignment. The proportion for GAA and GAG are based on sequence weights. ###### Click here for additional data file. 10.7717/peerj.6176/supp-3 ###### Predicted nucleotide proportions are adjusted based on sequence weights. \(A\) Example dataset consists of 7 sequences. Five sequences are very similar; therefore, they have a low weight. Two sequences out of seven are different, therefore, having a high weight. (B) Glutamic acid is encoded by only two codons: GAA and GAG. Only the third position of a codon varies, therefore, including information. The observed proportion for the A nucleotide in the 3rd position of the glutamic acid codon is 0.714 and for G it is 0.286. (C) Predictions based on codon usage give us proportion for A is 0.55 and for G 0.45. By comparing observed and predicted proportions we will get a signal as there is a difference. However, it is a false positive signal due to a biased dataset. (D) Adjusting predictions with sequence weights, we can account for the over-representation of similar sequences. Only nucleotides that were observed are adjusted. ###### Click here for additional data file. 10.7717/peerj.6176/supp-4 ###### Difference between "allowed gaps" and "skip gaps" parameter. Allowed gaps parameter is a threshold (percentage of gaps in one column) if a metrics (RMSD, MAXDIF, CHISQ) should be calculated to a certain position. Metrics are not calculated at positions (red crosses) where the percentage of gaps in one column exceeds the threshold (percentage of allowed gaps). By default, a position (column) must have less than 20% of gaps. Skip gaps parameter is only used in sliding window mode. It is a threshold for skipping columns during sliding window mode instead of terminating the calculation as in the first three positions (NA). The threshold is used to avoid sliding window calculation termination while encountering insertions in a few sequences. By default, if the proportion of gaps in a position exceeds 90% (in other words, when insertion occurs in less than 10%) then this position is skipped (transparent red column) and next position is included to the window. ###### Click here for additional data file. 10.7717/peerj.6176/supp-5 ###### Synplot2 analysis of non-structural polyproteins of alphaviruses using 3-codon sliding window (*n* = 1). \(A\) Non-structural polyprotein alignment of all 24 Alphaviruses in our dataset. (B) Non-structural polyproteins of 'New world' Alphaviruses. (C) Non-structural polyproteins of 'SFV Complex' Alphaviruses. Non-structural polyprotein sequences were aligned with MAFFT using the default settings at <http://www.ebi.ac.uk/Tools/msa/mafft/>. Codon alignment was generated with pal2nal (<http://www.bork.embl.de/pal2nal/>). ###### Click here for additional data file. 10.7717/peerj.6176/supp-6 ###### Synplot2 analysis of non-structural polyproteins of alphaviruses using 15-codon sliding window (n = 7). Same settings were used in the synplot2 publication ([@ref-13]). (A) Non-structural polyprotein alignment of all 24 Alphaviruses in our dataset. (B) Non-structural polyproteins of 'New world' Alphaviruses. (C) Non-structural polyproteins of 'SFV Complex' Alphaviruses. Non-structural polyprotein sequences were aligned with MAFFT using the default settings at <http://www.ebi.ac.uk/Tools/msa/mafft/>. Codon alignment was generated with pal2nal (<http://www.bork.embl.de/pal2nal/>). ###### Click here for additional data file. 10.7717/peerj.6176/supp-7 ###### Synplot2 analysis of structural polyproteins of alphaviruses. Significant signal was detected between codons 800--1000, which corresponds to a known UUUUUUA motif. The y-axis on the upper part of the synplot2 figure shows the p-value and the lower part shows obs/exp rato. Structural polyprotein sequences were aligned with MAFFT using the default settings at <http://www.ebi.ac.uk/Tools/msa/mafft/>. Codon alignment was generated with pal2nal (<http://www.bork.embl.de/pal2nal/>). ###### Click here for additional data file. 10.7717/peerj.6176/supp-8 ###### cRegions and Synplot2 analysis of structural polyproteins of 94 VEEV alphaviruses. \(A\) Two known regions of RNA secondary structures were detected with cRegions using sliding window size 18. (B) Zoomed region of the conserved stem-loops. Locations of the signals in the VEEV genome are also provided (Kim et al., 2014). (C) Zoomed region of the stem-loops adjacent to stop codon. Locations of the signals in the VEEV genome are also provided ([@ref-15]). (D) Similar to the work done by Kim et al. sliding window size 25 was used with synplot2 (n = 12). Structural polyprotein sequences were aligned with MAFFT using the default settings at <http://www.ebi.ac.uk/Tools/msa/mafft/>. Codon alignment was generated with pal2nal (<http://www.bork.embl.de/pal2nal/>). ###### Click here for additional data file. 10.7717/peerj.6176/supp-9 ###### The number of signals in randomly mutated sequences. A random 3000nt long protein-coding sequence was created with SMS v2 tool (<http://www.bioinformatics.org/sms2/random_coding_dna.html>) ([@ref-44]). A different number of random mutations (25--2000) were introduced into the original randomly generated protein-coding sequence 9 times with the SMS v2 mutate tool (<http://www.bioinformatics.org/sms2/mutate_dna.html>) ([@ref-44]). Each simulated dataset consisted of 10 sequences each having the same number of mutations. Orange circles show the number of signals over the threshold without a threshold correction and using window size 1. Red circles show the number of signals with sliding window size 2. The deep blue-green (teal) shows the number of signals over the threshold when applying threshold correction. ###### Click here for additional data file. We would like to thank Märt Roosaare, Mihkel Vaher and Siim Puustusmaa for giving feedback on paper during its composition. Authors also thank prof. Andres Merits for sharing expert knowledge on alphaviruses. Additional Information and Declarations ======================================= The authors declare that they have no competing interests. [Mikk Puustusmaa](#author-1){ref-type="contrib"} conceived and designed the experiments, performed the experiments, analysed the data, contributed reagents/materials/analysis tools, prepared figures and/or tables, authored or reviewed drafts of the paper, approved the final draft, wrote the python code and the web tool in php and javascript. [Aare Abroi](#author-2){ref-type="contrib"} conceived and designed the experiments, performed the experiments, analysed the data, authored or reviewed drafts of the paper, approved the final draft. The following information was supplied regarding data availability: GitHub repository: <https://github.com/bioinfo-ut/cRegions>

Background {#Sec1} ========== For bone reconstruction such as skull base reconstruction after tumor resection, an interconnected porous structure is critical to mimicking the bone extracellular matrix \[[@CR1]--[@CR8]\]. The pore size, porosity, and pore interconnectivity of porous bone scaffolds determine their performance in functions such as cell attachment and nutrient diffusion, which enhances soft tissue and bone ingrowth and eventually resistance to infection or deformation. Moreover, mechanical stability is mandatory for the mechanical support that is required during the repair and regeneration of damaged or degenerated bone \[[@CR7], [@CR9]\]. Porous scaffolds for biomedical applications have been successfully fabricated via the sol-gel process \[[@CR10]\], salt-leaching method \[[@CR8], [@CR11]--[@CR13]\], electrospinning \[[@CR14]--[@CR17]\], and microsphere-sintering technique \[[@CR18], [@CR19]\]. However, the lack of mechanical strength of the porous materials can cause instability of the pore structures and hence limit their biomedical applications, and thus the choice of scaffold material is crucial. The performance of porous scaffolds can be optimized by controlling their surface chemistry, because the interface between the porous scaffolds and cells determines the cellular behavior, such as cell adhesion, spreading, and proliferation \[[@CR6]\]. Collagen is the main organic component of bones, and is hence a promising candidate material for the surface modification of porous scaffolds by promoting cell attachment and chemotactic responses \[[@CR20]\]. High-density polyethylene (HDPE) shows excellent mechanical properties, and it has been widely used as an implant material for bone reconstruction \[[@CR18], [@CR21], [@CR22]\]. Medpor® (Porex Technologies Co., USA) is one such porous HDPE scaffold for bone tissue engineering, used as an alloplastic material for craniofacial reconstruction \[[@CR23], [@CR24]\]. However, HDPE is inert and hydrophobic, and exhibits poor reactivity with biomaterials such as collagen. Several efforts have been made to improve the reactivity of PE for biomedical applications. The grafting of acrylic acid onto the PE film was conducted to improve protein immobilization and cell seeding \[[@CR25]\]. It was also reported that plasma treatment effectively provides HDPE with a hydrophilic surface, which results in better reactivity with bioactive molecules \[[@CR26]\]. The carboxylic acid groups of poly(ethylene-co-acrylic acid) (PEAA) make it an outstanding candidate to support the reactivity with collagen. Besides this, PEAA is mechanically stable, owing to the strong hydrogen bonds in its carboxylic acid groups, which can be effective crosslinkers between polymer chains. In this study, the composite of HDPE and PEAA was chosen as scaffold material for cranial reconstruction owing to the high mechanical stability of HDPE and the high reactivity of PEAA with collagen. Before collagen grafting, the porous structure was prepared using a salt-leaching method, which can provide the proper pore size and high porosity. Osteoblast cells were then cultured on the collagen-grafted porous HDPE/PEAA scaffold, and the cell adhesion, proliferation, and differentiation were measured to investigate their bone tissue compatibility. Porous scaffolds of HDPE and HDPE/PEAA without collagen grafting were also fabricated and studied as controls. Methods {#Sec2} ======= Fabrication of collagen-grafted porous scaffolds {#Sec3} ------------------------------------------------ Porous HDPE/PEAA scaffolds were fabricated by using a salt-leaching method^10^. HDPE (Mw 85,000, Mn 13,500; Korea Petrochemical Industrial Co., Korea) and PEAA (acrylic acid 20 wt%; Sigma-Aldrich Co., USA) beads (w/w = 3:1) were mixed with sodium chloride (HDPE/PEAA:NaCl = 1:9) with a particle size of 200--500 μm, using a melt mixing machine (Brabender, Plasti-Corder Co.) at 160 °C. Then, the mixture was cast in a circular mold (diameter 13 mm, thickness 1.3 mm) using a heat press machine (Yoochang Co., Korea). The resulting HDPE/PEAA/NaCl composite was immersed in distilled water to leach out the NaCl, leaving pores in the composite. The salt-free porous HDPE/PEAA was washed with distilled water and air dried. For obtaining high reactivity between the scaffold and collagen, L-lysine was grafted onto the scaffold surface to improve the affinity of the carboxyl groups to the amine groups in collagen. Before the L-lysine grafting, the carboxylic groups on the HDPE/PEAA scaffold were activated by immersing the scaffold into a 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (0.25 wt%; Sigma-Aldrich Co., USA) and N-hydroxysuccinimide (0.25 wt%; Sigma-Aldrich Co., USA) aqueous solution for 6 h at room temperature. Afterwards, it was immersed in 3 wt% L-lysine aqueous solution with gentle stirring. The carboxyl groups of L-lysine, attached to the scaffold surface, were also activated by this same method. Collagen-grafted HDPE/PEAA (HDPE/PEAA/Col) was produced by immersing the HDPE/PEAA scaffold in 3 wt% collagen solution (in distilled water containing acetic acid, pH 4.3) for 6 h with gentle stirring, and then it was washed with distilled water and dried. Characterization of the scaffolds {#Sec4} --------------------------------- The surface morphology of the porous HDPE, HDPE/PEAA, and HDPE/PEAA/Col scaffolds was observed under a field emission scanning electron microscope (FE-SEM S4300; Hitachi, Japan) after sputter-coating with platinum. The chemical bonds and elemental composition were characterized by Fourier transform infrared (FT-IR; Mattson, Galaxy 7020A) spectroscopy and electron spectroscopy for chemical analysis (ESCA; ESCA LAB VIG microtech, Mt 500/1, and so forth, East Grinstead, UK), respectively. Tensile properties were measured via a universal testing machine (Instron, model 4465) with a Zwick Roell tensile tester equipped with a 1 kgf load cell, at 25 °C with an extension speed of 10 mm/min. The tensile strength and Young's modulus measure of each sample were calculated from the averages of 10 specimens. The porosity of the porous scaffolds was determined by using a mercury intrusion porosimeter (AutoPore IV 9520; Micromeritics Co., USA). The advancing and retreating contact angles of mercury were taken to be 140° and the surface tension was taken as 0.480 N/m (480 dynes/cm). Cell behavior {#Sec5} ------------- Cell behavior was observed by culturing osteoblast cells (5 × 10^4^ cells/mL; MC3T3-E1, ATCC) on the scaffolds, at 37 °C in a humidified atmosphere with 5 % CO~2~, in Dulbecco's modified Eagle's medium (Gibco, USA) supplemented with 10 % fetal bovine serum (Gibco, USA) and 1 % penicillin G-streptomycin (Gibco, USA). After both 1 and 2 days of incubation, calcein-AM (1 mM in dimethyl sulfoxide) and propidium iodide (1.5 mM in distilled water) solutions were added and the scaffolds were left standing for 15 min. The fluorescence images were visualized with a confocal laser scanning microscope (CLSM, Carl Zeiss, LSM 700, Germany). To evaluate the cytoskeletal organization of cells on the porous scaffolds, double staining was performed. After 3 days of incubating the cell solution with the scaffold samples, the cells were fixed with 4 % paraformaldehyde in PBS and permeabilized with 0.1 % Triton X-100 in PBS for 15 min. The samples were then incubated for 30 min in a PBS containing 1 % bovine serum albumin, followed by the addition of tetramethylrhodamine-5-isothiocyanate (TRITC)-conjugated phalloidin (Millipore, Cat. No. 90228). After 1 h, the samples were incubated with 4,6-diamidino-2-phenylindole (DAPI) (Millipore, Cat. No. 90229) for 5 min. The fluorescence images were taken with a confocal laser scanning microscope (CLSM 700). The cell viability and proliferation on the porous scaffolds were evaluated using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay and enzyme-linked immunosorbent assay (ELISA). For the MTT assay, the scaffold samples were immersed in 50 μL of MTT solution (5 mg/mL in PBS) for 4 h. After removing the solution, the water-insoluble formazan product was dissolved in 0.04 N HCl-isopropanol in the dark. ELISA was performed using 5-bromo-2-deoxyuridine (BrdU), which is incorporated during DNA synthesis in the cells. The BrdU ELISA was conducted according to the manufacturer's instructions (Roche Molecular Biochemicals, Germany). The absorbance was measured at 570 nm, using a kinetic microplate reader (EL × 800; Bio-T Instruments, Inc., Highland Park, USA). Cell differentiation was tested by several cell staining methods, using alizarin red S, von Kossa, and alkaline phosphatase (ALP) staining. The osteoblast cells (5 × 10^4^ cells/mL) were cultured for 15 days on the three porous scaffolds and then fixed using 10 % formaldehyde. For alizarin red S staining, the samples were treated with an alizarin red S solution and incubated for 20 min. For the von Kossa assay, the fixed samples were treated with 5 % AgNO~3~ solution for 20 min under ultraviolet radiation, followed by the addition of 5 % Na~2~S~2~O~3~ solution for 5 min. ALP staining was done by a standard procedure according to the manufacturer's instructions (Alkaline phosphatase, Leukocyte, Procedure No. 86; Sigma-Aldrich, USA), using an alkaline dye mixture (1 mL of sodium nitrate, 1 mL of FBB-alkaline solution, 1 mL of naphthol AS-BI alkaline solution, and 1 mL of deionized water) and a neutral red buffered solution for counterstaining \[[@CR27]\]. The digital images of the stained cultures were obtained with a digital camera (Canon A2000 IS, Japan) and an optical microscope (Carl Zeiss, Germany). Data analysis {#Sec6} ------------- The results are displayed as the mean ± standard deviation. The statistical significance of differences between the scaffolds was determined by a Student's two-tailed *t* test. Scheffe's method was used for multiple comparison tests at a level of 95 %. Results and discussion {#Sec7} ====================== Pore structure {#Sec8} -------------- The surface morphology of the porous HDPE, HDPE/PEAA, and HDPE/PEAA/Col scaffolds was observed by scanning electron microscopy. As shown in Fig. [1](#Fig1){ref-type="fig"}, interconnected pores were successfully formed in the scaffolds, and their pore sizes ranged between several microns and a few hundred microns. It is also seen that the collagen-grafted scaffold in Fig. [1(c)](#Fig1){ref-type="fig"} had slightly smaller pores than those without collagen grafting in Fig. [1(a)](#Fig1){ref-type="fig"} and [(b)](#Fig1){ref-type="fig"}.Fig. 1Surface morphologies of the porous HDPE (**a**), HDPE/PEAA (**b**) and HDPE/PEAA/Col (**c**) scaffolds The intrusion volume and porosity were measured to investigate the change of pore size by the scaffold materials and collagen grafting, and the results are shown in Table [1](#Tab1){ref-type="table"}. The porosity of the HDPE/PEAA scaffold was similar to that of HDPE, which was approximately 65 %. However, when collagen was introduced to the surface of the HDPE/PEAA scaffold the porosity decreased by 5 %, likely due to the high molecular weight of collagen.Table 1Intrusion volume and porosity of the porous HDPE, HDPE/PEAA and HDPE/PEAA/Col scaffoldsSubstrateIntrusion volume (mL/g)Porosity(%)HDPE2.0865.21HDPE/PEAA2.3166.75HDPE/PEAA/Collagen1.8859.28Standard deviation is within 10 % The pore characteristics are also key factors that affect the performance of porous scaffolds in bone reconstruction because the pore size and porosity of scaffolds affect the diffusion of nutrients and osteoblast cell attachment, migration, proliferation, and differentiation, which are vital for bone formation. Additionally, a porous surface is known to drive mechanical stability at the interface between the implant materials and the surrounding tissue \[[@CR28]\]. Even though there is disagreement about the optimum pore size of porous scaffolds, it is generally agreed upon that the pore size and porosity play essential roles in their compatibility to cells such as osteoblasts, and pores of a few hundred microns are highly required \[[@CR3]--[@CR5], [@CR8]\]. Therefore, on the basis of the results of Fig. [1](#Fig1){ref-type="fig"} and Table [1](#Tab1){ref-type="table"}, it can be concluded that the pore size of the HDPE-based scaffolds prepared by the salt-leaching method is appropriate for porous bone scaffolds. Surface chemistry {#Sec9} ----------------- FT-IR spectra of the HDPE, HDPE/PEAA, and HDPE/PEAA/Col scaffolds and of collagen are shown in Fig. [2](#Fig2){ref-type="fig"}. Both the HDPE and HDPE/PEAA spectra exhibited bands at 2849 and 2918 cm^−1^, assigned to hydrocarbons (CH, CH~2~). For the HDPE/PEAA scaffold (Fig. [2b](#Fig2){ref-type="fig"}), the vibrational band at 1700 cm^−1^ based on C = O was observed, but it did not appear for the HDPE scaffold (Fig. [2a](#Fig2){ref-type="fig"}), which proves that PEAA was well incorporated into the HDPE/PEAA scaffold. It is also seen that the HDPE/PEAA/Col scaffold (Fig. [2d](#Fig2){ref-type="fig"}) displayed the characteristic collagen peaks at 1661 and 1553 cm^−1^, assigned to the stretching vibration of the carbonyl group (C = O) within amide I (--**CO**NH--) and the coupling of N-H bending and C-N stretching of amide II (--CO**NH**--), respectively.Fig. 2ATR-FTIR spectra of (a) HDPE (---), (b) HDPE/PEAA (---), (c) Collagen (---), and (d) HDPE/PEAA/Col (---) Collagen grafting on the HDPE/PEAA scaffold was further confirmed by ESCA, and the elemental compositions of the HDPE, HDPE/PEAA, and HDPE/PEAA/Col scaffolds are shown in Table [2](#Tab2){ref-type="table"}. The atomic percentage of nitrogen was significantly increased on the surface of the HDPE/PEAA scaffold modified with L-lysine and subsequently with collagen. According to the FT-IR spectra and ESCA results, it can be confirmed that collagen grafting was successfully conducted on the porous HDPE/PEAA scaffold.Table 2Chemical composition of porous scaffolds calculated from their survey scan spectraSubstrateAtomic %C 1 sO 1 sN 1 sSi 2pCI 2pNa 1 sHDPE93.55.3\<0.11.2\--HDPE/PEAA83.211.81.02.40.30.5HDPE/PEAA/Collagen81.811.15.31.20.50.1 Tensile properties {#Sec10} ------------------ Figure [3](#Fig3){ref-type="fig"} represents the tensile strength and Young's modulus measures of the porous HDPE, HDPE/PEAA, and HDPE/PEAA/Col scaffolds. The porous HDPE scaffold showed higher strength and modulus values, owing to the high mechanical stability of HDPE. When PEAA was incorporated into the HDPE scaffold, its Young's modulus measure decreased significantly, while the tensile strength was slightly lowered. It is also shown that grafting collagen on the scaffolds does not affect their tensile properties. PEAA is widely used as a compatibilizer for polymer blends or composites because of its functionality. Its segment of acrylic acid provides unique properties, such as polarity, crosslink ability, and adhesion to polar substrates, as well as low softening and melting points \[[@CR29]\]. Kim et al. reported the addition of PEAA to polyethylene terephthalate/HDPE blends, which effectively improved their mechanical properties such as flexural yield strain and impact strength \[[@CR30]\]. PEAA was also reported as a compatibilizer of polylactic acid/recycled low-density polyethylene blends, enhancing the tensile properties of the composites \[[@CR31]\].Fig. 3Tensile properties of the porous HDPE, HDPE/PEAA scaffolds Cell viability and proliferation {#Sec11} -------------------------------- Cell behaviors on the HDPE, HDPE/PEAA, and HDPE/PEAA/Col scaffolds were investigated using several methods to examine their biocompatibility. First, the adhesion and cytotoxic effects of the three scaffolds were observed by using cell staining after a 1-day and 3-day incubation period. Figure [4](#Fig4){ref-type="fig"} shows the morphologies of osteoblast cells on the surface of the scaffolds. Calcein-AM, a highly lipophilic dye that can easily penetrate the cell membrane, interacts with cytosolic esterase in viable cells to result in green fluorescence. All the cells in Fig. [4](#Fig4){ref-type="fig"} exhibited strong green fluorescence, indicating the good viability of osteoblast cells. On the porous HDPE scaffold, only a few cells had adhered, and their growth appeared to be somewhat slow (Fig. [4a](#Fig4){ref-type="fig"} and [d](#Fig4){ref-type="fig"}). On the other hand, the HDPE/PEAA scaffold (Fig. [4b](#Fig4){ref-type="fig"} and [e](#Fig4){ref-type="fig"}) displayed slightly better cell adhesion and cell spreading, and these properties were further enhanced when collagen was introduced to the surface of the HDPE/PEAA scaffold (Fig. [4c](#Fig4){ref-type="fig"} and [f](#Fig4){ref-type="fig"}).Fig. 4Confocal laser scanning microscope images of calcein-AM dye-stained osteoblast cells cultured on the porous HDPE, HDPE/PEAA, and HDPE/PEAA/Col scaffolds The nucleus and actin of the osteoblast cells cultured on the three different scaffolds were observed by double staining to examine the cytoskeleton organization. As shown in Fig. [5](#Fig5){ref-type="fig"}, actin was stained with TRITC (red), whereas the nucleus was stained with DAPI (blue). It appeared that the cells cultured on the HDPE scaffold (Fig. [5a](#Fig5){ref-type="fig"}) expressed actin filaments slightly with a small number of cells. However, the cytoskeletons of the cells on the HDPE/PEAA scaffold (Fig. [5b](#Fig5){ref-type="fig"}) seemed more organized. The HDPE/PEAA/Col scaffold (Fig. [5c](#Fig5){ref-type="fig"}) showed a large number of cells cultured on the substrate that were clearly organized with stretched actin and stress fibers.Fig. 5Confocal laser scanning micrographs (actin in red, nucleus in blue) of osteoblast cells cultured on the porous HDPE (**a**), HDPE/PEAA (**b**), and HDPE/PEAA/Col (**c**) scaffolds Figure [6](#Fig6){ref-type="fig"} shows the MTT and BrdU assay results after a 3-day incubation of osteoblast cells on the three scaffolds. The HDPE/PEAA/Col scaffold showed significantly higher cell viability and proliferation (*p* \< 0.03 for MTT and *p* \< 0.02 for BrdU assays) than the HDPE and HDPE/PEAA scaffolds, suggesting that collagen plays an important role in cell growth and metabolism. Not only were a large number of osteoblasts alive, but they also proliferated actively on the collagen-containing biocompatible scaffold. Collagen has been mainly used to improve biocompatibility via surface modification for biomedical applications \[[@CR32]--[@CR34]\]. It was reported that collagen grafting successfully promoted cell proliferation by cell growth and cell division on both organic and inorganic materials.Fig. 6MTT (**a**) and BrdU (**b**) assays of osteoblast cells cultured on the porous HDPE, HDPE/PEAA, and HDPE/PEAA/Col scaffolds Cell differentiation {#Sec12} -------------------- Osteoblast cell differentiation is one of the most important parameters for confirming the osteogenesis of osteoblast cells. Alizarin red S, von Kossa, and ALP staining methods have been frequently utilized to characterize the interface between calcified bone tissue and the implant surface \[[@CR35]--[@CR37]\]. For alizarin red S staining, the calcification area in the cells is stained red from the formation of a calcium/alizarin red S complex. Figure [7](#Fig7){ref-type="fig"} shows the result of alizarin red S staining of osteoblast cells on the three porous scaffolds. It can be seen that osteoblasts on the HDPE, HDPE/PEAA, and HDPE/PEAA/Col scaffold were stained in red, with the HDPE/PEAA/Col scaffold showing the most intense dark red color, resulting from accelerated cell differentiation by collagen grafting.Fig. 7Alizarin red S staining of osteoblast cells cultured on the porous HDPE (**a**), HDPE/PEAA (**b**), and HDPE/PEAA/Col (**c**) scaffolds The von Kossa stain is also one of the ways to confirm mineralization in cell cultures by detecting phosphate in the calcification area, which is stained as a black spot. In Fig. [8](#Fig8){ref-type="fig"}, von Kossa staining images of osteoblast cells cultured on the porous scaffolds are displayed. Osteoblast cells on the HDPE/PEAA/Col scaffold showed the most intensive dark spots among the three scaffolds. It was therefore confirmed that collagen grafting is effective in triggering or accelerating osteoblast cell differentiation, which matches well with the alizarin red S staining results in Fig. [7](#Fig7){ref-type="fig"}. The HDPE/PEAA scaffold presented better cell differentiation (Figs. [7b](#Fig7){ref-type="fig"} and [8b](#Fig8){ref-type="fig"}) than the HDPE scaffold (Figs. [7a](#Fig7){ref-type="fig"} and [8a](#Fig8){ref-type="fig"}).Fig. 8Von Kossa assay of osteoblast cells cultured on the porous HDPE (**a**), HDPE/PEAA (**b**), and HDPE/PEAA/Col (**c**) scaffolds The differentiation of osteoblast cells was further proven by the synthesis of ALP in the cells, which appears as a blue spot (Fig. [9](#Fig9){ref-type="fig"}). ALP is an enzyme produced by osteoblast activity such as bone generation. Therefore, the amount of ALP synthesized can represent the vitality of osteoblast cells. The amount of ALP synthesized by the cells cultured on the HDPE/PEAA/Col scaffold (Fig. [9c](#Fig9){ref-type="fig"}) was higher than those on the HDPE and HDPE/PEAA scaffolds (Fig. [9a and b](#Fig9){ref-type="fig"}, respectively). Thus, the osteoblasts on the collagen-grafted scaffold had active ALP synthesis, indicating the scaffold's potential for bone generation applications.Fig. 9ALP activity staining of osteoblast cells cultured on the porous HDPE (**a**), HDPE/PEAA (**b**), and HDPE/PEAA/Col (**c**) scaffolds Conclusions {#Sec13} =========== For bone reconstruction, porous scaffolds were fabricated using HDPE/PEAA composites via a salt-leaching method. The surface of the porous HDPE/PEAA scaffold was modified using collagen to enhance bone tissue compatibility. The surface modification was confirmed via FT-IR spectroscopy and ESCA by detecting the nitrogen component in collagen. It was shown that the pore size and porosity are suitable for osteoblast attachment, as confirmed by the surface images and porosity results. The cell viability and proliferation were measured by MTT and BrdU assays, with results showing that the collagen-grafted HDPE/PEAA surface is favorable for the adhesion and proliferation of osteoblast cells. Furthermore, cell differentiation was studied using several staining methods, where it was seen that osteoblasts on the collagen-grafted scaffold have outstanding differentiation. It is concluded that collagen grafting on the porous HDPE/PEAA scaffold effectively improves its biocompatibility and potential use as a bone scaffold. This research was supported by Kyungpook National University Research Fund for the year 2013. Funding {#FPar1} ======= This research was supported by Kyungpook National University Research Fund for the year 2013. Availability of data and material {#FPar2} ================================= All data are available on Journal portals in submitted manuscript. No other supporting files/data are needed along with this submission. Authors' contributions {#FPar3} ====================== CSK and IKK designed the experiments. KHJ helped in writing the manuscript. HK and CBK have conceived the ideas of this study, and participated in its design. All authors read and approved the final manuscript. Competing interests {#FPar4} =================== The authors declare that they have no competing interests. Consent for publication {#FPar5} ======================= The manuscript has been submitted with the consent of all authors and data of any other person not included. Ethics approval and consent to participate {#FPar6} ========================================== Manuscript does not include human ethics value. Hence no consent is needed.

1.. Introduction {#s1} ================ The lead chalcogenides PbTe, PbSe and PbS have undergone much detailed study as thermoelectrics in recent years. These are all narrow band gap semiconductors, forming in the cubic rocksalt structure, that can be readily doped both p-type and n-type. Although in earlier years their maximum thermoelectric figure-of-merit ZT (ZT is defined as S^2^*σ*T/*κ*, where S is the Seebeck coefficient, *σ* the electrical conductivity, T the temperature and *κ* the thermal conductivity \[[@C1]\]) was thought to be unity or less, recent work has shown a spectacular ZT improvement when full optimization of both carrier concentration and nanostructuring is performed. p-type PbTe, in particular, has recently been found to have a ZT value exceeding 2 at high temperature \[[@C2]\]. Research on these materials continues to be quite active, and it is possible that still higher ZT values will be attained for these materials in the near future. In this paper we will discuss three narrow band gap (based on our calculations) alkaline earth lead compounds Ca~2~Pb, Sr~2~Pb and Ba~2~Pb that, while having undergone scant experimental study, may also prove to be excellent thermoelectrics at a range of temperatures from room temperature to 1000 K, if optimized in a manner similar to the lead chalcogenides. As with the lead chalcogenides, these materials are isostructural and isoelectronic; they form in an orthorhombic structure (space group 62, *Pnma*) as depicted in figure [1](#F0001){ref-type="fig"}. Lattice constants for these materials are presented in table [1](#TB1){ref-type="table"}. As the figure suggests, this is a layered structure and we therefore expect significant electronic structure anisotropy, to be discussed later. Nearest-neighbor distances in these materials exceed 3 Å, suggestive of weak bonding and hence soft phonons, favorable for thermoelectric performance (see section [4](#s4){ref-type="sec"}). Note also that the melting points for all three compounds are significantly above 1000 K, taking the values for Ca~2~Pb of 1476 K, Sr~2~Pb 1428 K and Ba~2~Pb 1201 K. This is similar to the lead chalcogenides and is favorable for high temperature performance. {#F0001} ###### Lattice constants (in Å) for the alkaline earth plumbides and stannides studied here. References as indicated. Compound *a* *b* *c* -------------------- ------- ------- -------- Ca~2~Pb \[[@C12]\] 8.072 5.100 9.647 Sr~2~Pb \[[@C13]\] 8.445 5.391 10.139 Ba~2~Pb \[[@C14]\] 8.651 5.691 10.618 Ca~2~Sn \[[@C15]\] 7.975 5.044 9.562 Sr~2~Sn \[[@C16]\] 8.402 5.378 10.078 Ba~2~Sn \[[@C14]\] 8.615 5.699 10.569 These materials also bear a certain similarity to the half-Heusler thermoelectrics which have also shown high thermoelectric performance, with a ZT of 1.5 at 700 K found for ZrHfNiSn \[[@C3]\]; for a review of these materials, see \[[@C4]\]. Similarly to the half-Heusler materials, these materials are comprised of metallic elements yet exhibit semiconducting behavior. In the Heusler materials this behavior is associated with a band gap associated with valence electron counts of 8 or 18 associated with the bonding of this compounds. In the present case, there is also an electron count rule with a gap at a valence electron count of 8. These compounds consist of highly electropositive alkaline earth elements and group IV elements, and as such it is convenient to describe the electron counting in terms of a nominal ionic Zintl type picture, e.g. Ca^2+^~2~Pb^4−^. As discussed below, the valence and conduction bands in these compounds have different character, consistent with this description. Ionic bonding generally implies soft lattices, and indeed the calculated bulk moduli (section [4](#s4){ref-type="sec"}) for these materials are rather low. These materials, like the lead chalcogenides, may represent a means of resolving one of the conundrums usually hampering thermoelectric performance: how to find materials with the usually conflicting properties of possessing soft phonons, indicating low lattice thermal conductivity, but also having a high melting point, allowing high temperature thermoelectric performance. In addition, with a substantially larger unit cell of 12 atoms, as opposed to the 2 atom unit cell of the lead chalcogenides, these materials may show particularly low thermal conductivity since a majority of the phononic specific heat will be comprised of the 33 optic modes (to be compared to the 3 optic modes of the lead chalcogenides), which generally have comparatively low group velocities and hence low thermal transport. The only experimental thermoelectric study of these materials dates back to 1961 \[[@C5]\], when Russell and Klein observed a room-temperature ZT of 0.2 in a Ca~2~Pb sample with a slight excess of calcium. Surprisingly, despite this significant ZT (probably at a substantially non-optimized doping level), no effort aimed at improving this value has been documented. Given this significant ZT, and that Ca~2~Pb has a significant band gap (estimated as approximately 0.46 eV \[[@C6]\]), it is likely that much higher ZT values are attainable at high temperature. In addition, good room-temperature performance when fully optimized is probable as well. Some theoretical results on these materials were reported in \[[@C7], [@C8]\]. As is well known, carrier mobility is an important factor affecting thermoelectric performance as it impacts the electrical conductivity, to which ZT is proportional. In this regard, there is one additional important piece of experimental data deriving from the aforementioned band-gap determination \[[@C6]\] of Ca~2~Pb. A sample exhibiting a substantial room temperature electrical conductivity of 200 Ω^−1^cm^−1^ was observed to enter the intrinsic regime (i.e. the resistivity )) at the comparatively low temperature of approximately 550 K. Since this is primarily a band structure, not scattering time, effect, it is accessible to our first principles calculations and we find for this sample, as a rough estimate, a carrier concentration of 10^18^ cm^−3^, which would suggest a very high room temperature mobility of roughly 1250 cm^2^ V^−1^ s^−1^. Given these favorable pieces of information, we chose to perform a theoretical study of this material, in the hopes of elucidating as much as possible about these materials. Sections [2](#s2){ref-type="sec"}--[4](#s4){ref-type="sec"} present the theoretical results and section [5](#s5){ref-type="sec"} a summary and conclusion. 2.. Theoretical results I: band structures, densities of states and thermopower {#s2} =============================================================================== As mentioned previously, Ca~2~Pb, Sr~2~Pb and Ba~2~Pb all form in the orthorhombic structure, and we have therefore undertaken a systematic study of each of these materials, using the linearized augmented plane-wave (LAPW) first principles code WIEN2K \[[@C9]\], within the generalized gradient approximation (GGA) of Perdew *et al* \[[@C10]\]. We have used the experimental lattice constants, as presented in table [1](#TB1){ref-type="table"}, and optimized the internal coordinates. Spin--orbit coupling was included throughout (excepting the optimization). The self-consistent charge densities were calculated, using 140 *k*-points in the irreducible wedge of the Brillouin zone, and sphere radii of 2.5--2.7 Bohr radius were used. Due to the well known underestimation of band gaps by most first principles approaches, we have here used an augmentation of the GGA known as a modified Becke--Johnson (mBJ) potential \[[@C11]\], in which a real-space, position dependent multiplicative factor is applied to the one electron potential derived from the GGA energy density functional. This approach has been shown \[[@C11]\] to give band gaps much closer to experimental values. This is an important consideration for narrow band gap semiconductors, for which high-temperature bipolar conduction can be a substantial concern and which therefore require highly accurate band gaps, as the mBJ produces. We have calculated the thermopower for each material from the converged results using Boltzmann transport theory within the 'constant scattering time approximation' (CSTA), in which the scattering time of an electron or hole is taken as independent of energy. This method has been used with much success \[[@C17]--[@C26]\] to describe the thermopower of a large number of semiconducting materials. We begin with the band structures, depicted in figure [2](#F0002){ref-type="fig"}. For all three materials both the conduction band minimum (CBM) and VBM are at the Γ point, with respective band gaps for Ca~2~Pb, Sr~2~Pb and Ba~2~Pb of 0.33, 0.38 and 0.18 eV. The Ca~2~Pb figure of 0.33 eV is somewhat lower than the resistivity-determined value of 0.46 eV, and we note that this last value derives from a single, non-optical measurement and therefore contains some uncertainty. The calculated direct band gaps for all these materials suggest, however, that band gap assessment via optical measurement should readily determine the band gaps of these materials. To our knowledge, the calculated values for Sr~2~Pb and Ba~2~Pb are the first reported for these materials. The smaller band gap of the Barium compound suggests that bipolar conduction may be more of a concern with this material, and that therefore performance for this material will potentially be optimal at lower temperatures---generally under 500 K. The other two compounds, in the absence of temperature-dependent band structure effects, should be largely free of bipolar conduction in the doping ranges where thermoelectric performance is expected to be optimum for temperatures below 800 K. ![The bandstructures of, from top/left, Ca~2~Pb, Sr~2~Pb and Ba~2~Pb, within the orthorhombic Brillouin zone. *k*-point labels from \[[@C27]\]. The energy zero is set to the valence band maximum (VBM).](TSTA11668636F02){#F0002} The band gap found for Ca~2~Pb is significantly larger than that found in the theoretical studies of Yang *et al* \[[@C8]\] and Migas *et al* \[[@C28]\] who found band gaps of 0.07 and 0.15 eV, respectively for this material. In these studies a PBE GGA functional was used, as was done here, but the mBJ potential was not, leading to lower, and potentially understated, calculated band gap values. The most significant effect of the inclusion of spin--orbit coupling is to reduce the calculated band gap, by an amount of the order of 0.1 eV. Such an effect has previously been shown to be important in properly describing the band structure and transport of the lead chalcogenides \[[@C18]\], so that its inclusion here is expected to improve the accuracy of the calculated band gaps, relative to future band gap determination experiments. Physically spin--orbit coupling is rather important in these compounds, due principally to the large atomic number of lead. In addition the relevant states of lead are p states and thus experience less electronic screening and hence a stronger nuclear charge, enhancing the spin--orbit effect, which is relativistic in origin. For all three materials, the valence band structure is rather anisotropic, with the slope near the *VBM* varying by as much as a factor of 5 or more, depending on direction. This anisotropy will be a significant consideration in the performance of the material, as (comparable to Bi~2~Te~3~) one direction can be expected to have significantly lower mobility and hence conductivity. The conduction band is slightly more isotropic but still contains significant anisotropy. We have attempted to understand the anisotropy in terms of the physical structure or bonding picture of these materials but have not been successful; for several of the compounds the high conductivity directions vary from the valence to the conduction band, confounding a simple bonding picture. The reason is probably related to the orthorhombic anisotropy of the physical structure, but no detailed explanation is readily apparent. With regards to band degeneracy (usually favorable for thermoelectric performance), four of the six band extrema contain no degeneracy, while the conduction bands for Sr~2~Pb and Ca~2~Pb contain a near-degeneracy, with two maxima of rather disparate mass less than 100 meV apart. Such a structure has previously been shown to be favorable for thermoelectric performance \[[@C29], [@C30]\], so that it is possible that the performance of n-type Sr~2~Pb and Ca~2~Pb may be enhanced by this feature. We will return to this issue later. Turning to the calculated densities of states (DOS) in figure [3](#F0003){ref-type="fig"}, we note that very near all six band edges the DOS rises rapidly, which provides a small energy scale that is generally favorable for high thermopower. All six band edges have a DOS peak of height over 10 per (unit cell eV) less than half an eV from the band edge. Interestingly, for several of the cases (see the inset figures) the bend edge contains a rather light dispersive band which is then augmented by a heavier band, generally less than 0.25 eV from the bend edge. Such a heavy band--light band combination has previously been shown \[[@C29]\] to be favorable for thermoelectric performance, as it typically allows both high carrier mobility and high thermopower. {#F0003} From the DOS plots, one sees that while the valence bands for all three compounds have substantial lead character, the conduction bands have primarily alkaline earth character. This reflects the fact that, unlike in the lead chalcogenides, where the lead is a cation, here the lead is an anion, receiving charge from the alkaline earth element. Electronegativity differences between the alkaline earth elements and lead are substantially larger than in the lead chalcogenides. Thus the band structure here, unlike in the lead chalcogenides, can be understood in terms of charge transfer from the electropositive alkaline earth elements to the lead. Since the conduction and valence bands are of different character, it is also possible they could experience different electron--phonon interactions and hence have different mobility. We also expect that alkaline earth vacancies will result in p-type doping into the lead bands, while lead vacancies would be n-type dopants. Moving to the thermopower, in figure [4](#F0004){ref-type="fig"} we present the conductivity averaged thermopower, for each of the three materials, for p-type and n-type, at the temperatures of 300, 500 and 800 K. The conductivity averaged thermopower is defined as follows, where *x*, *y* and *z* are the Cartesian axes: This expression can be easily evaluated from both the thermopower *S* and *σ*/*τ* obtained from our first principles calculations from the canonical Boltzmann transport expressions \[[@C1]\], under the assumption that the scattering time *τ* does not depend on direction. In the heavily doped regime typically of interest for thermoelectrics, scattering is dominated by electron--phonon scattering which is most typically a function of carrier concentration, but not of direction. We have chosen to plot this averaged thermopower because it is the effective thermopower that would be seen in a polycrystalline sample, as would most likely be used for applications, and because the thermopower in these materials is substantially anisotropic. This can actually be seen directly from the band structure plots in figure [2](#F0002){ref-type="fig"}, where, for example, the Γ-centered VBM between U and Z exhibits quite different dispersions in these two directions, leading to differing *σ*/*τ* and as a consequence, differing thermopower. {#F0004} We begin with Ca~2~Pb. At 300 K, Pisarenko behavior, i.e. logarithmic in carrier concentration, is observed for both p-type and n-type for carrier concentrations between 10^18^ and 10^19^ cm^−3^. This is consistent with the low carrier concentrations, in which non-degenerate behavior is expected. As one passes the 10^19^ cm^−3^ point, some curvature becomes apparent, which results both from the approach to the degenerate limit (in which for a parabolic band the thermopower is, to first order, ∼*p*^−2/3^) and also from the effect of more than one band, or band maximum. Optimal doping for thermoelectric performance at 300 K, for both p-type and n-type is expected to be in the range of 4 × 10^18^--1.5 × 10^19^ cm^−3^. At the higher temperatures of 500 and 800 K, bipolar conduction (a decrease in thermopower with decreasing carrier concentration, the opposite of the usual situation) begins to appear. At 500 K this is in a doping range significantly lighter than probable optimal doping, but at 800 K this bipolar effect begins to encroach upon the Seebeck coefficient, particularly for p-type. Optimal doping ranges are given as (at 500 K) 1--4 × 10^19^ cm^−3^ for p-type and 2--7 × 10^19^ cm^−3^ for n-type. At 800 K the corresponding ranges are 4 -- 8 × 10^19^ cm^−3^ for p-type and 6 × 10^19^--3 × 10^20^ cm^−3^ for n-type. Note that at 800 K the p-type performance (i.e. ZT) may well be degraded by the bipolar effects. Surprisingly, despite the comparatively small band gap the n-type material still attains a thermopower magnitude significantly exceeding 200 *μ*VK^−1^ so that this material may show good 1000 K n-type performance for doping levels in the range of 1--5 × 10^20^ cm^−3^. Moving to Sr~2~Pb, we observe Pisarenko behavior for both p-type and n-type at 300 and 500 K in the regime below carrier concentrations of 10^19^ cm^−3^, gradually transitioning to degenerate behavior as the carrier concentrations increases. As with the calcium compound, at 300 and 500 K the thermopower does not enter the bipolar regime until dopings far smaller than optimal. At 800 K the bipolar regime begins to appear nearer optimal doping, for p-type. For n-type, however, good performance is likely attainable at 800 K and even at 1000 K, as evidenced by the substantial regions of thermopower magnitude above 200 *μ*VK^−1^. Expected optimal doping regimes, indicated by the lines at thermopowers of ±200 and ±300 *μ* VK^−1^, are at 300 K: p-type--5 × 10^18^--2 × 10^19^ cm^−3^; n-type, 2.5 × 10^18^--10^19^ cm^−3^; 500 K, p-type--1.3 × 10^19^--6 × 10^19^ cm^−3^; n-type, 7 × 10^18^--3 × 10^19^ cm^−3^; 800 K, p-type--approximately 10^20^ cm^−3^; n-type, 2 × 10^19^--10^20^ cm^−3^. *N*-type optimal dopings at 1000 K are 6 × 10^19^--2 × 10^20^ cm^−3^. Finally we discuss Ba~2~Pb. With a calculated band gap of just 0.18 eV, we expect significant bipolar conduction at elevated temperature, and accordingly we find that the 800 K thermopower barely exceeds 200 *μ*VK^−1^ for p-type and does not reach this absolute value for n-type. High thermoelectric performance at 800 K is therefore not likely. Even at 500 K, bipolar conduction may well preclude good thermoelectric performance for n-type and could limit p-type performance as well. AT 300 K, however, the bipolar conduction is much less significant, with maximum thermopower magnitude exceeding 300 *μ*VK^−1^ for both p-type and n-type. We note also that, as Ba is the heaviest alkaline earth element in the three compounds considered in this paper, its lattice thermal conductivity is expected to be the lowest, a point underscored by our calculated bulk moduli (next section). This means that it is most probable that of the three materials studied here, Ba~2~Pb will be the best room temperature performance material. Optimal doping levels are likely as follows: 300 K: p-type -- 3 × 10^18^--10^19^ cm^−3^; n-type, 6 × 10^18^--3 × 10^19^ cm^−3^; 500 K, n-type, 3 --8 × 10^19^ cm^−3^. One additional area of interest regarding the thermopower concerns n-type Ca~2~Pb and Sr~2~Pb. As mentioned above, these may well have highly favorable thermoelectric performance at temperatures in the 800--1000 K range, despite the narrow band gaps of less than 0.4 eV. One factor supporting this observation is the likelihood of high electrical conductivity in these temperature ranges. Presented in figure [5](#F0005){ref-type="fig"} is a plot of the calculated direction averaged *σ*/*τ* versus direction averaged thermopower at 800 K for these two materials and compared with corresponding values (calculated within the same CSTA/GGA/mBJ potential framework) for the known high performance thermoelectrics PbTe and PbSe. While for p-type the performance of the lead chalcogenides appears more favorable than these alkaline earth plumbides, for n-type in the thermopower range around −200 *μ*VK^−1^ (the typical range for high performance thermoelectric), the *σ*/*τ* of Ca~2~Pb exceeds that of the lead chalcogenides by more than a factor of 2. Since the scattering time remains unknown, we cannot definitely assert high electrical conductivity in these compounds, but it is certainly possible. {#F0005} It is of interest to understand the reason for the high *σ*/*τ* results in n-type Ca~2~Pb and Sr~2~Pb, particularly since the CBM is located at Γ and hence has no pocket degeneracy. This is unlike in the high performance thermoelectrics Bi~2~Te~3~ and PbTe where the pocket degeneracy of the non-Γ-centered band extrema is a substantial contributor to transport. Part of the reason for these large *σ*/*τ* results can be seen from the band structure plots for these compounds in figure [2](#F0002){ref-type="fig"}, where subsidiary conduction band minima appear at T and on the Γ--*Y* line. Note also that these subsidiary minima are closer to the CBM in Ca~2~Pb than Sr~2~Pb, accounting for the slightly higher *σ*/*τ* in figure [5](#F0005){ref-type="fig"}. To shed further light on this we plot in figure [6](#F0006){ref-type="fig"} the Fermi surface of n-type Ca~2~Pb, at a *T* = 0 doping level of 3.1 × 10^20^ cm^−3^. In addition to the main ellipsoidal surface around Γ, one observes *six* additional Fermi surface pockets, four located around the zone boundary *T* point and two nearly spherical pockets located above and below Γ. It is this additional Fermi surface structure that is responsible for the large *σ*/*τ* in Ca~2~Pb and Sr~2~Pb. {#F0006} Since, as mentioned previously, the anisotropy of the electronic structure is a significant issue for these materials, we present in figure [7](#F0007){ref-type="fig"} plots of the electrical conductivity (divided by the scattering time, which is unknown) for these three materials, for both p-type and n-type. For Ca~2~Pb (top/left), the 500 K anisotropy is substantial for p-type, with approximately a factor of 5 difference between the conductivities in the *x* direction and that in the *y* and *z* directions. In general, such anisotropy will tend to degrade the electrical conductivity, and therefore thermoelectric performance. The n-type Ca~2~Pb results are more favorable, with only a factor of 3 anisotropy. For Sr~2~Pb, the 500 K anisotropy is also substantial for p-type, being roughly a factor of 10, and a more moderate 4 for n-type. For Ba~2~Pb (bottom/left), due to the probable lower optimal temperature range we have chosen to plot the 300 K results and find anisotropies of roughly 5 for p-type and roughly 2 for n-type. With the exception of the p-type Ca~2~Pb and the p-type Sr~2~Pb, these anisotropies are in the range of those of Bi~2~Te~3~, and therefore although the anisotropy is a concern for these materials, it is not necessarily sufficient to preclude high thermoelectric performance. The theoretical results of this section then suggest that all three materials may be useful thermoelectrics, with the n-type Sr and Ca compounds expected to be the best at temperatures from 500 to 1000 K and the Ba compound likely the best at low temperatures between 300 and 400 K. {#F0007} 3.. Theoretical results II: comparison of alkaline earth plumbides and stannides {#s3} ================================================================================ In this section we extend certain portions of the above analysis to the alkaline earth stannides Ca~2~Sn, Sr~2~Sn and Ba~2~Sn, which are isostructural and isoelectronic to the plumbides. While a full description of the potential performance of these stannides is beyond the scope of this paper, in figure [8](#F0008){ref-type="fig"} we present a comparison of the calculated DOS of Ba~2~Sn and Ba~2~Pb, Sr~2~Pb and Sr~2~Sn, and Ca~2~Pb and Ca~2~Sn. For all three plots the calculated DOS of the stannides and plumbides are very similar, with the main difference in fact being somewhat larger band gaps in the stannides. The agreement is particularly good within the first half of an eV in the valence and conduction bands, the relevant region for thermoelectric transport. This plot strongly suggests that the thermopower of those three compounds should also be favorable, and that these three stannides may also show good thermoelectric performance. In the high temperature range around 800 K the performance of these stannides for p-type may in fact exceed that of the plumbides due to the larger band gaps. Due to the presence of the lighter element tin, lattice thermal conductivities of these compounds are expected to be somewhat larger than those of the corresponding plumbides, but the isostructural and isoelectronic nature of all six compounds suggests that alloying on either site should be readily possible and would tend to reduce the lattice thermal conductivity. Note also that the larger band gaps associated with the stannides suggest that such alloying would tend to enlarge the band gap, and thereby enhance the p-type performance of the alkaline earth plumbides, which is particularly hampered by bipolar conduction. {#F0008} We have also performed calculations of the optical absorption of Sr~2~Pb and Sr~2~Sn, which are presented in figure [9](#F0009){ref-type="fig"}. It is worth noting that for both compounds the optical absorption is highest along the *b*-axis at low excitation energies, and in addition for both compounds the apparent onset of strong direct transitions, usually appearing at the band gap energy, here appears at energies of 1.2 eV, well above the true band gap. This will probably be a significant consideration in the interpretation of optical experiments designed to assess the actual band gap. Note also that for both compounds the onset of *b*-axis transitions is several tenths of an eV lower than for the other two directions. {#F0009} 4.. Theoretical results III: bulk modulus calculations and estimation of lattice thermal conductivity {#s4} ===================================================================================================== In the previous section we focussed on electronic structure and Boltzmann transport results, finding favorable band structures and thermopowers at a range of temperatures, depending upon the material. However, this is only part of what is necessary to produce a good thermoelectric, the other factors being low lattice thermal conductivity and acceptable carrier mobility. The mobility depends largely on the carrier scattering time, which is difficult to access in standard first principles approaches and can often be sample specific, depending on extrinsic factors such as point defects and surface oxidation. However, a reasonable guess can be made at the lattice thermal conductivity if one performs calculations to assess such factors as the bulk modulus. The bulk modulus has long been known as the determining factor in the velocity of the longitudinal (or 'compression', as opposed to 'shear') acoustic modes, which generally contribute a substantial fraction of the phononic heat transport in a given material. As with the electronic transport, a scattering time (related to mode anharmonicity, i.e. the Gruneisen parameters) enters as well, but factors such as the sound velocity, which directly contributes to the bulk modulus, play an important and easily assessed effect. In order to make this assessment we have performed calculations of the bulk moduli of each of the three materials studied. This is done by taking the experimental structure and altering each of the lattice constants by a specified amount: ±1%,±2%, and the self-consistent charge density computed. From this charge density one evaluates the total energy and then fits this energy to the well-known Birch--Murnaghan equation of state \[[@C31], [@C32]\] relating total energy, volume, bulk modulus *B* and the pressure derivative of the bulk modulus, d*B*/d*P*. Note that this calculation was performed with the standard GGA, not the mBJ potential which is inappropriate for structural calculation, as it does not derive from an energy functional. An additional point of note is that given that the energy differences associated with the above strains are fairly small (of the order of 10^−2^ Ryd), it is important to perform an optimization of any internal coordinates *independently for each strained structure*, to evaluate the actual energy sufficiently accurately. In general, other parameters, such as the number of *k*-points and LAPW sphere radii, were unchanged from the previous band structure calculations. In table [2](#TB2){ref-type="table"} we present our calculated GGA bulk moduli, and make a comparison with the values for the lead chalcogenides as calculated within the GGA by Zhang *et al* \[[@C33]\], matching compounds where the alkaline earth element and chalcogenide are of similar atomic number, for an overall comparison. The values decrease with increasing atomic mass, as expected, but what is more remarkable is that the values for the alkaline earth plumbides are roughly *half* of the values for the corresponding chalcogenides. These are indicative of low sound speeds. Note that our value for Ca~2~Pb is within 10% of the value calculated by Yang *et al* \[[@C8]\] and 15% of the value calculated by Migas *et al* \[[@C28]\]. Given that the chalcogenides have low room temperature lattice thermal conductivities in the 1.7--2.5 W m^−1^ K^−1^ range, this is suggestive that the alkaline earth plumbides considered here will have comparatively low lattice thermal conductivity as well. ###### Calculated GGA bulk moduli *B* (in GPa) for the alkaline earth plumbides (our calculations) and the lead chalcogenides (from \[[@C33]\]). Compound *B* (GPa) Density (g cm^−3^) ---------- ----------- -------------------- Ca~2~Pb 29.7 4.8 Sr~2~Pb 23.4 5.5 Ba~2~Pb 18.9 6.1 PbS 55.7 7.6 PbSe 49.2 8.1 PbTe 40.4 8.2 One possible reason for the lower bulk modulus values, relative to the lead chalcogenides, can be found from the atomic character plots depicted in figure [3](#F0003){ref-type="fig"}. All three of these compounds are substantially ionic, with the primary character in the valence band being lead (the more electronegative element) and that in the conduction band being the alkaline earth element. As is well known, ionic compounds tend to be weakly bonded, with the attendant low sound speeds and elastic moduli. These compounds have substantially larger interatomic distances than the lead chalcogenides, likely due to the larger ionic radii of the alkaline earth elements relative to the chalcogenides, causing lower bulk moduli. In this vein we can make a crude estimate of the possible lattice thermal conductivity for these materials. As is well known, for temperatures above the Debye temperature (of order 100--200 K for these soft materials) the lattice thermal conductivity can be estimated \[[@C33]\] as Here *M*~avg~ is the average atomic mass, in atomic mass units, *θ* the Debye temperature, *γ* the Gruneisen parameter, *δ* the cube root of the volume per atom and *n* the number of atoms per unit cell. *A* is a constant and if *δ* is measured in Å, *T* and *θ* in K and *κ*~lattice~ measured in W m^−1^ K^−1^, then *A* is 3.1 × 10^−6^. We may readily estimate the Debye temperature of each compound (for Ba~2~Pb, 116 K; for Sr~2~Pb, 143 K; for Ca~2~Pb, 180 K) from the computed bulk modulus and find that for the three respective (Ba, Sr, Ca) compounds, at 300 K, the lattice thermal conductivity is 1.6/*γ*^2^~Ba~2~Pb~ W m^−1^ K^−1^, 2.5/*γ*^2^~Sr~2~Pb~ W m^−1^ K^−1^, and 3.5/*γ*^2^~Ca~2~Pb~ W m^−1^ K^−1^. With, experimentally speaking, for most materials Gruneisen parameters falling between one and two, one finds potential 300 K lattice thermal conductivity values between 0.4 and 1.6 W m^−1^ K^−1^ for the Ba compound, 0.6 and 2.5 W m^−1^ K^−1^ for the Sr compound, and 0.9--3.5 W m^−1^ K^−1^ for the Ca compound. As suggested previously, these are comparatively low values, particularly for the Ba compound, which we expect to exhibit the best room temperature performance. 5.. Discussion and conclusion {#s5} ============================= The previous sections make clear that the alkaline earth plumbides and environmentally friendly stannides considered here, about which comparatively little is known experimentally, are quite likely to have the three characteristics necessary for good thermoelectric performance: high thermopower, low lattice thermal conductivity and good carrier mobility. One concern for practical applications, however, is the reactivity of the alkaline earth element, particularly at the elevated temperatures necessary for waste heat recovery. While calcium, strontium and barium have melting points of 1000 K or more (note that the corresponding chalcogens all have melting points below 725 K), it is possible that samples of the thermoelectrics considered here will oxidize in air, so that effort may be necessary to isolate the samples from air and water vapor. In any case, future experimental study is expected to reveal much about the potential of these materials for thermoelectric applications. To summarize, in this paper we have shown that the alkaline earth plumbides Ca~2~Pb, Sr~2~Pb and Ba~2~Pb and stannides Ca~2~Sn, Sr~2~Sn and Ba~2~Sn, presently virtually unexplored as thermoelectrics, have substantial potential for thermoelectric performance, both at room temperature and at high temperature, and therefore warrant experimental exploration. We await the results of such experimental inquiries. This research was supported by the US Department of Energy, EERE, Vehicle Technologies, Propulsion Materials Program.

Background {#Sec1} ========== Whilst adolescence and young adulthood is often thought of as a period of relative good health, burden of disease studies indicate that10--24 year olds remain at substantial risk of morbidity and mortality \[[@CR1]\]. Burden of disease in this age group is not driven by physical illness, but by mental illness, self-harm and suicide. Meanwhile, the leading risk factors contributing to this burden of disease in young people include risky alcohol use, illicit drug use and unprotected sex \[[@CR2], [@CR3]\]. These risk factors are not only associated with poor outcomes for youth in the short-term, such as obesity and mental health problems, but are also associated with morbidity and mortality experienced later in life \[[@CR4]--[@CR7]\]. Evidence has consistently shown that these risk behaviours are linked to the development of later chronic non-communicable conditions, such as heart disease, cancer and Type II diabetes, which cause considerable disease burden amongst older adults \[[@CR8]--[@CR10]\]. Experts have therefore suggested that chronic health disease risk factor management should begin in adolescence, as evidence is accumulating on the benefits of early intervention for future health gains \[[@CR11]\]. Identifying clusters of potentially modifiable risk behaviours associated with morbidity and mortality amongst adolescents, and risk for later chronic disease in adulthood, would ideally guide the development of preventive interventions designed to reduce the burden of disease across the lifespan. Previous research has linked certain clusters of risk factors with indicators of mental health in young people. For example, the combination of concurrent risky alcohol use, drug use and smoking is related to increased depressive symptom severity in young people aged 15--30 years \[[@CR12]\], as has the clustering of physical inactivity, smoking, low-quality diet, and abnormal (high or low) body mass index (BMI) \[[@CR4]\]. The cluster of smoking, alcohol misuse, marijuana use, sexual activity, violence, and suicidality has also predicted a range of psychopathology in children and adolescents aged 9--17 years, including mood, anxiety and disruptive disorders \[[@CR13]\]. In another study, risky alcohol use, illicit drug use, smoking, sleep deprivation, overweight/underweight, sedentary behaviour, high media use, and truancy were related to a range of poor mental health outcomes, including depression, anxiety and suicidality amongst adolescents (mean age 14.9 years) \[[@CR14]\]. Recent research has therefore recognised the importance of lifestyle risk factors in the determination of adolescent mental health. In the adult literature, the importance of these risk factors has led to the development of lifestyle risk indices comprised of factors related to disease burden in adults \[[@CR7], [@CR15]\]. To our knowledge, a similar index of adolescent-specific risk factors for burden of disease has not been constructed. It is envisaged that this lifestyle risk factor index could be used to easily identify adolescents at risk for chronic diseases of both adolescence and adulthood in a rigorous and standardised way. Those identified as at risk could then be targeted by healthy lifestyle interventions designed to reduce the burden of disease across the lifespan. Using a large nationally representative sample of Australian adolescents, the current study will construct a lifestyle risk index comprising the leading risk factors for adolescent burden of disease (i.e., risky alcohol use, drug use and unprotected sex) \[[@CR3]\]. The extent to which this index predicts indicators of the leading causes of adolescent morbidity and mortality (i.e., depression, psychological distress, self-harm and suicide attempt) will also be investigated. The study will also investigate whether the addition of conventional (i.e., smoking and BMI) and emerging (i.e., sleep duration) risk factors associated with disease burden in adulthood \[[@CR7], [@CR14], [@CR16], [@CR17]\] improves the measurement properties of this index. Methods {#Sec2} ======= Sample {#Sec3} ------ The 2013--2014 s Australian Child and Adolescent Survey of Mental Health and Wellbeing was based on a stratified, multistage area probability sample of households where there was at least one child aged 4--17 years. The survey represents the most up-to-date, detailed snapshot of mental health and wellbeing in adolescents in Australia. In total 6310 parents and carers of eligible households participated in the survey, representing an overall response rate of 55%. In addition, 2967 (89%) of young people aged 11--17 years, for whom their parents or carers had given written consent, also gave written consent to complete a youth self-report questionnaire. Comparison with 2011 Australian Census data indicated that the sample was broadly representative of the Australian population in terms of demographic characteristics \[[@CR18]\]. The survey data were weighted to represent the Australian population of children and adolescents aged 4--17 years based on 2013 data from the Australian Bureau of Statistics. The current study uses data from youth aged 13--17 years (*n* = 2314). Those aged 11--12 did not provide data on the relevant risk behaviours and were therefore excluded from the current study. The survey received ethical approval from the Australian Government Department of Health. The survey methods have been discussed in more detail elsewhere \[[@CR18]\] and full questionnaires used in the survey can be accessed here: <https://youngmindsmatter.telethonkids.org.au/for-researchers/>. Measures {#Sec4} ======== Key variables accounting for morbidity/mortality in adolescence {#Sec5} --------------------------------------------------------------- The 2013--2014 s Australian Child and Adolescent Survey of Mental Health and Wellbeing included a self-report module based on the Diagnostic Interview Schedule for Children (DISC-IV) which was used to provide a 12-month diagnosis of major depressive disorder (MDD) \[[@CR19]\]. The DISC-IV is designed for use in children and adolescents aged 6--17 years and has been shown to have acceptable test-retest reliability in community samples \[[@CR19]\]. Questions from the Youth Risk Behavior Surveillance System (YRBSS) were also included in the survey and were used to provide data for the suicide attempt and self-harm outcomes \[[@CR20]\]. Participants were asked whether they had attempted suicide in the past 12 months (suicide attempt) and whether they had deliberately harmed or injured themselves without intending to end their own life during the past 12 months (self-harm). Psychological distress in the past 4 weeks was measured by the 10-item Kessler Psychological Distress scale (K10) \[[@CR21]\], which has been used previously in samples of adolescents in the Australian general population \[[@CR22]\], and demonstrated excellent reliability in the current sample (Cronbach's α = 0.91). Those reporting severe psychological distress (scores 21 and above) were compared with those with scores below this standard cut-off \[[@CR23]\]. The dichotomised version of the K10 has been used extensively, and has been shown to predict serious mental illness in both adolescents and adults with good accuracy \[[@CR22], [@CR24], [@CR25]\]. Risk factors {#Sec6} ------------ Self-reported risky alcohol use, drug use, smoking, and unprotected sex were collected as part of the YRBSS. To evaluate an index based on these risk factors, each was coded as 0 (not at risk) or 1 (at risk). Alcohol risk was defined as consuming four or more standard drinks on a single occasion in the past 30 days, based on the Australian National Health and Medical Research Council's definition of single occasion risky drinking \[[@CR26]\]. Respondents were asked to select which drugs they had used in the past 30 days from a list including cannabis, meth/amphetamine, cocaine, ecstasy and prescription drugs for non-medical or non-prescribed purposes. Drug risk was defined as the use of any of these illicit drugs in the past 30 days. Those who had smoked cigarettes in the past 30 days were considered at smoking risk. Respondents were asked whether they used any form of protection to prevent pregnancy or sexually transmitted infections. Sexual risk was defined as those who reported using no form of protection the last time they had sexual intercourse. Youth were also asked to self-report their current height and weight \[used to derive BMI (height (m)/weight^2^ (kg)\] and the average amount of sleep each week night and each weekend night (hours). Those with a BMI less than 18.5 or higher than 30 were considered at weight risk, according to standard guidelines of underweight and obesity from the Australian Government Department of Health \[[@CR27]\]. Both ends of the BMI spectrum were considered given evidence suggesting that the relationship between mental health and BMI is U-shaped, with those who are both underweight and overweight at higher risk for mental health problems such as depression \[[@CR28]\]. In addition, both weight gain and weight loss are considered symptoms of major depression \[[@CR29]\]. Finally those who slept less than 7 hours and more than 11 hours per night on average were considered at sleep risk, according to Australian sleep guidelines for adolescents \[[@CR30]\]. Sleep problems, which include both hypersomnia and insomnia, are associated with mental health problems, and are both considered symptoms of major depression \[[@CR29]\]. We constructed a basic risk index, which consisted of risky alcohol use, drug use and unprotected sex, the leading risk factor associated with adolescent disease burden \[[@CR2], [@CR3]\]. The basic risk index scores ranged from 0 to 3. Three other specific risk indices were then created, comprising the basic index + smoking, the basic index + BMI and the basic index + sleep. Each of the specific risk index scores ranged from 0 to 4. The basic index + smoking + BMI, the basic index + smoking + sleep, and the basic index + BMI + sleep were also constructed (scores ranging from 0 to 5). Finally, an index consisting of all six risk factors (risky alcohol use, drug use, unprotected sex, smoking, BMI and sleep) was constructed (ranging from 0 to 6). Table [1](#Tab1){ref-type="table"} includes a description of each of the risk factors, as well as a summary of how each of the risk indices were composed.Table 1Lifestyle factors, risk scoring methods and risk indices derived from the 2013--2014 Australian Child and Adolescent Survey of Mental Health and Wellbeing (*n* = 2314)Risk indicesLifestyle factorRisk Scoring MethodBasic riskBasic risk + BMIBasic risk + SleepBasic risk + SmokingBasic risk + BMI + SleepBasic risk + BMI + SmokingBasic risk + Sleep + SmokingBasic risk + BMI + Sleep + SmokingAlcohol use*1 = 4 or more drinks on a single occasion in past month*✓✓✓✓✓✓✓✓Illicit drug use*1 = past 30-day illicit drug use*✓✓✓✓✓✓✓✓Unprotected sex*1 = no protection used against pregnancy or sexually transmitted diseases the last time respondent had sexual intercourse*✓✓✓✓✓✓✓✓BMI*1 = Body mass index \< 18.5 or \> 30*✓✓✓✓Sleep*1 = \< 7 h or \> 11 h/day*✓✓✓✓Smoking*1 = past 30-day cigarette smoking*✓✓✓✓ Statistical analyses {#Sec7} -------------------- All analyses were weighted to account for the complex survey design and were conducted using SAS 9.4. The prevalence of each risk factor and the four leading causes of disease burden in adolescence (self-reported MDD, suicide attempt, self-harm, severe psychological distress) were first analysed by key demographic variables: age, sex, rurality (classified as greater capital cities or rest of state, as per the Australian Bureau of Statistics' Australian Statistical Geography Standard), country of birth and current education status (whether or not the participant currently goes to school, as reported by the parent). The extent to which the risk factors co-occurred was also investigated using tetrachroic correlations. Preliminary analyses focused on whether the basic risk index (alcohol use, drug use and unprotected sex) predicted the four disease burden outcomes with greater precision than any single risk factor, or combination of two risk factors, that comprised this index. The predictive utility of the basic risk index was then compared with the seven other indices constructed as described above and in Table [1](#Tab1){ref-type="table"}. The predictive utility of these various combinations of risk factors was assessed using receiver operating characteristic (ROC) curves created with logistic regressions that controlled for demographic characteristics (age, sex, country of birth and education status) which were likely to be related to both the risk factors and indicators of burden of disease in young people.. The area under the curve (AUC) was calculated to describe how well each index classified the four causes of disease burden in adolescence, and to compare these indices in terms of predictive value. An AUC of 0.5 indicates that an index is no better than chance at predicting risk, whilst an AUC of 1.0 indicates that an index predicts risk perfectly. Areas under the curve for each index were compared with the basic risk index (reference index) using the ROCCONTRAST statement in SAS 9.4 which implements a nonparametric approach as recommended by methodological experts \[[@CR31]\]. An optimal at-risk threshold (or cut-off) was then established for the index that predicted the four outcome variables with the highest precision. This optimal threshold was selected by comparing AUCs characterising the relationship between all possible thresholds on the best performing index and the four disease burden outcome variables of interest. A threshold of one risk factor was first imposed on the best performing index, and AUCs were constructed describing the precision with which one risk factor predicted each of the disease burden outcome variables. A threshold of two risk factors was then imposed and AUCs were again constructed describing the precision with which a total of two risk factors predicted each of the disease burden outcomes variables. This procedure was then replicated for each possible threshold on the best performing risk index. Basic demographic characteristics of those identified as at-risk according to this optimal threshold were then described using univariate and multivariate logistic regression models. Results {#Sec8} ======= Overall, 48.7% of the sample were female, the mean age was 15.0, 63.5% of the sample were from metropolitan areas, 85.8% were Australian born and 93.0% of the sample were currently attending school. Descriptive statistics by causes of disease burden outcomes and major risk behaviours are presented in Table [2](#Tab2){ref-type="table"}. Based on tetrachoric correlations (Table [3](#Tab3){ref-type="table"}), risky alcohol use, drug use, smoking, unprotected sex, and sleep were all significantly and highly clustered risk factors. BMI did not appear to cluster with these other factors.Table 2Prevalence of lifestyle risk factors by demographic characteristics in the 2013--2014 Australian Child and Adolescent Survey of Mental Health and Wellbeing (n = 2314)Major depression % (SE)Self-harm % (SE)Suicide attempt % (SE)Severe psychological distress % (SE)Risky alcohol use % (SE)Illicit drug use % (SE)Unprotected sex % (SE)Sleep % (SE)Smoking % (SE)BMI% (SE)Sex Male5.4 (0.6)4.4 (0.6)1.3 (0.3)3.7 (0.6)12.6 (1.0)5.6 (0.7)1.0 (0.3)18.7 (1.2)6.2 (0.8)27.1 (1.5) Female15.5 (1.1)13.7 (1.0)4.0 (0.6)9.2 (0.9)12.5 (1.0)5.5 (0.7)2.1 (0.5)21.9 (1.3)8.2 (0.8)29.2 (1.6)Age 134.2 (1.2)5.3 (1.3)1.0 (0.6)2.1 (0.8)1.0 (0.7)1.9 (0.9)0.5 (0.5)15.5 (2.2)1.2 (0.7)42.5 (3.0) 147.3 (1.4)7.9 (1.5)2.3 (0.9)7.4 (1.5)4.2 (1.2)2.7 (0.8)0.5 (0.5)17.2 (2.2)4.8 (0.3)31.9 (2.6) 1510.7 (1.8)8.3 (1.6)2.2 (0.8)5.7 (1.4)8.1 (1.6)5.0 (1.3)1.6 (0.7)18.6 (2.4)5.8 (1.4)26.7 (2.7) 1615.6 (1.4)12.1 (1.3)3.9 (0.7)8.0 (1.0)20.9 (1.6)10.4 (1.2)2.4 (0.6)24.2 (1.6)11.4 (1.2)20.6 (1.6) 1713.8 (1.4)11.2 (1.3)3.7 (0.7)8.7 (1.1)27.8 (1.9)7.5 (1.1)2.7 (0.7)25.6 (1.8)12.4 (1.6)20.7 (1.7)Area City10.5 (0.8)8.7 (0.8)2.4 (0.4)6.6 (0.7)11.7 (0.9)5.7 (0.6)1.5 (0.3)20.4 (1.2)6.6 (0.6)28.4 (1.3) Rural10.0 (1.1)9.4 (1.1)3.1 (0.7)6.0 (0.9)14.0 (1.2)5.2 (0.8)1.7 (0.5)20.0 (1.6)8.2 (1.2)27.8 (2.1)Country of birth Australia10.8 (0.7)9.2 (0.7)3.0 (0.4)6.7 (0.6)13.5 (0.8)6.0 (0.6)1.7 (0.3)19.7 (1.0)7.7 (0.6)27.7 (1.2) Overseas7.8 (1.3)7.8 (1.5)0.6 (0.4)4.7 (2.5)6.9 (1.3)2.8 (0.9)0.8 (0.4)23.8 (2.5)4.2 (1.1)30.5 (2.9)Education Not attending school19.9 (2.7)15.5 (2.5)4.8 (1.6)9.7 (2.1)37.9 (3.3)16.6 (2.6)8.1 (1.9)28.4 (3.0)28.5 (3.4)18.6 (2.8) Attending school9.6 (0.6)8.5 (0.6)2.5 (0.3)6.1 (0.6)10.6 (0.7)4.7 (0.5)1.1 (0.3)19.6 (1.0)5.6 (0.5)28.9 (26.5)Total10.4 (0.6)9.0 (0.86)2.6 (0.3)6.4 (0.5)12.5 (0.7)5.5 (0.5)1.6 (0.3)20.3 (1.0)7.2 (0.6)28.1 (1.1)NB: Risky alcohol use defined as 4 or more standard drinks in past 30 days; illicit drug use risk defined as any drug use in past 30 days; unprotected sex risk defined as no protection against pregnancy of sexually transmitted diseases the last time the respondent had sexual intercourse; BMI (body mass index) risk defined as less than 18.5 or greater than 30; sleep risk defined as less than 7 hours per day or more than 11 hours per day on average; smoking risk defined as cigarette smoking in the past 30 daysTable 3Tetrachoric correlations between lifestyle risk factors included in the 2013--2014 Australian Child and Adolescent Survey of Mental Health and Wellbeing (*n* = 2314)Risky Alcohol UseIllicit Drug UseUnprotected SexBMISleepSmokingAlcohol use1.000.69^b^0.25^a^−0.14^a^0.13^a^0.65^b^Illicit drug use1.000.47^b^0.010.27^b^0.77^b^Unprotected sex1.000.010.080.51^b^BMI1.00−0.040.05Sleep1.000.30^b^Smoking1.00^a^*p* \< 0.01; ^b^ *p* \< 0.0001NB: Risky alcohol use defined as 4 or more standard drinks in past 30 days; illicit drug use risk defined as any drug use in past 30 days; unprotected sex risk defined as no protection against pregnancy of sexually transmitted diseases the last time the respondent had sexual intercourse; BMI (body mass index) risk defined as less than 18.5 or greater than 30; sleep risk defined as less than 7 hours per day or more than 11 hours per day on average; smoking risk defined as cigarette smoking in the past 30 days Preliminary analyses focused on validating the basic risk index comprising alcohol use, drug use and unprotected sex. The basic risk index was found to predict the four disease burden outcomes with fair to good precision (Table [4](#Tab4){ref-type="table"}), and with more accuracy than any of the composite risk factors either alone (i.e., alcohol use, drug use or unprotected sex) or in combination (i.e., alcohol use + drug use, alcohol use + unprotected sex or drug use + unprotected sex) (results available on request). Table [4](#Tab4){ref-type="table"} presents the AUCs for the basic risk index, as well as comparisons between this index and the AUCs for each of the calculated risk indices, in terms of their ability to predict the four disease burden outcomes. When comparing the different indices, the basic risk + sleep + smoking index predicted the disease burden outcomes with the greatest precision, and with greater precision than the basic risk index alone (all contrast *p*s \< .01). The basic risk + sleep + smoking index was therefore selected as the most precise risk index predicting adolescent disease burden. Based on the basic risk + sleep + smoking index, 68.1% of the sample reported no risk behaviours, 22.7% reported one risk behaviour, 4.8% reported two risk behaviours, 3.2% reported three risk behaviours, 1.1% reported four risk behaviours, and 0.1% reported all five risk behaviours.Table 4Area under the curves for each lifestyle risk index and causes of adolescent disease burden in the 2013--2014 Australian Child and Adolescent Survey of Mental Health and Wellbeing (*n* = 2314)Risk indicesBasic risk AUC (95% CI)Basic risk + BMI AUC (95% CI)Basic risk + Sleep AUC (95% CI)Basic risk + Smoking AUC (95% CI)Basic risk + BMI + Sleep AUC (95% CI)Basic risk + BMI + Smoking AUC (95% CI)Basic risk + Sleep + Smoking AUC (95% CI)Basic risk + BMI + Sleep + Smoking AUC (95% CI)Major depression0.706 (0.671--0.742)0.702 (0.667--0.736)0.753 (0.718--0.787) ^c^0.713 (0.678--0.748)0.739 (0.705--0.773) ^c^0.709 (0.675--0.743)0.757 (0.723--0.791) ^c^0.743 (0.709--0.776) ^c^Self-harm0.717 (0.684--0.751)0.710 (0.677--0.743)0.743 (0.711--0.776) ^b^0.723 (0.690--0.757)0.729 (0.697--0.762)0.717 (0.684--0.750)0.747 (0.715--0.780) ^b^0.736 (0.703--0.768) ^a^Suicide attempt0.765 (0.708--0.823)0.766 (0.708--0.824)0.826 (0.777--0.875) ^c^0.791 (0.736--0.847) ^a^0.820 (0.775--0.864) ^c^0.798 (0.742--0.855) ^a^0.838 (0.793--0.884) ^c^0.836 (0.793--0.880) ^c^Severe psychological distress0.701 (0.655--0.747)0.693 (0.649--0.738)0.759 (0.717--0.802) ^c^0.709 (0.663--0.754)0.737 (0.694--0.779) ^b^0.701 (0.656--0.747)0.768 (0.727--0.810) ^c^0.743 (0.702--0.785) ^ba^contrast with basic risk index *p* \< 0.05; ^b^ contrast with basic risk index *p* \< 0.01; ^c^ contrast with basic risk index *p* \< 0.001NB: Risky alcohol use defined as 4 or more standard drinks in past 30 days; illicit drug use risk defined as any drug use in past 30 days; unprotected sex risk defined as no protection against pregnancy of sexually transmitted diseases the last time the respondent had sexual intercourse; BMI (body mass index) risk defined as less than 18.5 or greater than 30; sleep risk defined as less than 7 hours per day or more than 11 hours per day on average; smoking risk defined as cigarette smoking in the past 30 days A threshold of one or more risk behaviours on the basic risk index + sleep + smoking index was the best predictor of depression \[AUC = 0.747 (95% CI: 0.715, 0.779); sensitivity = 0.667 (95% CI: 0.605--0.730); specificity = 0.718 (95% CI: 0.696--0.741)\], self-harm \[AUC = 0.731 (95% CI: 0.700, 0.762); sensitivity = 0.637 (95% CI: 0.569--0.705); specificity = 0.710 (95% CI: 0.688--0.732)\] and severe psychological distress \[AUC = 0.753 (95% CI: 0.714, 0.793); sensitivity = 0.705 (95% CI: 0.626--0.784); specificity = 0.705 (95% CI: 0.683--0.727)\]. A threshold of two risk behaviours was a marginally better predictor for suicide attempt \[one behaviour AUC = 0.797 (95% CI: 0.761, 0.832); sensitivity = 0.827 (95% CI: 0.726--0.928); specificity = 0.692 (95% CI: 0.670--0.715); two behaviour AUC = 0.806 (95% CI: 0.754, 0.858); sensitivity = 0.626 (95% CI: 0.497--0.756); specificity = 0.919 (95% CI: 0.907--0.931)\], but this difference was not statistically significant (*p* = 0.689). A threshold of one or more risk behaviours on the basic risk index + sleep + smoking was therefore selected as the optimal cut-off in terms of predicting each the four causes of disease burden in adolescence. Overall, 31.9% of the sample scored above this threshold. Table [5](#Tab5){ref-type="table"} displays the demographic characteristics of adolescents who were classified as 'at risk' according to the one or more threshold on the basic risk + sleep + smoking index. The odds of being classified as 'at risk' increased with age, and were greater amongst those not attending school.Table 5Demographic characteristics of at risk group (i.e., those reporting one or more behaviours on the basic risk + sleep + smoking index) in the 2013--2014 Australian Child and Adolescent Survey of Mental Health and Wellbeing (*n* = 2314)% (SE)Unadjusted OR (95% CI)Adjusted OR (95% CI)Sex Male30.4 (1.5)1.17 (0.98--1.40)0.87 (0.72--1.06) Female33.9 (1.5)\[ref\]\[ref\]Age 1316.7 (2.4)\[ref\]\[ref\] 1422.4 (2.4)1.43 (0.93--2.21)1.42 (0.92--2.19) 1526.5 (2.6)1.79 (1.126--2.78)^b^1.75 (1.13--2.71)^a^ 1643.1 (1.9)3.78 (2.62--5.45)^c^3.55 (2.46--5.14)^c^ 1751.0 (2.1)65.20 (3.60--7.50)^c^4.05 (2.77--5.92)^c^Area City31.2 (1.4)\[ref\]\[ref\] Rural33.6 (1.9)1.12 (0.91--1.37)1.14 (0.92--1.42)Country of birth Australia32.3 (1.2)1.06 (0.81--1.39)1.05 (0.79--1.40) Overseas31.0 (2.7)\[ref\]\[ref\]Education Not attending school66.5 (3.3)4.74 (3.51--6.41)^c^2.67 (1.98--3.62)^c^ Attending school29.5 (1.1)\[ref\]\[ref\]Total31.9% (1.1)----^a^*p* \< 0.05; ^b^ *p* \< 0.01; ^c^ p \< 0.001NB: Risky alcohol use defined as 4 or more standard drinks in past 30 days; illicit drug use risk defined as any drug use in past 30 days; unprotected sex risk defined as no protection against pregnancy of sexually transmitted diseases the last time the respondent had sexual intercourse; BMI (body mass index) risk defined as less than 18.5 or greater than 30; sleep risk defined as less than 7 hours per day or more than 11 hours per day on average; smoking risk defined as cigarette smoking in the past 30 days Discussion {#Sec9} ========== In a large, nationally representative sample of Australian adolescents, this study investigated the clustering of lifestyle risk factors, and the extent to which these clusters predicted the key causes of excessive adolescent burden of disease. A lifestyle risk index comprised of risky alcohol use, drug use, unprotected sex, sleep duration and smoking predicted each of the disease burden outcomes with the best precision. When an empirically-derived threshold of one or more risk behaviours was imposed on this index, the prediction of these health outcomes ranged from fair to good. Such a high level of predictive precision was surprising, especially given the complexity of the disease burden outcomes investigated. These results suggest that this lifestyle risk index represents a useful summary metric in the context of health promotion and non-communicable disease prevention. Major depression is predicted to become the largest contributor to burden of disease by 2030 \[[@CR32]\], and is associated with other non-communicable diseases that contribute excessively to the burden of disease in adults, such as heart disease, cancer and Type II diabetes \[[@CR33]\]. Prevention programs targeting these novel and potentially modifiable lifestyle risk factors in adolescence may therefore be instrumental in reducing the considerable burden of disease associated with poor mental health across the lifespan. The current study proposes a lifestyle risk factor index that could be used in these prevention programs to identify those adolescents at risk for poor mental health outcomes and who my benefit from a healthy lifestyles intervention. Of note, this study identified sleep as an important risk factor in the prediction of adolescent disease burden. This is consistent with similar studies of the adult population \[[@CR7]\]. Recent "calls to action" have focused on the importance of sleep, particularly given its relationship with risk behaviours associated with weight gain \[[@CR17]\]. Longitudinal studies have also found that poor sleep quality in early adolescence is associated with earlier onset of alcohol and cannabis use \[[@CR16]\]. There is consistent evidence to therefore suggest that sleep is an important risk factor contributing to disease burden, and should be a target for future health promotion activities. These findings have implications for the types of lifestyle factors that may be targeted by multiple health behaviour change interventions focusing on adolescents. Some of the lifestyle factors that form the risk index were adolescent-specific contributors to burden of disease (i.e., unprotected sex and drug use) \[[@CR2], [@CR3]\]. Risky alcohol use and sleep duration, however, have also been identified as important predictors of adult health outcomes \[[@CR7], [@CR8]\]. These findings suggest that focusing on these latter risk factors may be useful for addressing concurrent adolescent health as well as future morbidity and mortality associated with chronic diseases of adulthood. Thus, these lifestyle risk factors may therefore be particularly important to target in adolescent preventive interventions given their potential for both short- and long-term impact. Consistent with previous research \[[@CR4], [@CR13], [@CR14], [@CR34]--[@CR40]\], this study indicated that lifestyle risk factors tend to cluster together in adolescence. This clustering may occur through direct causation (i.e., unprotected sex as a result of alcohol or drug use) or shared aetiologies (i.e., social disadvantage). Given that these lifestyle risk factors are highly clustered, modifying one factor may also lead to changes in another. Historically, interventions aimed at preventing lifestyle risk factors have targeted a single factor \[[@CR41]\]. However, the clustering of lifestyle risk factors suggests that multiple risk factors should be targeted together, rather than in isolation. Interventions designed specifically to target multiple risk factors are less resource intensive and more cost-effective than single factor interventions \[[@CR41]\]. There is a growing literature on multiple health behaviour change interventions among adolescents and adults, but many of these are limited to small number of risk factors, and few include emerging risk factors, such as sleep duration which was identified as important in the current study \[[@CR38], [@CR41]--[@CR45]\]. Future research should focus on developing and evaluating multiple health behaviour change interventions for adolescents and young adults, perhaps including the risk factors identified by the current study. The current study also provides evidence as to the setting and timing of multiple health behaviour change interventions for adolescents. Youth outside of school settings are much more likely to be at risk in terms of the lifestyle factors included in this study. Whilst previous research indicates that school may be the best place for implementing "universal" prevention strategies \[[@CR46]\], "selected" or "indicated" risk factor prevention could be considered as a means of targeting adolescents who are no longer attending school. Delivering interventions to high risk adolescents outside of school settings is a current challenge that should be prioritised. In the current study, the prevalence of these lifestyle risk factors increased with age, with a particularly sharp increase between the ages of 15 and 16 years. Although chronic disease prevention is advocated across the lifespan, the present findings suggest than an optimal time to deliver preventive interventions is in early adolescence, before the age 15, when these lifestyle risk factors are more likely to be initiated and before they become entrenched. The findings from this study need to be considered within the context of its strengths and limitations. This study included data from a large, nationally representative sample of adolescents from the general population. In addition, health outcomes related to the leading global burdens of disease in adolescents were measured. Depression was assessed using a self-administered structured diagnostic questionnaire, and other health outcomes were assessed using standardised measures. Limitations of this study include its cross-sectional design. Longitudinal studies are needed to explore how these lifestyle risk factors co-vary over time to predict morbidity and mortality in adolescence through to adulthood. Risk factors and health outcomes were all self-reported, with the possibility of over- or under-reporting by participants especially with the querying of sensitive topics. Future research should also attempt to use objective measures to validate self-report of physical activity, sedentary behaviour and sleep patterns and sleep duration. There is an absence of energy balance risk behaviours in our risk index, and we included BMI as a proxy for these behaviours. Physical inactivity, diet and sedentary behaviour have been found to co-occur in adolescents \[[@CR47]\] and to predict poor outcomes in adolescence \[[@CR17]\] and the incidence of later cardiovascular disease, cancers and diabetes \[[@CR48]\]. These measures were not available in the present dataset, however future research should examine these risk factors alongside those investigated in the present study as a means of further guiding intervention efforts. Finally, whilst for some lifestyle factors, risk was defined with reference to national guidelines (i.e., alcohol use, sleep and BMI) for others there were no guidelines available to determine risk thresholds (i.e., drug use, unprotected sex and smoking). However, due to their illegality and/or the dire consequences associated with these behaviours, it could be argued that any involvement in these behaviours constitutes significant risk. Conclusions {#Sec10} =========== In a large, nationally representative sample of Australian adolescents, a lifestyle risk index comprised of risky alcohol use, drug use, unprotected sex, sleep duration and smoking was a fair to good predictor of health outcomes associated with the leading fatal (suicide attempt and self-harm) and non-fatal (major depression and severe psychological distress) global burdens of disease amongst adolescents. This study indicates that this lifestyle risk index represents a useful summary metric in the context of adolescent health promotion and non-communicable disease prevention. Lifestyle risk factors were found to cluster in adolescence, providing further support for the implementation of multiple health behaviour change interventions over those with a single behaviour focus. Future research should focus on determining whether this lifestyle risk index predicts adolescent health outcomes in a longitudinal framework. AUC : Area Under the Curve BMI : Body Mass Index DISC-IV : Diagnostic Interview Schedule for Children 4th Edition K10 : Kessler Psychological Distress Scale MDD : Major Depressive Disorder YRBSS : Youth Risk Behavior Surveillance System Not applicable. Funding {#FPar1} ======= Dr. Mewton is funded by an Australian Rotary Health Postdoctoral Fellowship. Australian Rotary Health had no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript. Availability of data and materials {#FPar2} ================================== The datasets analysed during the current study are available from the Telethon Kids Institute. Data are released under special restricted access, which requires users to have obtained ethics approval for their research. Details can be accessed here: <https://youngmindsmatter.telethonkids.org.au/siteassets/media-docs%2D%2D-young-minds-matter/data-access-statement2.pdf> LM, KC, MS, FK-L, LT and MT contributed to the concept and design of the study. LM analysed and interpreted the data and wrote the first draft of the manuscript. LM, KC, MS, FK-L, LT and MT contributed to subsequent drafts and read and approved the final manuscript. Ethics approval and consent to participate {#FPar3} ========================================== This study was approved by the University of New South Wales Human Research Ethics Committee. Adolescents and their parents provided informed written consent to participate in the Australian Child and Adolescent Survey of Mental Health and Wellbeing. Consent to publish {#FPar4} ================== Not applicable. Competing interests {#FPar6} =================== The authors declare that they have no competing interests. Publisher's Note {#FPar7} ================ Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Introduction ============ *Orcula* [@B30], the type genus of the family Orculidae is a group of small (5--10 mm), pulmonate land snails with ovate--cylindrical shells and 3--4 lamellae within the aperture and the body whorl. Taxa are primarily found in mountainous regions in relatively humid habitats, most commonly in deciduous forests. There are at least 50 names described within the genus and 14 of them are considered as valid on species level ([@B27]). The Alps are inhabited by eight *Orcula* species, and this area is considered as the centre of the diversity of the genus. The type species, *Orcula dolium* ([@B14]), has the widest distribution within the genus, occurring from Eastern France to Eastern Slovakia and North-Eastern Hungary and several subspecies are recognized ([@B35], [@B21], [@B27]). Other Alpine *Orcula* species have much narrower areas and occur in the Alps of Austria, Italy and Slovenia only. Non-Alpine species include *Orcula jetschini* ([@B34]) from Romania (Banat, Transylvania, Crişana), *Orcula schmidtii* ([@B37]) from Montenegro, Albania and northwestern Greece, *Orcula wagneri* [@B66] from Albania, Macedonia (FYROM) and Kosovo and *Orcula zilchi* [@B68] is distributed from South-Western Bulgaria to North-Western Turkey. The anatomy of the Alpine and the Illyric *Orcula* species is well-known (see [@B21] and [@B59]). On the other hand, the reproductive anatomy of the two Eastern European species (*jetschini* and *zilchi*) remained unpublished. [@B21] presented compelling data regarding the utility of the epiphallus and penial caecum (= ''Flagellum'' or ''Penisflagellum'') in taxonomic studies in *Orcula*. [@B22] noted that extra-Alpine species had much stronger shell sculpture than Alpine species. These conchological and anatomical data, however, did not provide the resolution required to properly subdivide the genus. Recently, [@B59] evaluated the taxonomic positions of most *Orcula* species. He concluded that the genus can be divided into two groups based on the morphology at the epiphallus--vas deferens transition. The transition is abrupt in *Orcula conica* ([@B55]), *Orcula fuchsi* [@B78], *restituta* ([@B74]), *spoliata* ([@B55]) and *Orcula dolium*, whereas it is gradual in *gularis* ([@B55]), *austriaca* [@B79], *tolminensis* [@B72], *wagneri* and *schmidtii*. It is not possible to assign some *Orcula dolium* specimens to one group or another. Independent of these two groups, [@B59] delineated five ''clusters'' (species groups) on the basis of anatomical and conchological characters (1: *conica*, 2: *fuchsi*, 3: *dolium* (s. l.) + *spoliata*, 4: *austriaca* (s. l.) + *tolminensis*+ *gularis*, 5: *schmidtii*+ *wagneri*). In Schileyko's phylogenetic scheme *Orcula conica* (unique shell shape and peculiar position of the penial caecum) and *Orcula fuchsi* (unique structure of the epiphallus) are the most basal members of the genus. Recently, living specimens of *Orcula jetschini* and *Orcula zilchi* were made available for study. Anatomical investigation of these species allowed us to fully evaluate the taxonomic relationships within *Orcula*. We present data here that establishes subgenera within *Orcula* based on shell and genital characters. These divisions are further supported by unpublished molecular data (Harl et al. in prep.). Furthermore, we discuss the orculid species reported from Romania. Two species, namely *Pupa (Orcula) jetschini* [@B34] and *Sphyradium dobrogicum* [@B24] were originally described from Romania. Two other species (*Orcula dolium* and *Orcula gularis*) were also reported from Romania by [@B6] and later by other authors. Distributions of the last two species (see [@B73]), unreliable data sources and inaccessible or lost voucher material make Romanian occurrence data questionable. Material and methods -------------------- The comprehensive map ([Fig. 9](#F9){ref-type="fig"}) showing the distribution of *Orcula dolium*, Alpine endemic *Orcula* spp., *Orcula schmidtii*--*wagneri*, *Orcula jetschini* and *Orcula zilchi* were compiled by literature sources ([@B4], [@B24], [@B29], [@B31], [@B36], [@B40], [@B41]; [@B42], [@B43], [@B44], [@B45], [@B46], [@B52], [@B71], [@B47], [@B64], [@B65]), museum collections (HNHM, MMM, NHMW, SMF) and personal communications: O. Gargominy (France), W. de Mattia (Italy), P. Subai (Greece, Montenegro, France, Germany). Records of shells from deposits of the Tisza River ([@B52]) were excluded because of the unreliability of their origin. Photographs of several focal planes were made with a Wild Makroskop M420 and a Nikon DS Camera Control Unit DS-L2. The different layers were combined with Helicon Focus 4.75 Pro to obtain one completely focused image. Shells were directly observed without coating under a low vacuum SEM (Miniscope TM-1000, Hitachi High-Technologies, Tokyo). Teleoconch sculpture was noted on the dorsal or dorsolateral area of the penultimate whorl. Abbreviations ------------- **HNHM** Magyar Természettudományi Múzeum (Budapest, Hungary) **MMM** Munkácsy Mihály Múzeum (Békéscsaba, Hungary) **MNINGA** Muzeul Național de Istorie Naturală "Grigore Antipa" (Bucharest, Romania) **NHMSB** Natural History Museum, Sibiu (Romania), Bielz collection **NHMSK** Natural History Museum, Sibiu (Romania), Kimakowicz collection **NHMW** Naturhistorisches Museum Wien (Vienna, Austria) **SMF** Senckenberg Forschungsinstitut und Naturmuseum (Frankfurt am Main, Germany). **SP** Collection Péter Subai (Aachen, Germany) Results ======= Systematics Family Orculidae Pilsbry 1918 ----------------------------------------- ### Genus. Orcula Held 1837 http://species-id.net/wiki/Orcula 1. OrculaHeld, Isis: 919. (1837) #### Type species. *Pupa dolium* [@B14], by subsequent designation Gray: 1847: 176. #### Diagnosis. Shell yellowish--greenish to dark brown; cylindrical to conical and elongated; 8--10 weakly convex whorls; sculpture of first 0.5--1.0 protoconch whorl usually smooth, but may be of fine spiral lines, which may be extremely weak; teleoconch axial sculpture variable, ranging from irregular growth lines to equally spaced, conspicuous radial structure; apertural barriers: one parietal and 1--3 columellar lamellae; palatal side of the aperture smooth or with strong tooth or thickening parallel to the apertural lip; parietal callus weak, subangularis sometimes present; palatalis plicae missing. Penis cylindrical, penial caecum of variable length and shape; penial appendix absent; interior of penis, epiphallus and caecum with longitudinal folds; retractor muscle attaches to the penis-epiphallus junction on the opposite side of the penial caecum; diverticulum absent; distal part of vas deferens sometimes slightly swollen, entering epiphallus terminally; bursa copulatrix long, club-like. #### Habitat. *Orcula* species occur in humid limestone areas, usually forests, or rocky boulder fields at high altitudes. Animals live under stones, leaf litter or decaying wood, or at the base of large rocks. #### Remarks. Detailed anatomical and conchological diagnoses were provided by [@B21] and [@B58]. According to [@B29], the genera *Orcula*, *Orculella* [@B63] and *Schileykula* [@B22] cannot be distinguished from each other based on conchological characters alone. Some African genera, such as *Fauxulus* Schaufuss, *Fauxulella* Pilsbry and *Anisoloma* Ancey have very similar genital tracts but usually possess sinistral shells with several apertural lamellae and denticles (see [@B58], [@B59]). In general, species of *Schileykula* and *Orculella* usually inhabit dry limestone areas in the Mediterranean. The only exceptions known are the closely related *Orculella bulgarica* (Hesse) and *Orculella aragonica* (Westerlund) which both prefer very humid, marshy stream banks ([@B20], [@B1]). ![Shells and a living specimens of Orculidae species. **a** *Orcula (Orcula) dolium* ([@B14]), Hungary, Bükk Mts., Farkasnyak, Vöröskő, leg. Németh, L., 21.07.1984 **b** *Orcula (Illyriobanatica) jetschini* ([@B34]), Romania, Jud. Bihor, Munţii Pădurea Craiului, Şuncuiuş, Valley of Crişul Repede, in front of Peştera Vantului (cave), limestone, leg. Bata, Danyik, Deli, 11.04.2011. **c** *Orcula (Hausdorfia) zilchi* [@B68], Turkey, Vil. Bursa, between Bozüyük and İnegöl, by the "Mezit 7" bridge, limestone rocks and beach forest next to the road, 580 m, 39°55.724\'N, 29°43.939\'E, leg. Páll-Gergely, B., 30.09.2007 **d** *Sphyradium doliolum* ([@B9]), Romania, Jud. Tulcea, Forest near the Cocoş Monastery, 145 m, 45°12.835\'N, 28°24.415\'E. Leg: Németh, L. & Páll-Gergely, B. 26.05.2011 **e** same locality as b. Photos: J. Harl (**a**--**d**) and T. Deli (**e**).](ZooKeys-301-025-g001){#F1} ![SEM of shells of various *Orcula* taxa. **a** protoconch of *Orcula (Orcula) dolium* (Draparnaud, 1801), Hungary, Bükk Mts., Farkasnyak, Vöröskő, leg. Németh, L., 21.07.1984 **b** protoconch of *Orcula (Illyriobanatica) schmidtii* ([@B37]), Albania,Mirditë Mts., 1 km NE of Ndërshenë, beneath the Gurit te Çikut peak, 1350 m, 41°49.952\'N, 20°06.034\'E, leg. Erőss, Fehér, Kontschán, Murányi, 21.10.2002 **c** protoconch of *Orcula (Hausdorfia) zilchi* [@B68], Turkey, Vil. Bolu, Abant Gölü N, 1030 m, 40°38.756\'N, 31°21.531\'E, leg: Páll-Gergely, B., 17.05.2006 **d** protoconch of *Orcula (Orcula) austriaca* [@B79], Austria: Niederösterreich, Piestingtal, Waldegg, 412 m, 47°52.293\'N, 16°2.722\'E, Duda, M., Haring, E., Harl, J., Kruckenhauser, L., Sattman, H, 10.09.2009 **e** columellar lamellae of *Orcula (Hausdorfia) zilchi*, locality: see figure **2c f** sculpture of *Orcula (Orcula) tolminensis* [@B72], Austria, Kärnten, Karawanken, Eisenkappel, Kupitzklamm, 674 m, 46°27.979\'N, 14°36.915\'E, leg. Duda, M., Haring, E., Harl, J., Kruckenhauser, L., Sattman, H., 29.07.2009.](ZooKeys-301-025-g002){#F2} ![SEM of the shell sculpture of various *Orcula* taxa. G *Orcula (Orcula) dolium* **H** *Orcula (Orcula) austriaca* **I** *Orcula (Illyriobanatica) jetschini*, Romania, Jud. Bihor, Munţii Pădurea Craiului, Şuncuiuş, Valley of Crişul Repede, in front of Peştera Vantului (cave), limestone, leg. Bata, Danyik, Deli, 11.04.2011. **J** *Orcula (Illyriobanatica) wagneri* [@B66], Albania, Malësia e Madhe, 11 km from Bogë, north of Tërthorës pass, 1800 m, 42°23.537\'N, 19°43.782\'E, leg. Erőss, Fehér, Kontschán, Murányi, 20.10.2002. **K** *Orcula (Illyriobanatica) schmidtii*, locality: see figure **2b L** *Orcula (Hausdorfia) zilchi*, locality: see figure **2c**.](ZooKeys-301-025-g003){#F3} ![Genital anatomy of *Orcula (Illyriobanatica) jetschini* ([@B34]), locality: see figure **1b**.](ZooKeys-301-025-g004){#F4} {#F5} ![Genital anatomy of *Orcula (Hausdorfia) zilchi* [@B68]. Bulgaria, Strandzha Mts., Kondolovo village, 42°6.150\'N, 27°39.896\'E, leg. Irikov A, 28.04.2012.](ZooKeys-301-025-g006){#F6} {#F7} ![Spermatophores. **A** *Orcula (Illyriobanatica) jetschini* ([@B34]) **B** *Orcula (Illyriobanatica) wagneri* and **C** *Orcula (Hausdorfia) zilchi* **D** egg. *Orcula (Hausdorfia) zilchi*; spermatophore and egg from different individuals. Scale = 1 mm.](ZooKeys-301-025-g008){#F8} ![Distribution map of *Orcula*. Subgenus *Orcula*: *Orcula (Orcula) dolium* (green) (green circles indicate fossil records), Alpine endemic species (blue); subgenus *Illyriobanatica*: *Orcula (Illyriobanatica) wagneri* and *Orcula (Illyriobanatica) schmidtii* (orange), *Orcula (Illyriobanatica) jetschini* (red); subgenus *Hausdorfia*: *Orcula (Hausdorfia) zilchi* (black). Number **1**: type locality of *Sphyradium dobrogicum* [@B24]; Number **2** locality of *Orcula gularis* ([@B55]) according to [@B6].](ZooKeys-301-025-g009){#F9} ### Subgenus. Orcula #### Diagnosis. Shell smoothish with irregular growth lines; apex somewhat conical, not blunt; aperture with 2--3 columellar lamellae; penial caecum simple and usually longer than half the length of the penis; its base often not conspicuously thickened. #### Content. *austriaca*, *conica*, *dolium*, *fuchsi*, *gularis*, *pseudodolium* [@B72], *restituta*, *spoliata*, *tolminensis*. #### Remarks. The soft anatomy of various *Orcula* taxa has been described in the following papers: *austriaca* ([@B21], [@B59]), *austriaca faueri* [@B35] ([@B21], [@B59]), *austriaca pseudofuchsi* [@B35] ([@B21], [@B59]), *conica* ([@B61], [@B21], [@B59]), *dolium* ([@B60], [@B61], [@B63], [@B21], [@B69], [@B25], [@B53], [@B57], [@B58], [@B59]), *dolium brancsikii* [@B12] ([@B53]), *dolium edita* Ehrmann, 1933 ([@B59]), *dolium gracilior* [@B79] ([@B21], [@B59]) *dolium infima* Ehrmann, 1933 ([@B59]), *dolium pseudogularis* A. J. Wagner, 1912 ([@B21]), *fuchsi* ([@B21], [@B59]), *gularis* ([@B61]; republished by [@B25], [@B21], [@B59]), *pseudodolium* ([@B21]), *restituta* ([@B21]), *spoliata* ([@B21], [@B59]), *tolminensis* ([@B21]). The penial caecum of *Orcula (Orcula) restituta* is very short compared to other *Orcula (Orcula)* species, but the shell is similar to that of *Orcula (Orcula) gularis*. Prior to [@B35], *restituta* was considered a subspecies of *gularis*. A third columellar lamella is rarely present, but can occur in a small percentage of individuals within a population. [@B8]: 84) noted a third columellar fold in only one individual of thousands in each *Orcula dolium titan* ([@B8]) and *Orcula dolium dolium*. #### Distribution. Most species have limited distributions in the Alps (mainly Austria). *Orcula dolium* is widely distributed in Central Europe, in the Alps (eastern France, Switzerland, Southern Germany, Northern Italy, Austria, Slovenia, Northern Croatia, and Slovenia) and the Western Carpathians (Northern Hungary, Slovakia, Eastern Czech Republic). The Croatian records of *Orcula dolium* and *Orcula gularis* ([@B64], [@B65], [@B79]) have not been verified by recent investigations. Our knowledge of the distribution of *Orcula dolium* is distorted due to misidentified material. Probably all reports of this species (living and fossil) from Spain (e.g. [@B39]) refer to *Orculella aragonica* (see [@B1]). Italian (Toscana) records ([@B76], [@B77]) refer to a yet unknown *Orculella* species (see photo in [@B77]). [@B13] reported *Orcula dolium* from the Balkan Peninsula, South, Central and West Europe, the Crimea, Western Ukraine, Central Asia, Tunisia, Ethiopia and northen Iran. This distribution is much broader than that of *Orcula dolium* and probably refers to the distribution of the family Orculidae. [@B38] and [@B67] speculated that *Orcula dolium* occurs in Ukraine. This supposition has been included in distribution maps ([@B73]), but to date the taxon's occurrence in Ukraine has not been verified data ([@B3]). [@B62] mentioned that during careful collections around Munkács (Mukachevo, southwest Ukraine), Traxler was not able to find the species. *Orcula dolium* was more widely distributed during the Pleistocene. The northernmost localities were published by [@B40] (Czech Republic, ca, 30 km north of Prague) and [@B44] (Germany, northern Baden-Württemberg). The southernmost locality was reported by [@B42] from the Serbian Kisiljevo. [@B56] described *Orcula dolium* var. *pliopedemontana* from the middle Pliocene sediments at Ceresole d'Alba (Italy: ''Villafranchiano''). The description is unfortunately insufficient and the taxonomic position of this form is uncertain ([@B19], [@B51]). More recently, [@B11] made no mention of the species from the same sediment layers. ### Orcula (Illyriobanatica) Páll-Gergely & Deli subgen. n. #### Type species. *Pupa (Orcula) jetschini* M. von [@B34]. #### Diagnosis. Shell usually with strong axial sculpture (irregular ribs), with two columellar lamellae, apex rather rounded, not attenuate. The penial caecum usually consists of two parts (''tubercles'') and its length is less than half that of the penis. #### Etymology. The name of this new subgenus refers to its distribution in the Illyrian and Banatic biogeographical regions. It is feminine. #### Content. *jetschini*, *schmidtii* and *wagneri*. #### Distribution. Montenegro, Albania, Northwestern Greece, Kosovo, Macedonia (*Orcula wagneri* and *Orcula schmidtii*) and the western part of Romania (*Orcula jetschini*). #### Remarks. The reproductive anatomy of *Orcula schmidtii transversalis* ([@B75]) was described by [@B28] and [@B54]. According to [@B28], the penial caecum of *Orcula schmidtii transversalis* is short, but simple. The caecum appears double in the illustration provided by [@B54]. ### Orcula (Illyriobanatica) jetschini (Kimakowicz 1883) http://species-id.net/wiki/Orcula_jetschini 1. Pupa (Pupilla) dolium, Bielz, Fauna der Land- und Süsswasser-Mollusken Siebenbürgens, 89. (1863) 2. Pupilla dolium, --- Bielz, Fauna der Land- und Süsswasser-Mollusken Siebenbürgens: 94--95. (1867) 3. Pupa (Orcula) jetschiniKimakowicz, M. von. --- Verhandlungen und Mittheilungen, 33: 44--46. (1883) 4. Orcula jetschini, --- Pilsbry: Manual of Conchology: 5, 17., Plate 2, fig. 10--11 . \[''Transylvania, restricted to the southwestern part: Vajda-Hunyad and Bad Gyogy (Kimakowicz), Judenberg near Zalatna (Jickeli). Cerna valley at Mehadia in the Banat (Jetschin) ''\]. (1922) 5. Orcula jetschini, --- Soós: A Kárpát-medence Mollusca faunája: 155--156, plate 6, fig. 18. (1943) 6. Orcula jetschini, --- Kerney et al. Die Landschnecken Nord- und Mitteleuropas: 102--103. (1983) 7. Orcula (Orcula) dolium(partim: citation of [@B6]), --- Grossu, Gastropoda Romaniae 2: 221--223. (1987) 8. Orcula (Orcula) jetschini, --- Grossu, Gastropoda Romaniae 2.: 223--224. (1987) 9. Orcula jetschini, --- Harl et al. Archiv für Molluskenkunde, 140 (2): 184, Plate 4, fig. J. (2011) 10. Orcula jetschini, --- Welter-Schultes, European non-marine molluscs: 145. (2012) #### Material. RO, Jud. Bihor, Munţii Pădurea Craiului, Şuncuiuş, Valley of Crişul Repede, in front of Peştera Vantului (cave), limestone, leg.: Bata, Danyik, Deli, 11.04.2011. (anatomically examined); RO, Gyalui-havasok (Munţii Gilăului), Runki szakadék (gulch of Runk), leg. Papp, J., 22.07.1959, HNHM 73030/3; RO, Bihar Mts (Munţii Apuseni)., Felsőgirda (Gârda de Sus), Ordincus valley., leg. Kovács, Gy., 30.05.1985, HNHM 68284/2; RO, Muntii Bihorului, Baita, Piatra Graitoare, environment of the Crisu Baitei River, leg. Kovács, Gy., 23.08.1974, HNHM 68283/2; Černath. (Chernathal) bls Badern (?) v. Mehadia, leg. Jetschin 1882, NHMSK 4874/5; Forstgra (Forstgartens) bei im Černathal, Banat, leg. Jetschin 1882, NHMSK 4875/2; Klausenburg, Györgyfalvaer Wald, leg. Marzlof 1891, NHMSK 7470/8; Zalathna gegen den Judenberg, leg. Barth 1866--1906, NHMSK 7468/8; Banat, Herkulesbad, leg. Deubel 1895 May--Juni, NHMSK 7469/6; Steierdorf bis zur Höhle Panur, leg. Jetschin 1885, NHMSK 4876/4; Černathal b. Mehadia, leg. Jetschin 1885, NHMSK 4877/4 (''*mut. albina*''); Gyógybad nächst Broos. Orm. (?) 1887, NHMSK 4873/3; Hideg-Szamos, NHMSB 51/72, 50136--50137; Györgyfalvaer Wald b. Klausenburg, NHMSB 51/82, 50389--50391; Unter-Grohob bei Körösbánya NHMSB 51/92, 50713; Klausenburg, Bükk, NHMSB 51/13, 48225--48226; Klausenburg ?? (not legible on label) Wald, NHMSB 51/23, 48500; RO, Bihor Mts., Valea Boghii (valley), 46°36.610\'N, 22°39.542\'E, leg. Páll-Gergely, B. 09.08.2007.; RO, Jud. Bihor, Bălnaca Groşi, cliffs at Bíró Lajos cave., leg. Domokos, T. 18.04.2004, MMM 04503/1; Munţii Pădurea Craiului, Şuncuius valley of Crişul-Repede, under shrubs., leg. Domokos, T. & Deli, T., 08.07.2005, MMM 04505/2; Munţii Pădurea Craiului, Şuncuiuş valley of Mişid-brook near brook-Alnetum., leg. Domokos, T.& Deli, T., 08.07.2005, MMM 04506/3; Munţii Apuseni, Gârda de Sus, Ordincus valley., leg. Domokos, T. & Kovács, Gy., 30.05.1985, MMM 04501/2; Munţii Apuseni, Gârda de Sus, Ordincus valley., leg. Domokos, T. & Deli, T., 30.11.2009, MMM 92488/1; Munţii Zărandului, Troaş, Pietroasia, floating debris., leg. Domokos, T., 07.06.2002, MMM 04502/10; Munţii Zărandului, Troaş, Valea Galsa floating debris, leg. Domokos, T. et al., 27.05.2005, MMM 04504/11; Jud. Arad, above Obârşia (Munţii Metaliferi) (1.9 km W of Arad-Hunedoara board), forest clearing (*Corylus*), 700 m, leg. Deli, T. & Domokos, T, 07.03.2007, MMM 90866/2; Jud. Arad, between Pojoga and Căprioara (7 km SE Săvârsin), in gorge--forest, 120 m, leg: Deli, T., Domokos, T., Páll-Gergely, B., Subai, P., 15.04.2007, MMM 91092/2; Jud. Arad, between Pojoga and Căprioara (7 km SE Săvârsin), in gorge--forest, 120 m., leg: Deli, T., 12.06.2007, MMM 90814/1; Jud. Caraş-Severin, between Moldova Nouă and Padina Mate, forest (*Fagus*, *Caprinus*, *Ruscus*) with limestone rocks, 300 m, leg. Boldog, G., Deli, T., Kóra, J., 04.07.2007, MMM 92489/1; Mehadia Mts. (Munţii Mehedinţi), Cerna valley, Jelărăului gorge, above Băile Herculane, large flotsam deposit, leg. Boldog, G., Deli, T., Kóra, J., 08.07.2007, MMM 92494/6; Munţii Mehedinţi, Cerna valley, Jelărăului gorge, above Băile Herculane, large flotsam deposit, leg. Deli T., Horváth, É., Lennert, J., Páll-Gergely, B., Subai, P., 04.05.2008, MMM 92490/5; Munţii Vâlcan, N Tismana, near Monastery Tismana, bank of Tismana brook, flotsam deposit., leg. Deli, T., Domokos, T., Páll-Gergely, B., Subai, P., 15.04.2007, MMM 92491/1; Munţii Vâlcan, Piscuri-valley, 1,4km N Vâlcele (NE Tismana), flotsam deposit., leg. Deli T., Domokos T., Páll-Gergely B., Subai, P., 17.04.2007, MMM 92492/1; Munţii Vâlcan, N of Runcu, Cheile Sohodol, 4.5 km upstream of the gorge entrance, limestone walls., leg. Boldog, G., Deli, T., Kóra, J., 06.07.2007, MMM 92493/1. #### Description of the genitalia. Two specimens were anatomically examined. Penis slim, with the retractor muscle attached at its distal end; penial caecum very small, vestigial, consisting of two "tubercles"; epiphallus very long and cylindrical; there is clear distinction between the vas deferens and the epiphallus; vas deferens long and relatively thick; a slender retractor muscle is attached near the proximal end. Vagina short and thick, but pedunculus relatively long; bursa copulatrix extremely long, with the distal end slightly expanded. In one specimen an elongated, simple spermatophore was found with the apical portion slightly thickened. #### Distribution. *Orcula (Illyriobanatica) jetschini* is known only from western part of Romania (Banat, Crişana and Western Transylvania). The Hungarian record ([@B52]) is apparently based on a flotsam specimen so its origin is suspect ([@B70]). [@B46] reported the species from Moldova Nouă, which lies close the Serbian border and, therefore, it is expected to occur in Serbia. The Pleistocene distribution of *Orcula dolium* included the ''Požarevac Danube Area'' ([@B42]), which is just on the other bank of the Danube River, but temporal and spatial sympatry of *Orcula (Illyriobanatica) jetschini* and *Orcula (Orcula) dolium* is not verified. #### Ecology. The species inhabits deciduous forests. It is found most commonly between small stones and leaf litter on the forest floor or under hazelnut (*Corylus*) bushes. The species is known from non-limestone bedrock, such as the Zarand Mountains. #### Conservation status. Least concern (LC) according to IUCN criteria ([@B17]). #### Remarks. All living specimens found were covered in mud, causing them to appear like tiny grains of soil. The ribbed shell is possibly an adaptation for camouflaging. The photographs herein are of cleaned shells. ### Orcula (Illyriobanatica) wagneri Sturany 1914 http://species-id.net/wiki/Orcula_wagneri 1. Orcula wagneriSturany in Sturany and Wagner, --- Denkschriften der Kaiserlichen Akademie der Wissenschaften, mathematisch-naturwissenschaftliche Klasse 91: 45, Plate 15, fig. 82b. (1914) 2. Orcula wagneri, --- Harl et al., Archiv für Molluskenkunde, 140 (2): 186, Plate 5, fig. A--G, J. (2011) 3. Orcula wagneri, --- Audibert, Folia Conchyliologica 14: 21--25. Figure 1: habitat, figure 2, and Figure 1, 2, 4, 5: shells with possible signs of parasites. (2011) 4. Orcula wagneri, --- Schileyko, Ruthenica, 22 (2): 152-253, 156, figure 17 (genitalia). (2012) 5. Orcula wagneri, --- Welter-Schultes, European non-marine molluscs:146. (2012) #### Material. Albania, Bjeshkët e Nemuna (Prokletije Mts), above village Okol, near pass Qafa e Pejës, W slope of Mt. Maja e Popluks, at a spring on limestone, 1660 m, 42°27.343\'N, 19°46.478\'E, leg. Barina, Z., Puskás, G., Sárospataki, B., 16.07.2010., HNHM 98841. #### Description of the genitalia. One specimen was dissected. Penis cylindrical and slim, with a short, but thick penial caecum, the proximal portion broader than the short and slimmer distal portion; retractor muscle attaches at the penis--epiphallus transition; epiphallus more than twice as long as the penis and much thicker, its transition to the vas deferens is gradual, barely discernable; there is a slim retractor muscle attached to the proximal portion of the epiphallus; proxim\\al portion of the vas deferens thicker than the distal part. Vagina and free pedunculus extremely short; bursa copulatrix almost twice as long as the combined length of the penis--epiphallus complex. #### Conservation status. *Orcula (Orcula) wagneri* is listed as Near Threatened (NT), being close to the criteria threshold for Vulnerable ([@B48]). #### Remarks. Our observations on the genitalia agree with that of [@B59], who investigated the anatomy of *Orcula wagneri* from the Tomor Mountains (''Maja e Tomorit Mt., S Albania''). A partially reabsorbed, elongated, spermatophore was located in the bursa copulatrix ([Fig. 8B](#F8){ref-type="fig"}). ### Orcula (Hausdorfia) Páll-Gergely & Irikov subgen. n. #### Type species. *Orcula zilchi* [@B68] (by monotypy). #### Diagnosis. Shell with conical apex and strong axial sculpture (irregular axial growth lines), with three columellar lamellae (columellar, supracolumellar and one short lamellae above), palatalis reaches its maximum height on the dorsolateral side. Penial caecum very long with thickened base, canal connecting the proximal end of the epiphallus to the penial caecum. #### Etymology. The new subgenus is named in honour of Dr Bernhard Hausdorf (University of Hamburg), who first noted the unusual shell characters of *Orcula zilchi* and questioned its generic status ([@B29]). It is feminine. #### Distribution. See under *Orcula (Hausdorfia) zilchi*. #### Remarks. According to [@B59], the penial caecum of *Orcula fuchsi* is long and slender, with a thickened base, making it similar in morphology to *Orcula zilchi*. However, the characteristic canal connecting the proximal end of the epiphallus with the penial caecum of *Orcula zilchi* is lacking in *Orcula fuchsi*. The long caecum of this species is also illustrated by [@B21], but its base is not conspicuously thickened. This may vary between populations or during an individual's life history. ### Orcula (Hausdorfia) zilchi Urbański 1960 http://species-id.net/wiki/Orcula_zilchi 1. Orcula zilchiUrbański, J., Bulletin de la Société des Amis des Sciences et des Lettres de Poznan (Série D) 1: 57. \[''Am rechten Ufer des Ropotamo, etwa 3 km vor seiner Mündung (etwa 30 km südlich von Burgas) ''\]. (1960) 2. Orcula zilchi, --- Damjanov and Likharev, Fauna Bulgarica, Gastropoda terrestria, vol. IV: 115. (1975) 3. Orcula(?) *zilchi*, --- Hausdorf, Archiv für Molluskenkunde125 (1/2): 14, Plate 1, fig. 1. \[''Westanatolien: V. Kütahya, Safa 2 km R Domaniç''\]. (1996) 4. Orcula zilchi, --- Páll-Gergely, Zoology in the Middle East, 50: 91. (2010) 5. Orcula zilchi, Harl et al. --- Archiv für Molluskenkunde, 140 (2): 186--187, Plate 4, fig. F, G. (2011) 6. Orcula zilchi, --- Welter-Schultes, European non-marine molluscs:146. (2012) #### Material. Bulgaria, Strandzha Mts., Kondolovo village, 42°6.150\'N, 27°39.896\'E, leg. Irikov, A., 28.04.2012. (anatomically examined); Bulgaria, Silkosiya Reserve, near Kosti Village, 23.06.2001, leg., А. Irikov; Turkey, Vil. Bolu, Abant Gölü N, 1030 m, 40°38.756\'N, 31°21.531\'E, leg. Páll-Gergely, B., 17.05.2006.; Turkey, Vil. Bursa, between Bozüyük and İnegöl, by the ''Mezit 7'' bridge, limestone rocks and beach forest next to the road, 580 m, 39°55.724\'N, 29°43.939\'E, leg. Páll-Gergely, B., 30.09.2007; Bulgaria, floating debris 6 km N of Malko Tarnovo, 210 m, UTM NG 45, 42°5.028\'N, 27°25.698\'E, leg. Dedov & Subai 8.5.2008, SP 22168/2 (juv.) #### Description of the genitalia. Two specimens were dissected. Penis cylindrical, relatively long; retractor muscle short, attaches on the proximal portion; penial caecum very long, with a thickened base and a cylindrical distal portion; an additional canal (?) connects the proximal end of the epiphallus with the penial caecum; epiphallus long, with a thickened distal part; the separation between the vas deferens and epiphallus is distinct; vas deferens relatively thick. Vagina cylindrical and relatively short; bursa copulatrix extremely long with a pointed end. A developing egg covered with small calcareous crystals was found in the uterus of the figured specimen. In the other specimen, an elongated, simple bursa copulatrix was found with a slightly thickened apical part. #### Distribution. South-Eastern Bulgaria and North-Western Turkey. #### Ecology. The type series (12 shells) of *Orcula zilchi* was collected by Urbański on the floodplain of the Ropotamo River in leaf litter and under decaying wood. It was found in association with *Sphyradium doliolum* ([@B9]), *Oxychilus deilus rumelicus* Hesse, *Laciniaria plicata* (Draparnaud), *Bulgarica denticulata thessalonica* (Rossmässler), *Euxina persica paulhessei* (Lindholm), *Euxina circumdata* (L. Pfeiffer), *Cochlodina laminata* (Montagu). Atanas Irikov visited the type locality (very humid forest with rocks along the river) on two occasions, with collection time totalling 6--8 hours. Besides *Orcula zilchi* he collected all other species previously reported from the Ropotamo area. We were able to find *Orcula zilchi* only in deciduous forests. In Bulgaria (near Kondolovo village), living specimens were collected in an oriental beech (*Fagus orientalis*) forest in shady and moist microhabitats between the leaf litter and soil. These conditions were very similar to the Abant Gölü locality (Turkey). The other Turkish locality (between Bozüyük and İnegöl) was slightly different, with a deciduous forest at the base of a large limestone rock, on a slope covered with smaller stones and larger rocks. The species is very rare wherever it has been encountered yet, especially in Turkey. On two occasions, in 2007 and 2010, Barna Páll-Gergely spent about 4--5 hours at the locality in vil. Bursa, but found only one specimen in 2007. The other locality (Vil. Bolu) was visited in 2005 and 2006 for similar lengths of time and only one specimen was found in 2006. Atanas Irikov collected 9 living specimens and about 10 empty shells in an hour near Kondolovo in Bulgaria. #### Conservation status. Listed as Vulnerable (V) under IUCN criteria ([@B49]). Deforestation and disturbance of the forests are the main threat to this species. #### Remarks. Two of four living specimens had beetle (possibly drilid beetle) larvae in the body whorl. The dissected specimens were collected about 23 km south-southwest of the type locality. The Strandzha Mountains (incl. the collecting site) belongs to the drainage of the Ropotamo River. It is reasonable to suppose that Urbański's population was "washed down" from somewhere in the Strandzha Mts. and settled a temporary subpopulation in the Ropotamo floodplain. This might be a reasonable explanation why A. Irikov could not find this species in the type locality. ### Genus. Sphyradium Charpentier 1837 http://species-id.net/wiki/Sphyradium 1. Sphyradium doliolum([@B9]) 2. Bulimus doliolumBruguière Encyclopédie méthodique: 351. (1792) 3. Sphyradium dobrogicumGrossu, **new synonym** Travaux du Muséum d'Histoire Naturelle ''Grigore Antipa'' 28: 7--13. Bucureşti. \[Dobrogea, département de Tulcea, près du Monastère Cocoş de la Forêt Luncaviţa.\] (1986) 4. Sphyradium dobrogicum, --- Grossu: Gastropoda Romaniae 2: 228--230, fig. 120. \[''pădurea Luncaviţa în apropierea Mănăstirei Cocoş, jud. Tulcea''\]. (1987) 5. Orcula dobrogica, --- Welter-Schultes, European non-marine molluscs:143. (2012) #### Remarks. *Sphyradium dobrogicum* was described based on a single shell. The holotype could not be located in the collection of the Grigore Antipa National Museum of Natural History (Bucharest) during a recent search (2012). It could still be in Grossu's house (Oana Popa, pers. comm., 2011) but, at present, the holotype seems to be lost. [@B5] and [@B73] assigned the species to *Orcula* without supporting evidence. *Sphyradium dobrogicum* has a domed apex, a ribbed shell and very weak lamellae (see original description and drawing), indicating that it may represent a dwarf specimen of *Sphyradium doliolum*. We visited the type locality of *Sphyradium dobrogicum* in 2011 but found only *Sphyradium doliolum*. Based on available information we suggest using *Sphyradium dobrogicum* as a synonym of *Sphyradium doliolum*. *Orcula gularis* and *Orcula dolium* in Romania ----------------------------------------------- ***Orcula gularis***: [@B6], [@B7]) reported *Pupa (Pupilla) gularis* from Gușterița, which is presently part of Sibiu. [@B25], [@B26]) cited this record in his account of *Sphyradium gularis*. [@B24] discussed Bielz's specimens housed in the museum in Sibiu, but specimens were not located in the collection of Bielz in the NHMS. Although [@B5] reports the species from Romania, recent authors consider this record as erroneous ([@B16]) or simply ignore it ([@B73]). Indeed, the occurrence of *Orcula gularis* in Romania, more than 650 km from its main distribution area is very unlikely. However, to our knowledge, no one has searched for the species at the respective Romanian locality which was well-defined by Bielz. Nevertheless, despite intensive collecting efforts by competent malacologists over the last 150 years, *Orcula gularis* has not been encountered in Transylvania. *Orcula gularis* is a very characteristic species with a strong palatal tooth in the aperture. Today, with literature available it is very difficult to misidentify *Orcula jetschini* as *Orcula gularis* Based on recent available literature, these two species easily can be distinguished from each other. Bielz also had non-Romanian comparative material of *Orcula gularis* at his disposal: Krain (NHMSB 139634--139635), Oberkrain (NHMSB 139636--139639), Karnthen (NHMSB 139640--139641), Hohewand (NHMSB 139642--139644) and Austria (NHMSB 139645--139646) (Ana Mesaroş, pers. comm.). On the other hand, the understanding of the Orculidae was insufficient at that time. For example, Bielz confused *Orcula jetschini* with *dolium* (see below). Perhaps, a thickening behind the aperture lip of the examined specimens led Bielz to misidentify them as *Orcula gularis* instead of *Orcula jetschini*, the species inhabiting that area. ***Orcula dolium***: *Orcula dolium* has been recorded from Romania by several authors. [@B6] reported it from ''Vajda-Hunyad am Kaczanyas'' (Hunedoara), ''Valea-Ordinkusi bei Skerisora'' (Ordâncuşa valley at Scărișoara), ''nördlich von Unter-Grohot bei Körösbánya'' (Baia de Criș), ''Vormága'' (Vărmaga) and ''Collegiumwald von Nagy-Enyed'' (Aiud). In the collection of Bielz (NHMS) we only found the sample from Körösbánya, one sample from ''Hideg-Szamos'' and more samples from Klausenburg (Cluj Napoca). [@B33] reported *Orcula dolium* from a few localities in the Bihor Mountains, and [@B25] most recently from a number of localities, namely from northern Oltenia and the Banat area, and a remote locality in Tulcea County (Luncaviţa, around the Cocoş Monastery). [@B62] speculated that records of *Orcula dolium* by Bielz and [@B33] from the Apuseni Mountains should be assigned to *Orcula jetschini*. In fact, Bielz revised his original labels from *Pupa dolium* to *jetschini* (NHMS). Two Romanian samples of Grossu in the collection of MNINGA are labelled as *Orcula dolium*: MNINGA GST/923, Horezu, Valcea, leg. Grossu (3 shells) and MNINGA 28142, Bucegi, ''Sertarul 114'', ex Licherdopol (''*Orcula dolium* var. *implicata*'', 3 shells). Both samples are actually *Sphyradium doliolum*. We cannot explain the Luncaviţa locality (see remarks under *Sphyradium dobrogicum*). We collected near the localities mentioned by Grossu (Oltenia and Banat) and found only *Orcula jetschini*. [@B42] reported *Orcula dolium* from the Pleistocene of the Serbian Kostolac, very close to Grossu's localities. It is possible that some of the specimens Grossu examined are Pleistocene fossils, but it seems unlikely that *Orcula dolium* still lives in the Banat-Oltenia area in Romania. [@B62] marked Torna (southeast Slovakia) as the easternmost locality of *Orcula dolium*. Discussion ========== In this paper we describe the genitalia of the Eastern European *Orcula jetschini* and *Orcula zilchi* for the first time. We also examine and describe the anatomy of *Orcula wagneri* from a locality that lies 200 km north of populations examined by [@B59]. This additional information and data published by other authors (mainly [@B22] and [@B59]) allows us to review the taxonomic relationships of the entire genus. The genus can be subdivided into three subgenera (*Orcula*, *Illyriobanatica* subgen. n. and *Hausdorfia* subgen. n.) based on the shell characters and the morphology and size of the penial caecum, which serves as the primary diagnostic character. Unpublished results of molecular phylogenetic analysis (Harl et al. in prep.) of most *Orcula* species and subspecies indicate the monophyly of these three groups. This subdivision is in good agreement with biogeographic information. Three species included herein have some shell and anatomical characters which differ from characters used to the features mentioned in the diagnoses of certain subgenera: (1) the shell sculpture of many populations of *Orcula (Illyriobanatica) wagneri* is almost smooth, which is unusual in the subgenus. (2) The penial caecum of *Orcula (Orcula) restituta* is very short compared to other species assigned to the subgenus. (3) The penial caecum of *Orcula (Illyriobanatica) schmidtii transversalis* is short, but simple ([@B28]), not ''tuberculated'' as the other forms of the subgenus. The caecum of *Orcula schmidtii transversalis* seems to be double in the illustration of [@B54]. Based on available literature, the occurrence of *Orcula (Orcula) dolium* and *Orcula (Orcula) gularis* in Romania is discussed. Most purported Romanian ''voucher specimens'' are lost or we were unable to examine them. As the published literature is based on possibly misidentified specimens, and the verified distributional ranges of *Orcula gularis* and *Orcula dolium* lie far from the Romanian records, we suggest deleting *Orcula gularis* and *Orcula dolium* from the Romanian faunal list. The occurrence of only one *Orcula* species, namely *Orcula (Illyriobanatica) jetschini* is verified from Romania. Supplementary Material ====================== ###### XML Treatment for Orcula ###### XML Treatment for Orcula ###### XML Treatment for Orcula (Illyriobanatica) ###### XML Treatment for Orcula (Illyriobanatica) jetschini ###### XML Treatment for Orcula (Illyriobanatica) wagneri ###### XML Treatment for Orcula (Hausdorfia) ###### XML Treatment for Orcula (Hausdorfia) zilchi ###### XML Treatment for Sphyradium We are very grateful to those colleagues who provided access to their museum collections: A. Eschner (NHMW), E. Neubert (NMBE), R. Janssen (SMF), A. Mesaroş (NHMS), Z. Fehér (HNHM), to O. Merkl (HNHM) for determining the beetle larvae found in *Orcula zilchi*, to T. Gaudényi, P. Subai, V. Štamol, W. de Mattia, I. Fritzsche, O. Gargominy, B. Lecaplain, A. Thomas, A. Wagner, M. Szekeres, O. Popa (MNINGA) for providing information, K. Auffenberg (Florida Museum of Natural History) for correcting the English and for his valuable advices, to Z. Fehér and B. Hausdorf for reviewing the manuscript, to László Németh for help in the field and to the Biodiversity Heritage Library for the multitude of rare literature made available to us ([www.biodiversitylibrary.org](http://www.biodiversitylibrary.org)). [^1]: Academic editor: Eike Neubert

Introduction ============ Almost all available antidepressants bring about their effects by increasing monoamine neurotransmission, and many drugs that increase monoamines in the synaptic cleft have been shown to have antidepressant properties.^[@bib1]^ Despite the considerable noise in placebo-controlled clinical trials, such trials showed a statistically significant advantage for switching patients with selective serotonin reuptake inhibitor (SSRI)-resistant depression to a non-SSRI rather than another SSRI antidepressant,^[@bib2]^ which suggests important interindividual variation in the response to specific monoaminergic drugs. As a result, biomarkers predicting outcomes of specific monoaminergic drug classes have the potential to reduce the current trial-and-error method that commonly delays effective treatment. However, until now no such marker has been consistently identified for any monoaminergic antidepressant class. Thus, studies are needed that build a framework for guiding the selective, personalized antidepressant therapy by relating clinical symptoms and brain circuitry responses to serotoninergic and catecholaminergic neurotransmission. In this study, we applied tryptophan depletion (TD) and catecholamine depletion (CD) to elucidate the common and differential symptoms and regional cerebral glucose metabolic changes these challenges induce in subjects with a history of major depressive disorder (MDD). The study compared data acquired in two previously published experiments from the same laboratory that used the identical neuroimaging procedure with positron emission tomography (PET) and \[^18^F\] fluorodeoxyglucose (^18^FDG) after TD or CD in subjects with fully remitted MDD (rMDD). TD, which putatively lowers central serotonergic transmission, was induced by depleting the serotonin precursor, tryptophan, through oral loading with all essential amino acids, except tryptophan. CD, which is expected to reduce central dopamine and norepinephrine neurotransmission, was achieved by administering α-methyl-paratyrosine (AMPT),^[@bib3]^ a competitive inhibitor of tyrosine hydroxylase, the rate-limiting enzyme in the synthesis of catecholamines.^[@bib4]^ CD has been shown to induce depressive symptoms in a relatively high proportion of subjects with rMDD, but generally does not affect mood in healthy controls.^[@bib5]^ On the basis of previous studies suggesting functional interactions between serotonin and norepinephrine neurons in animal models of depression,^[@bib6],\ [@bib7],\ [@bib8]^ reductions of serotoninergic and catecholaminergic neurotransmission by means of TD and CD are likely to induce both common and distinct effects on the spectrum of depressive symptoms.^[@bib3],\ [@bib6],\ [@bib9],\ [@bib10],\ [@bib11],\ [@bib12],\ [@bib13],\ [@bib14],\ [@bib15],\ [@bib16],\ [@bib17],\ [@bib18],\ [@bib19]^ Previous neuroanatomical findings of serotonin,^[@bib20]^ norepinephrine^[@bib21]^ and dopamine^[@bib22]^ have suggested that monoaminergic neurotransmission is involved in a wide range of cerebral functions including cognition, attention, mood, reward processing, appetite and sleep. As a result, deficiency of monoamines may conceivably explain the wide range of depressive symptoms including cognitive dysfunction, depressed mood and appetite and sleep disturbances. Functional neuroimaging studies have associated the reductions of specific monoamines with changes in hemodynamic or metabolic activity in distinct cerebral networks.^[@bib23]^ The limbic--cortical--striatal--pallidal--thalamic network is of particular interest. This network connects the orbitofrontal cortex (OFC), medial prefrontal cortex (PFC), amygdala, hippocampus, ventromedial striatum, ventral pallidum and the mediodorsal and midline thalamic nuclei^[@bib23],\ [@bib24],\ [@bib25],\ [@bib26]^ that showed altered neurotransmission under the depletion of serotonin^[@bib27]^ and catecholamines.^[@bib28]^ In an exploratory fashion, the current study investigated how this altered neurotransmission differed quantitatively between TD and CD, and how these differences related to the type and severity of induced symptoms. Materials and methods ===================== The current study compared two previously published experiments. The first study applied TD,^[@bib27]^ the second study used CD in subject samples selected via the same entrance criteria.^[@bib28]^ Participants ------------ Both studies (TD and CD) used the identical study procedure with respect to the study design and participant selection, that is, a double-blind, placebo-controlled crossover design in fully remitted, unmedicated depressed patients (rMDD). TD compared the effects of TD versus placebo and CD compared the effects of CD versus placebo. During the depletion procedures, the cerebral glucose metabolism was measured by PET and ^18^FDG. The experimental group in both studies comprised individuals aged 18--56 years who met DSM-IV criteria for MDD in full remission (rMDD). The healthy controls had no history of any psychiatric disorder and no major psychiatric disorder in first-degree relatives. Diagnosis was established by the Structured Clinical Interview for DSM-IV^[@bib29]^ and confirmed by an unstructured interview with a psychiatrist. The subjects were recruited through the outpatient clinical services of the National Institute of Mental Health and by advertisements in local newspapers and posters on the National Institutes of Health campus. Exclusion criteria included major medical illnesses, pregnancy, psychotropic drug exposure (including nicotine) within 3 months, substance abuse within 1 year, lifetime history of substance dependence, psychiatric disorders other than MDD and structural brain abnormalities on magnetic resonance imaging (MRI). Inclusion criteria required that rMDD subjects had remained in remission without medications for at least 3 months and had manifested depression onset before 40 years of age. Written informed consent was obtained as approved by the institutional review board of the National Institute of Mental Health. With the exception of two rMDD subjects and five healthy controls who participated in both studies, the TD and CD study comprised independent subject samples. The TD study included 28 rMDD subjects (19 women, 9 men) and 27 healthy controls (18 women, 9 men). The CD study comprised 17 rMDD subjects (16 women, 1 man) and 13 healthy controls (12 women, 1 man). There were significantly more men in the rMDD group in the TD compared with the CD study (*P*=0.04). No PET data were obtained in one subject with rMDD and one healthy control subject in the TD study, and in two subjects with rMDD in the CD study. Tryptophan depletion -------------------- Subjects underwent two identical sessions that were separated by at least 8 days to avoid carryover effects. TD was induced by administration of 70 white capsules containing an amino-acid mixture consisting of isoleucine (4.2 g), leucine (6.6 g), lysine (4.8 g), methionine (1.5 g), phenylalanine (6.6 g), threonine (3.0 g) and valine (4.8 g) at 0700 hours (see Neumeister *et al.*^[@bib27]^ for details). Placebo administration at 0700 hours comprised 70 white capsules with 31.5 g of lactose. Patients were restricted from eating upon completion of PET at about 1600 hours on day 1. Behavioral measures included the Hamilton Scale of Depression (HAMD), the Montgomery--Åsberg Depression Rating Scale (MADRS) and the Beck Anxiety Inventory (BAI). Study raters were blinded. Catecholamine depletion ----------------------- Subjects underwent two identical sessions separated by at least 1 week, in which they received either a body-weight-adjusted AMPT dose or placebo (see Hasler *et al.*^[@bib28]^ for details). To reduce risk of adverse reactions, a body-weight-adjusted AMPT dose of 40 mg kg^−1^ of body weight orally, to a maximum of 4 g, over 22 h was used. Each session took 3 days and was performed on an inpatient basis at the National Institutes of Health Clinical Center. To reduce the risk of crystalluria during AMPT administration, subjects received sodium bicarbonate, drank at least 2 l of water daily and underwent urinalysis twice daily. Behavioral measures included the HAMD, the MADRS and the BAI. Study raters were blinded. Statistical analysis of behavioral data --------------------------------------- To compare the depletion effects of TD and CD on behavioral measures, differences in behavioral measures (ΔHAMD, ΔMADRS and ΔBAI) between challenge and placebo were calculated first for each subject and time point. These behavioral differences (ΔHAMD, ΔMADRS and ΔBAI) were then modeled with full factorial linear mixed models with restricted maximum likelihood estimations to account for the repeated measurements in the same subjects. Schwarz\'s Bayesian criteria were used to determine the best fitting covariance structure for each set of measures in cases where the typical compound symmetry approach used by analysis of variance did not provide the appropriate structure for the data. The effects of depletion type, diagnosis, depletion type-by-diagnosis and time on the ΔHAMD, ΔMADRS and ΔBAI scores were assessed with linear mixed models with an autoregressive covariance structure. Subject number and depletion sequence were included as random effects in all models. In addition, the factor gender was included in all models to regress-out this possible confounder since there were significantly more male subjects in the TD study. Furthermore, a *post hoc* analysis involving only the females alone was performed to prove that the statistical analysis adequately controlled for the gender difference. *Post hoc* *t*-tests involved a Tukey correction for multiple comparisons. Additional analyses assessed the different items measured with the HAMD, the MADRS and the BAI in detail using *t*-tests in rMDD to test for differences between TD and CD across symptom dimensions. The significance thresholds for these contrasts were set at alpha=0.05, two tailed. SAS 9.3 (SAS Institute, Cary, NC, USA) was used for all analyses. The means of the data are reported with their associated s.e. PET imaging ----------- The PET imaging methods have been described in detail in our previous reports on the same participant cohort.^[@bib27],\ [@bib28]^ Both studies used the same procedures with respect to PET imaging. The PET images were acquired when the peak behavioral response was expected, that is, in the TD study, PET was measured 6 h after the administration of the amino-acid mixture/placebo, because the peak effects were expected at 5--7 h; in the CD study, PET images were acquired 30 h after the administration of the first AMPT/placebo dose, which corresponded to the time period when peak behavioral response was expected.^[@bib3]^ Scanning was performed with a GE Advance scanner in three-dimensional mode (35 contiguous slices, 4.25 mm thick; three-dimensional resolution=6 mm full-width at half-maximum; GE Healthcare, Waukesha, WI, USA) and a slow bolus (over 2 min) injection of ^18^FDG. In order to obviate the need for arterial blood sampling, cerebral glucose utilization was quantified using a method that combines the left ventricular chamber time--tissue radioactivity data that were measured with dynamic PET imaging of the heart with venous blood sampling in order to provide ^18^FDG input function.^[@bib30]^ This method has been validated previously by comparing it to more invasive approaches that use arterial plasma sampling.^[@bib30]^ During image processing, the left ventricular time--radioactivity curve was extended in time to include the time of the brain emission scan by obtaining venous blood samples 25, 30, 35 and 50 min after the ^18^FDG injection. The mean radioactivity of these samples was divided by the mean left ventricular radioactivity concentration between 25 and 35 min post injection. This ratio was used to scale the 50-min venous sample concentration, which then was appended to the left ventricular curve in order to complete the input function that was used to generate parametric images of the regional cerebral metabolic rates for glucose (rCMRglu), as described by Moore *et al.*^[@bib30]^ To provide an anatomical framework for the analysis of the PET images, structural MRI scans were acquired with a 3.0-T scanner (Signa; GE Medical Systems, Waukesha, WI, USA) applying a three-dimensional magnetization-prepared rapid acquisition gradient-echo sequence (echo time 2.982 ms; repetition time 7.5 ms; inversion time 725 ms; voxel size 0.9 × 0.9 × 1.2 mm). PET imaging: region-of-interest analysis ---------------------------------------- To compare the effects of TD versus placebo with CD versus placebo on cerebral metabolism, first a region of interest (ROI)-based analysis (with *P*-values corrected for the number of ROIs), and then a voxelwise analysis (with *P*-values corrected for the number of independent comparisons across the entire brain) were performed. For the ROI analysis, MEDx (Medical Numerics, Sterling, VA, USA) software was used. ROIs were selected according to previous results in rMDD^[@bib5],\ [@bib31]^ and untreated, symptomatic patients with MDD,^[@bib32]^ which showed alterations in the OFC, posterior cingulate cortex (PCC), medial thalamus, and dorsolateral prefrontal cortex (DLPFC), ventral striatum, anterior PFC, pregenual PFC, subgenual PFC, ventrolateral PFC, anteromedial PFC, amygdala, hippocampus and anterior insula. The ROIs were defined *a priori* on an MRI template and were placed on each patient\'s registered MRI, using anatomical definitions described previously.^[@bib27]^ A binary mask of the gray matter was used to restrict all further analyses to gray matter voxels. To account for nonspecific global effects, the whole-brain metabolism was used to normalize the regional measures (the quantitative measures of whole-brain metabolism obtained under each depletion type revealed no significant difference in global metabolism for either CD or TD, as reported previously^[@bib27],\ [@bib28]^). The normalized mean metabolic activity was then obtained for each ROI in each subject and each session, and the regional differences (ΔrCMRglu) between sessions (TD versus placebo and CD versus placebo) were calculated for each subject. The statistical models that were applied to compare the ΔrCMRglu in each ROI included the main effects of depletion, diagnosis and their interaction. The significance level was Bonferroni corrected for the number of 13 ROIs. The significance threshold was set at alpha=0.05, two tailed. All *P*-values are reported before correction for multiple comparisons. PET imaging: voxelwise analysis ------------------------------- For the whole-brain analyses, we used Matlab (Matlab version 8, release 14; The MathWorks, Natick, MA, USA), SPM8 (Wellcome Trust Centre for Imaging, London, England; <http://www.fil.ion.ucl.ac.uk/spm/software/spm8>), and the toolbox aslm.^[@bib33]^ PET images were coregistered to the MRIs and spatially normalized to the Montreal Neurological Institute brain template with SPM8. Images were filtered with a 6-mm Gaussian smoothing kernel in order to compensate for interindividual anatomical variability. The statistical analysis of whole-brain metabolism involved a flexible factorial model in SPM8 with the factors depletion, diagnosis and subject. Two additional regressors were included to account for the oversampling of female subjects in the CD study and for the nonspecific fluctuations in the whole-brain metabolism. Clusters with a voxel-level threshold of *P*\<0.05, whole-brain corrected for false discovery rate are reported for regions without *a priori* hypotheses. PET imaging: correlational analysis ----------------------------------- Depression and anxiety items showing significant differences between TD and CD were analyzed in an exploratory *post hoc* analysis. Spearman rank correlations were calculated that assessed associations between those items and metabolism in the ROIs in rMDD subjects. The significance threshold was set at alpha=0.05, two tailed. Results ======= The clinical and demographic characteristics of the subject samples are detailed in [Table 1](#tbl1){ref-type="table"}. Behavioral effects of TD compared with CD ----------------------------------------- Both TD and CD induced more depressive symptoms in rMDD subjects compared with controls, but the depletion effect of TD compared with CD did not differ significantly as measured on either the HAMD (*P*=0.37) or the MADRS (*P*=0.27). In addition, there was no depletion type (TD versus CD)-by-diagnosis interaction on either scale\'s total score. Furthermore, TD and CD induced more anxiety symptoms as assessed using the BAI in subjects with rMDD compared with controls, but the effect did not significantly differ between depletion types (*P*=0.14). There was no depletion type-by-diagnosis interaction evident on the change in anxiety ratings. Repeating the analyses in female subjects alone did not alter the results. To assess the between-subject variation between the TD and CD study, we also calculated one-factor analyses of variance with the factor depletion type (two levels; TD and CD) and the dependent variables MADRS and BAI, respectively, including all measures during the placebo conditions. There were no significant differences in between-subject variations in MADRS scores (F(1,167)=2.53, *P*=0.11) and BAI scores (F(1,166)=0.42, *P*=0.52), respectively. Detailed analysis of HAMD and MADRS items in rMDD subjects ---------------------------------------------------------- Individual items from the HAMD and the MADRS showing significant differences in the TD effect compared with the CD effect in rMDD are displayed in [Figure 1a](#fig1){ref-type="fig"}. Compared with CD, TD induced stronger effects on depressed mood (*t* (1,254) =2.52, *P*=0.01), hopelessness (*t* (1,252)=3.15, *P*=0.002), apparent sadness (*t* (1,254)=2.56, *P*=0.01) and reported sadness (*t* (1,254)=3.07, *P*=0.002). The depletion effect of CD was stronger compared with TD on work and activities (*t* (1,254)=2.85, *P*=0.005), concentration difficulties (*t* (1,254)=2.92, *P*=0.004) and lassitude (*t* (1,252)=2.89, *P*=0.004). Repeating the analyses in female subjects only weakened the stronger effects of TD compared with CD on depressed mood (*t* (1,194)=1.77, *P*=0.08) and apparent sadness (*t* (1,194)=1.93, *P*=0.06), potentially due to the reduction in sample size. In addition, TD had stronger effects on hypochondriasis (*t* (1,194)=2.21, *P*=0.03) and CD had stronger effects on the inability to feel (*t* (1,194)=2.01, *P*\<0.05) in females. The results for the other items remained unchanged between the entire-group comparison versus the females-alone comparison. Detailed analysis of BAI items in rMDD subjects ----------------------------------------------- Anxiety items showing significant differences in the TD effect compared with the CD effect in rMDD are displayed in [Figure 1b](#fig1){ref-type="fig"}. Compared with the TD condition, CD induced significantly greater feelings of flushing (*t* (1,248)=2.44, *P*=0.02), palpitations (*t* (1,248)=2.07, *P*=0.04), fear (*t* (1,248)=3.38, *P*=0.0008), choking (*t* (1,248)=2.55, *P*=0.01), tremulousness (*t* (1,248)=3.03, *P*=0.003), dyspnea (*t* (1,247)=2.39, *P*=0.02) and diaphoresis (*t* (1,248)=2.05, *P*=0.04). In the analyses limited to female subjects, no significant difference was found on either the palpitations or the diaphoresis items, but the results on the other items were similar to those found in the entire sample. ROI analyis of PET data ----------------------- No difference was found between TD- and CD-induced effects on mean whole-brain glucose metabolism (*P*=0.26). [Figure 2](#fig2){ref-type="fig"} shows *a priori* defined ROIs that showed a significant difference in the TD compared with the CD effect on normalized regional metabolism. In the OFC there was a decrease in metabolism induced by CD compared with TD across groups (left: F (1,76)=20.7, *P*\<0.0001; right: F (1,76)=17.4, *P*\<0.0001). In addition, glucose metabolism in the left OFC in rMDD subjects was higher than in healthy controls (F (1,76)=7.30, *P*=0.009). In the right PCC, there was a significant depletion type-by-diagnosis interaction (F (1,74)=6.99, *P*=0.01) that was attributable to a higher TD effect than CD effect in rMDD and higher metabolism induced by TD in rMDD compared with controls. The right medial thalamus showed increased metabolism in rMDD compared with controls across studies (F (1,68)=8.58, *P*=0.005). The CD-induced increase in metabolism in the ventral striatum was higher than the increase under TD (left: F (1,76)=5.53, *P*=0.02; right: F (1,76)=10, *P*=0.002). There was a significant depletion type-by-diagnosis interaction on the regional glucose metabolism of the left anterior PFC (F (1,76)=4.82, *P*=0.03) that was attributable to a CD-induced metabolic increase compared with a TD-induced metabolic decrease in rMDD subjects. In the pregenual PFC, female subjects showed a significant higher metabolism compared with male subjects across studies and groups (left: F (1,76)=5.8, *P*=0.02; right: F (1,76)=6.61, *P*=0.01). The right subgenual PFC showed a higher metabolism across groups induced by TD (F (1,76)=3.98, *P*\<0.05) and a significant depletion type-by-diagnosis interaction (F (1,76)=4.78, *P*=0.03) that was attributable to a higher metabolism induced by TD compared with CD in controls; in addition, female subjects showed a higher metabolism in this region compared with male subjects across studies and groups (F (1,76)=4.42, *P*=0.04). The left ventrolateral PFC showed a TD-induced increase in metabolism across groups compared with a decrease in the CD study (F (1,76)=5.26, *P*=0.02). In the right anterior insula, there was a significant depletion type-by-diagnosis interaction (F (1,76)=5.26, *P*=0.02) that was attributable to a decreased metabolism induced by CD in rMDD subjects compared with controls and an increased metabolism induced by CD compared with TD in controls. After applying Bonferroni corrections, the effects in the right and left OFC and in the right ventral striatum remained significant. After repeating the analyses in female subjects alone, the depletion type-by-diagnosis interaction in the left anterior PFC was reduced to a nonsignificant trend (*P*=0.13); the depletion type-by-diagnosis interaction in the right anteromedial PFC (*P*=0.08) and the right subgenual PFC (*P*=0.06) and the effect of depletion type on the left ventrolateral PFC (*P*=0.06) remained as trend effects. In females only, metabolism was higher after CD compared with TD across groups in the right anterior PFC (F (1,57)=11.92, *P*=0.001) and in the right hippocampus (F (1,57)=5.05, *P*=0.03), and higher after TD compared with CD across groups in the PCC (left: F (1,57)=4.27, *P*=0.04; right: F (1,55)=6.31, *P*=0.02). All other results remained essentially unchanged from the comparisons involving the entire group with respect to statistical significance. Voxelwise analysis of PET data ------------------------------ A depletion type-by-diagnosis interaction was evident in the right lingual gyrus that survived applying correction for false discovery rate (peak: *t* (1,83)=4.65, size=34 voxel, *P*\<0.05, Brodmann area 18). *Post hoc* tests showed this interaction was attributable to a significant decrease in regional metabolism under CD in the rMDD subjects (*t* (1,54)=2.04, *P*\<0.05) but not in the controls (*P*=0.68), and no change under TD in either the rMDD subjects (*P*=0.25) or the controls (*P*=0.56). Correlational analysis of depression and anxiety items with PET data -------------------------------------------------------------------- [Table 2](#tbl2){ref-type="table"} shows Spearman rank correlations (rho) of depression and anxiety symptoms where significant differences between TD and CD had been found, and the corresponding regional glucose metabolism changes in the ROIs. As there were no significant correlations of TD-induced changes in depression and anxiety symptoms with corresponding changes in regional glucose metabolism, all values correspond to findings that were induced by CD. After applying Bonferroni corrections, only the correlation between the apparent sadness and the left anterior PFC remained significant. We also assessed how the relationship between changes in specific depression and anxiety items and changes in regional glucose metabolism were moderated by the depletion type and thus the differences in neurotransmitter levels. Results can be found in the [Supplementary Information](#sup1){ref-type="supplementary-material"} ([Supplementary Table S1](#sup1){ref-type="supplementary-material"}). In addition, we compared the significant correlation coefficients with the corresponding coefficients of the ROI in the complementary hemisphere; results can be found in the [Supplementary Information](#sup1){ref-type="supplementary-material"} ([Supplementary Table S2](#sup1){ref-type="supplementary-material"}). Discussion ========== We believe the current study is the first to compare the behavioral and neural effects of TD versus CD in unmedicated rMDD subjects and healthy controls. These challenges putatively reflected depletions in serotoninergic and catecholaminergic neurotransmission, respectively, which are of interest in patients with MDD because currently available antidepressant treatments enhance the function of one or both of these systems. [Table 3](#tbl3){ref-type="table"} displays the main findings of this study. Although brain activity was correlated with distinct depressive symptoms following CD, there was no direct relationship between specific symptoms and brain activity following TD. The behavioral and neural effects of CD and TD showed some shared effects that are compatible with the literature.^[@bib3],\ [@bib6],\ [@bib9],\ [@bib10],\ [@bib11],\ [@bib12],\ [@bib13],\ [@bib14],\ [@bib15],\ [@bib16],\ [@bib17],\ [@bib18]^ However, the different study design of the current study has to be considered as it was conducted in fully rMDD patients. Both depletion methods induced depressive symptoms as measured using the HAMD and MADRS, and both increased cerebral glucose metabolism in the medial thalamus, the OFC and the ventral striatum. These findings suggest that interactions between the monoamine systems are involved in the pathogenesis of depression. The medial thalamus has an important relay function in connecting sensory and basal ganglia inputs with prefrontal cortical structures. The abundant serotoninergic and dopaminergic innervation of the thalamus has been shown to participate in complex synergistic or opposing interactions, potentially contributing to the similar impact of TD and CD on thalamic glucose metabolism, and conceivably on several depressive symptoms. For example, the thalamus receives serotonergic afferents from the dorsal and median raphe nucleus,^[@bib34],\ [@bib35]^ which participate in the neural processing underlying anxiety-related behaviors^[@bib36],\ [@bib37]^ and the generation of various stages of the sleep--wake cycle.^[@bib38]^ In addition, the dopaminergic system modulates neural transmission within the limbic--thalamo--cortical circuits that involve regions of the medial and orbital prefrontal cortex, ventral striatum and amygdala, which modulate reward-related learning and motivation.^[@bib39]^ Dysfunction of this pathway may underlie a range of depressive symptoms including lack of motivation, including problems related to work and activities. For example, the OFC contains dopaminergic terminals and receptors^[@bib40]^ and blockade of these receptors reduces the break point of rats responding on a progressive ratio schedule of reinforcement, a classic test of incentive motivation.^[@bib41]^ In addition, dopamine depletion in the OFC impaired responding for delayed reward.^[@bib42]^ Moreover, the OFC also receives serotonergic innervation from the dorsal and median raphe nuclei, and the reciprocal innervation from the OFC enables the OFC to regulate not only its own serotonin input but the serotonin input to the rest of the forebrain, which has been associated with the capability of animals to adapt to changing reward contingencies.^[@bib43]^ Notably, depletion of serotonin has been shown to impair this flexibility of the reward system.^[@bib43]^ In addition, it has been suggested that serotonin and dopamine modulate different functions in the OFC with orbitofrontal serotonin preventing competing, task-irrelevant stimuli from biasing task-based responding, processes that may hold relevance for the attentional biases toward negative stimuli extant in MDD.^[@bib44],\ [@bib45]^ Our results further appear in line with the clinical observation that motivational deficits and the inability to concentrate are closely connected.^[@bib46]^ TD more specifically induced symptoms of sadness and depressed mood. Serotonin has a well-known function in the emotional inhibition and regulation, and acute TD has been shown to induce a negative attentional and mnemonic bias both in rMDD subjects and in healthy controls.^[@bib47],\ [@bib48],\ [@bib49]^ In the right PCC, TD compared with CD induced a metabolic increase compared with a decrease in rMDD and a significant difference between rMDD and controls in TD only. The PCC has a specific role in the regulation of pain, which has been related theoretically to negative effect. Consequently, serotonin deficiency may both increase negative emotion and reduce emotional control. In addition, given that the PCC is a critical node in serotonin neurotransmission^[@bib50]^ and is also implicated in self-referential processes as a hub of the default mode network, this study adds to the evidence that serotonin deficiency has an important role in default mode network overactivity in depression.^[@bib51]^ Nevertheless, the correlation between TD-induced changes in regional metabolism and TD-induced depressive symptoms was not significant, suggesting that the relationship between serotonin, brain metabolism and sad/depressed mood is complex. CD specifically induced symptoms of reduced activity, impaired concentration and lassitude and specifically increased brain metabolism in specific PFC regions and the ventral striatum. Dopamine depletion in monkeys leads to cognitive and attention deficits in the primate PFC,^[@bib52],\ [@bib53]^ potentially resembling the concentration problems encountered in some depressed humans. This is supported by our study that found that the CD-induced increase in brain metabolism in the anterior PFC was correlated with concentration difficulties induced by CD. The findings of increased lassitude and of work and activity problems conceivably support hypotheses^[@bib54],\ [@bib55]^ that a dopamine deficit underlies dysfunctional reward processing, which leads to amotivation-related symptoms in MDD. As shown in [Table 2](#tbl2){ref-type="table"}, the CD-induced increase in brain metabolism in the striatum was directly associated with lassitude, and the CD-induced increase in brain metabolism in the ACC was associated with self-reported sadness. This argues for a relatively direct mechanistic relationship between lack of dopamine and/or norepinephrine in the pathogenesis of distinct depressive symptoms. In the left anterior PFC, the glucose metabolism increased under CD but decreased under TD. The anterior PFC has been linked to the processing of affective salience of sensory stimuli in previous studies,^[@bib56],\ [@bib57]^ which may relate to reduced activity and lassitude depressive symptoms. However, this effect was weakened to a trend level when repeating the analysis in female subjects only. One of the most consistent functional imaging findings in MDD is increased metabolism in the subgenual and ventrolateral PFC.^[@bib58],\ [@bib59],\ [@bib60]^ We found metabolic increases in these brain regions to be more pronounced under TD compared with CD, suggesting that serotonin deficiency is more important in the pathogenesis of the hyperactivity of these brain regions that are both part of the extended visceromotor networks, which participates in regulating the neuroendocrine, autonomic and experiential aspects of emotion.^[@bib24]^ We found some interhemispheric differences in our correlational analysis of CD-induced changes in mood and neurotransmission, with the most prominent being the positive correlation between changes in depressed mood and changes in regional glucose metabolism in the left DLPFC and a corresponding negative correlation in the right DLPFC. Interhemispheric differences in the DLPFC have been reported in several studies measuring resting state^[@bib25],\ [@bib61]^ and have been the neurobiological basis for therapeutic brain stimulation paradigms with transcranial magnetic stimulation^[@bib62]^ and transcranial direct current stimulation.^[@bib63]^ Specifically, hypoactivity in the left DLPFC has been linked to negative emotional judgment and hyperactivity in the right DLPFC to attentional modulation.^[@bib64]^ Our findings suggest that interhemispheric differences in the DLPFC in depression are related to a deficit in catecholaminergic neurotransmission. Unexpectedly, the voxelwise analysis identified reduced glucose metabolism under CD in the right lingual gyrus in the unmedicated rMDD sample. Reductions in this region have been previously reported in a study in young MDD adults^[@bib65]^ and abnormal focal magnetic low-frequency activity has been found in untreated patients with MDD.^[@bib66]^ The current study adds that abnormality in this brain region is associated with reduced catecholaminergic neurotransmission. Interestingly, no difference was found in the current study between TD and CD on global levels of anxiety, as measured with the BAI. We found stronger effects of CD compared with TD on several anxiety items including greater feelings of flushing, palpitations, fear, choking, tremulousness, dyspnea and diaphoresis. A possible explanation for this observation is that somatic anxiety symptoms in response to threat and stress are modulated more potently by central catecholaminergic pathways than via central serotonergic pathways.^[@bib67]^ The current study had several strengths that are noteworthy. First, it compared two experiments that took place at the same scanning site and used the same PET imaging procedure for both the TD and the CD studies. Second, the fact that a sample of subjects with rMDD off medication was assessed allowed us to investigate behavioral and neural effects of serotonin- and catecholamine-related pathways unbiased by medication. Further, we could interpret the findings as risk factors for a depressive relapse. Finally, in contrast to our findings, previous studies involving SSRI and norepinephrine reuptake inhibitor pharmacotherapy of MDD patients found surprisingly small differences between serotoninergic and catecholaminergic agents on depressive symptoms.^[@bib68]^ Possible reasons for this discrepancy could be the insufficient specificity of chronic SSRI and norepinephrine reuptake inhibitor administration, as the 6--8-week duration of therapeutic trials allows for adaptive changes to occur in other neurotransmitter systems and for a placebo effects to increase, which both may blur differences across challenges.^[@bib69]^ The shorter time frame needed for comparing the effects of acute TD versus acute CD thus may offer greater sensitivity than clinical trials using monoaminergic antidepressant drugs, because the acute nature of both depletion methods ensures a higher specificity for serotonin and catecholamine systems, respectively, along with relatively smaller placebo effects. In future studies, refining our approach using a placebo-controlled, double-blind, crossover design would enable within-subject comparisons with increased power to detect differences between depletion paradigms. The current study had several limitations that merit comment. The study sample was relatively small and contained a majority of female subjects. We took this into account by including a regressor for gender in each analysis and by repeating all analyses with female subjects only. In addition, PET imaging of glucose metabolism did not specifically assess central serotonin or catecholamine concentrations. Instead, we assessed the central effect of TD indirectly by measuring plasma total and free tryptophan levels and the central effect of CD by assessing serum prolactin levels, which is the standard method to assess the effect of central CD.^[@bib28],\ [@bib70]^ A preferable design would have been a within-subjects design, whereby all participants underwent both depletion procedures. However, the two experiments took place at the same scanning site and used the same PET imaging procedure for both the TD and the CD studies, increasing the comparability of the two studies. In addition, an analysis of between-subject variance between TD and CD during the placebo condition did not reveal any significant differences. Finally, a noninvasive method for comparing the depth of CD versus TD within the brain is not available. Given the similar amount of depressive symptoms induced by both methods and the similar effect of both methods on metabolism in at least some brain structures, we assumed that TD and CD, as used in our study, were comparable regarding the hypotheses we aimed to test. Taken together, these data suggest that serotonin and catecholamines have both common and distinct roles in the neurobiology of depressive symptoms. This study further suggests that the development of psychopathological and neuronal markers predicting response to selective monoamine inhibition may be feasible. Finally, this study provides a rationale for the use of antidepressants with primary pharmacological actions involving both serotonergic and catecholaminergic mechanisms in some patients.^[@bib71]^ This research was supported by the Intramural Research Program of the National Institutes of Mental Health and the University Hospital and Hermann Klaus Foundation, Zurich. We thank Eveline Nüesch for statistical advice. [Supplementary Information](#sup1){ref-type="supplementary-material"} accompanies the paper on the Translational Psychiatry website (http://www.nature.com/tp) WCD is currently an employee of Janssen Pharmaceuticals of Johnson & Johnson. The remaining authors declare no conflict of interest. Supplementary Material {#sup1} ====================== ###### Click here for additional data file. {#fig1} {#fig2} ###### Demographic and clinical characteristics of unmedicated subjects with remitted major depressive disorder (rMDD) and healthy controls *Characteristic* *TD (*n=*55)* *CD (*n=*30)* ------------------------- --------------- --------------- ------------ -------------   *HC* *rMDD* *HC* *rMDD* Sex no., f/m 18/9 19/9 12/1 16/1 Age, mean (s.d.), years 34.2 (11.2) 39.8 (12.7) 39.1 (9.6) 39.2 (10.8) MADRS at study entry 0.6 (1.2) 1.4 (1.8) 0.4 (0.9) 1.6 (1.9) HAMD at study entry 0.9 (1.1) 1.3 (1.4) 0.4 (0.8) 1.6 (1.1) BAI at study entry 0.8 (1.2) 2.3 (2.6) 0.5 (1.1) 1.9 (1.7) Abbreviations: BAI, Beck Anxiety Inventory; CD, catecholamine depletion; f/m, female/male; HAMD, Hamilton Depression Scale; HC, healthy control; MADRS, Montgomery--Åsberg Depression Rating Scale; NA, not applicable; no., number; TD, tryptophan depletion. ###### Spearman rank correlations (rho) of changes in depression and anxiety symptoms with changes in regional glucose metabolism   *Apparent sadness* *Reported sadness* *Concentration difficulties* *Lassitude* *Depressed mood* *Work and activities* *Feeling hot* *Heart pounding/racing* *Feeling of choking* *Hands trembling* ------------------------- -------------------- -------------------- ------------------------------ --------------- ------------------ ----------------------- --------------- ------------------------- ---------------------- ------------------- DLPFC, left 0.17 0.07 0.03 **−**0.50 **0.57\*** **−0.55\*** 0.02 **−**0.05 0.18 **−**0.37 DLPFC, right **−0.66\*\*** **−0.57\*** **−**0.34 **−**0.09 **−0.63\*** **−**0.44 0.22 **−**0.11 **−**0.14 **−**0.30 Anterior PFC, left **0.79\*\*\*** **0.73\*\*** **0.71\*\*** 0.24 0.45 0.36 **−**0.46 **−**0.28 **−**0.50 **−**0.02 Anterior PFC, right **−**0.06 **−**0.14 0.17 **−**0.02 0.02 **−**0.23 **0.63\*** **−**0.05 **−**0.23 **−**0.14 Hippocampus, left **−**0.20 **−**0.19 **−**0.03 0.05 **−**0.07 **−**0.15 **0.63\*** **−**0.22 **−**0.27 **−**0.24 Hippocampus, right 0.50 0.46 **0.70\*\*** 0.09 **0.63\*** 0.09 **−**0.09 **−**0.27 **−**0.18 **−**0.11 Ventral striatum, left **0.53\*** **0.55\*** 0.19 **−0.65\*\*** **0.58\*** **−**0.19 0.03 0.09 0.05 **−**0.27 Ventral striatum, right **−**0.31 **−**0.16 **−**0.41 **−0.64\*** **−**0.12 **−**0.09 0.25 **−**0.13 0.00 **−**0.32 Pregenual PFC, right **−**0.34 **−**0.31 **−**0.12 **−**0.12 **−**0.31 0.00 **−**0.17 **−0.53\*** **−**0.41 **−**0.35 PCC, left **−**0.37 **−**0.32 **−**0.42 **−**0.16 **−**0.32 0.08 0.29 0.49 0.14 **0.55\*** PCC, right 0.39 0.42 0.30 0.33 0.05 **0.59\*** **−**0.38 **−**0.13 **−**0.18 0.10 Anteromedial PFC, left **−**0.02 **−**0.09 0.09 **−**0.01 0.07 **−**0.13 **−**0.27 0.49 **0.54\*** 0.39 Anterior insula, left 0.28 0.23 **−**0.01 **−**0.22 **0.59\*** 0.03 **−**0.01 0.20 0.23 0.10 Abbreviations: CD, catecholamine depletion; DLPFC, dorsolateral prefrontal cortex; PCC, posterior cingulate cortex; PFC, prefrontal cortex; TD, tryptophan depletion. Items were chosen according to our previous analyses, that is, where we had found significant differences between TD and CD. Note that there were no significant correlations of TD-induced changes in depression and anxiety symptoms with corresponding changes in regional glucose metabolism, so all values correspond to findings that were induced by CD. Statistically significant correlations are indicated in bold. Significance at \**P*\<0.05; significance at \*\**P*\<0.01; significance at \*\*\**P*\<0.001. ###### Summary of the main findings of the study, categorized by common and differential effects. Effects on behavior relate to subjects with remitted depression only   *Common effects* *Differential effects* ----------------------------- --------------------------------------------------------------- ------------------------------------------------------------- ----------------------------------------------------------     *TD\>CD* *CD\>TD* Behavior Global HAMD Global MADRS Global BAI Depressed mood Sadness Work and activities Concentration difficulties Lassitude Cerebral glucose metabolism Whole brain Ventral striatum Medial thalamus, right OFC, left Pregenual PFC, right Ventrolateral PFC, left PCC, right OFC Anterior PFC, left Subgenual PFC, right Abbreviations: BAI, Beck Anxiety Inventory; CD, catecholamine depletion; HAMD, Hamilton Depression Scale; MADRS, Montgomery--Åsberg Depression Rating Scale; OFC, orbitofrontal cortex; PCC, posterior cingulate cortex; PFC, prefrontal cortex; TD, tryptophan depletion.

**What do we already know about this topic?** Although there is considerable interest in possibilities to enhance informal care, the risk of a negative health impact among informal caregivers should be taken into account especially when the intensity of the care increases. **How does your research contribute to the field?** This study provides a broad overview of characteristics and needs of informal caregivers that supports researchers in selecting risk factors for perceiving a high burden to eventually prevent overburdening. **What are your research's implications toward theory, practice, or policy?** To relieve the perceived burden of informal caregivers, a focus should be set on sharing care tasks, where a better cooperation with and support from the concerning partners in the municipalities is a necessity. Introduction {#section1-0046958018775570} ============ The problem of overburdening among informal caregivers has been increasing in developed countries, such as the Netherlands, where the number of caregivers reporting to perceive a high burden has been increased by 50% in the period from 2001 to 2008.^[@bibr1-0046958018775570]^ In the Netherlands, recent changes introduced by the Dutch government, such as the objective of developing a "participatory society"^[@bibr2-0046958018775570]^ and allocating more of the coordinating tasks to municipal authorities under the Social Support Act,^[@bibr3-0046958018775570]^ have led to a greater focus on community development. Governments of other developed countries are also encouraging citizens' initiatives to support each other.^[@bibr4-0046958018775570]^ One of these initiatives that is becoming more popular these days concerns a greater focus on the possibilities of informal caregivers to take over formal care provision tasks.^[@bibr5-0046958018775570]^ Informal caregivers are unpaid nonprofessionals who support chronically ill, disabled, and other people in need in their immediate environment, such as family members, friends, or neighbors, over a lengthy period of time.^[@bibr6-0046958018775570]^ Caregivers often provide the support, because of their personal connection to the person in need.^[@bibr7-0046958018775570]^ The support they provide can be physical, social, and/or emotional support, which is based on the care recipients' needs and the abilities, capacity and willingness of the caregiver to provide the different types of support.^[@bibr8-0046958018775570],[@bibr9-0046958018775570]^ In less intensive stages of caregiving, informal care is often perceived to be a source of positive influence on the lives of caregivers.^[@bibr10-0046958018775570]^ The Dutch Expertise Centre for Informal Care^[@bibr11-0046958018775570]^ found that informal caregivers often experience a sense of meaningfulness in life and feel more positive about themselves. Caregivers who receive some form of appreciation for their activities from the care recipient also mentioned positive influences of their care provision activities.^[@bibr12-0046958018775570]^ However, as the intensity of the care increases, more caregivers report to perceive high burdens. When the caregiver burden, defined as a multidimensional response, eg, physical or mental, to the negative appraisal and perceived stress resulting from taking care of an ill individual in their immediate environment, becomes too intense, informal caregivers will be unable to provide the care.^[@bibr13-0046958018775570]^ The perception of a high burden, when a caregiver is subjected to an excessive level of burden, has a negative influence on their physical, psychological, emotional, and functional state of health.^[@bibr10-0046958018775570]^ A significantly lower quality of life was found among informal caregivers,^[@bibr14-0046958018775570]^ which is also reflected by other indicators (eg, mental health problems such as stress, tension, anxiety and depression), and physical health problems (such as back injury, sleep disruption and hypertension).^[@bibr7-0046958018775570],[@bibr10-0046958018775570],[@bibr15-0046958018775570]^ According to the Dutch National Institute for Public Health and the Environment, reporting a high burden is a predictor of an impending decline of the health status of caregivers.^[@bibr16-0046958018775570]^ Based on Pearlin's stress process model and various studies researching the topic of informal care, different characteristics can be identified that can explain a caregiver's burden. First, background factors were seen as important, eg, the context of the caregiving situation and personal characteristics. A higher perceived burden was found to be associated with contextual factors such as collectivistic cultures that value the group as a whole,^[@bibr7-0046958018775570],[@bibr17-0046958018775570]^ the nonexistence of a social care system,^[@bibr18-0046958018775570]^ coresidence with the care recipient,^[@bibr19-0046958018775570]^ and a small social network with low levels of social support.^[@bibr20-0046958018775570]^ At a personal level, a high perceived burden was associated with a higher age,^[@bibr19-0046958018775570]^ female gender,^[@bibr20-0046958018775570][@bibr21-0046958018775570]-[@bibr22-0046958018775570]^ a lower socioeconomic status,^[@bibr7-0046958018775570]^ a lower educational level,^[@bibr23-0046958018775570]^ and poorer physical or mental health status of the caregiver.^[@bibr20-0046958018775570]^ In addition, stressors, such as relationship factors,^[@bibr7-0046958018775570],[@bibr10-0046958018775570]^ role strains,^[@bibr24-0046958018775570],[@bibr25-0046958018775570]^ and psychological strains,^[@bibr20-0046958018775570]^ were found to contribute to the burden perceived by informal caregivers. Mediating factors such as well-developed coping resources might reduce the influence of stressors.^[@bibr19-0046958018775570],[@bibr26-0046958018775570]^ The specific combination of a caregiver's background, stressors, and mediators determines the physical and mental health outcomes and the burden they perceive as shown in Pearlin's stress process model.^[@bibr20-0046958018775570]^ Although studies have focused on identifying important characteristics, a comprehensive overview aimed at exploring the involvement of all possible characteristics contributing to the risk of perceiving a high burden is currently lacking. In view of the increased recognition of the importance of supporting caregivers, it is necessary to be aware of potential risk factors. This study supports researchers in selecting these risk factors for perceiving a high burden. The considerable number of interventions that are available to support informal caregivers, such as training and education programs, approaches to care planning, support groups, individual counseling and mindfulness,^[@bibr7-0046958018775570],[@bibr27-0046958018775570][@bibr28-0046958018775570][@bibr29-0046958018775570]-[@bibr30-0046958018775570]^ specifically enhancing coping styles, can be more targeted at the risk factors among caregivers, which should eventually prevent overburdening. The aim of this study was to explore the characteristics and needs of adult and senior informal caregivers reporting a low or high burden, as well as the contributions made by these characteristics to perceiving a high burden. These insights are essential for the development of future evidence-based interventions to reduce the burden perceived by informal caregivers. Methods {#section2-0046958018775570} ======= Study Design, Participants, Recruitment, and Data Collection Procedure {#section3-0046958018775570} ---------------------------------------------------------------------- This quantitative study had a cross-sectional design. In this study, secondary data analyses were performed on the gathered data from the Limburg Health Monitor 2012. The monitor is a nationwide survey that is repeated every 4 years, where information is collected from a representative sample of adult and senior Dutch citizens. It provides the Dutch government with information about the overall health status of Dutch citizens, including physical, mental, and social health domains, and based on this information, health policy targets are set.^[@bibr31-0046958018775570]^ The Health Monitor is performed among two samples: adults (17-64 years) and seniors (65+). From all adults and seniors who participated, only informal caregivers were selected and included in this study. To answer the current research aims, data from both samples, adults and seniors, were used. Selected citizens received the Health Monitor questionnaire and an information letter at home by the end of September 2012.^[@bibr32-0046958018775570]^ Participants who were providing informal care were included in the present study. Data, Instruments, and Measures {#section4-0046958018775570} ------------------------------- The question, "Are you currently providing informal care?" (0 = yes; 1 = no) was used to include all informal caregivers. The main outcome was the burden perceived by the respondent, as assessed by the question, "Do you currently feel burdened?" rated on a 5-point scale from perceiving no burden to being overburdened, which was recoded into low burden (0 = no, little, or some burden perceived) and high burden (1 = relatively severe or very severe perceived burden or being overburdened). Other assessed variables are described below, most of them based on validated indicator scales. The background variables were age (1 = 17-24; 2 = 25-39; 3 = 40-54; 4 = 55-64; 5 = 65-74; 6 = 75-79; 7 = 80+), gender (1 = male; 2 = female), education (1 = low, ie, "primary, basic vocational, lower general school, or no education"; 2 = intermediate I, ie, "higher secondary education, preparatory academic education, or medium vocational school"; 3 = intermediate II, ie, "higher vocational school"; 4 = high, ie, "university level"), annual income (1 = maximum of 15 200 euros; 2 = 15 201-19 400 euros; 3 = 19 401-24 200 euros; 4 = 24 201-31 100 euros; 5 = minimum of 31 101 euros), marital status (1 = married/registered partnership; 2 = unmarried; 3 = divorced; 4 = widow/widower), and social contacts with family, friends, or neighbors (1 = at least once a week; 2 = less than once a week), social network based on Wenger's classification^[@bibr33-0046958018775570]^ (1 = locally integrated; 2 = family dependent; 3 = local self-contained; 4 = wider community focused; 5 = private restricted). Three relationship factors for informal care provision, explaining the connection between informal caregiver and care recipient, were included. The first was the type of support provided, "Which activities do you carry out?" with answering options as follows: support for housekeeping, preparing meals, support for personal care, support for medical care, company, consolation and distraction, accompaniment and transport, administrative support, and other (0 = not providing the activity; 1 = providing the activity). The second factor was the recipient of the care provided, "To whom are you providing informal care?" with answering options as follows: partner, child, parent (or in-law), other family members, and neighbors/friends (0 = not providing care to the recipient; 1 = providing care to the recipient). The third was the duration of the care, "For how long have you been providing informal care?" with options shorter or longer than 3 months (1 = less than 3 months; 2 = more than 3 months), and "How many hours a week do you spend on providing care?" where caregivers had to fill in the mean number of hours, which was recoded into 3 categories (1 = 1-5 hours a week; 2 = 6-15 hours a week; 3 = \>16 hours a week). The physical and mental health status of the informal caregivers were assessed using 4 indicator scales. Quality of life was assessed by the question, "How good is your health?" (1 = good or very good perceived health; 2 = poor or very poor to moderate perceived health). Chronic conditions were assessed by presenting respondents with a list of conditions from which they could select those they suffered from (0 = no chronic conditions; 1 = one; 2 = two; 3 = three; 4 = four or more). Fear and depression were assessed with the Kessler Psychological Distress Scale, which includes 10 questions to be answered on a 5-point scale from always to never, indicating depressive symptoms, eg, feeling sad, restless, worthless, or tired, recoded into 3 categories (1 = no low symptoms; 2 = moderate symptoms; 3 = high symptoms).^[@bibr34-0046958018775570]^ Loneliness was assessed with 11 questions to be answered on a 3-point scale (yes--more or less--no) for perceiving loneliness, eg, lacking a good friend, having a small social network, feeling of emptiness, and feeling abandoned, recoded into 3 categories (1 = no low; 2 = moderate; 3 = severe).^[@bibr35-0046958018775570]^ Assessment of caregivers' social roles focused on work situation, ie, having a paid job, by asking, "Which situation applies to you?" with answering options of having a paid job for \<12 hours, 12 to 20 hours, 20 to 32 hours, \>32 hours a week, being retired, unemployed, incapacitated, on social benefit, being homemaker, and being a student, (0 = unemployed, people without a paid job; 1 = employed, people with a paid job). Financial difficulties were also assessed: "Did you experienced financial difficulties in the last 12 months?" with the answers of no difficulties at all, no difficulties but I need to watch my expenditures, yes some difficulties, and yes many difficulties (0 = no difficulties; 1 = some and many difficulties). Family life was assessed by checking whether the caregivers' households included children living at home, using the question, "With whom are you living together?" with answering options such as partner, children younger than 18 years, children aged 18 years and above, parent(s), other adult(s), and living alone (0 = no children at home; 1 = children at home). The assessed coping indicators concerned mastery and self-management. Mastery was assessed using 7 questions on a 5-point scale from totally agree to totally disagree indicating the feeling of self-control, eg, ability to solve problems, control over things that happen in life, ability to control the future, recoded into low and high sense of mastery (1 = low; 2 = high).^[@bibr36-0046958018775570]^ Self-management was assessed using 6 questions on a 6-point scale from never to very often, regarding self-management activities of the informal caregivers, such as taking the initiative to enter into or maintain contact with other people and actively engaging in leisure activities, recoded into 3 categories (1 = low; 2 = moderate; 3 = high).^[@bibr37-0046958018775570]^ Finally, the needs of the caregivers were assessed by asking about 5 practical and emotional needs, eg, for information and advice regarding the execution of their caregiver role, replacement to take over care provision tasks, emotional support, relaxing activities, and advocacy regarding the representation and fulfillment of the interests of caregivers. The question was, "Apart from the support that you might already have, do you need another kind of support to help you with your caregiving tasks?" (0 = no need; 1 = need). Data Processing and Analysis {#section5-0046958018775570} ---------------------------- Bivariate analyses were performed, using Pearson's chi-square, to explore the associations between perceiving a high burden and each of the variables described above. Subsequently, a logistic regression analysis was performed. The specific sampling, used to optimize the representativeness of the sample, needed to be taken into account during the analysis, which was done by carrying out the logistic regression analysis using a complex sampling procedure. Perceived burden among caregivers was included as the dependent variable and the other variables were selected as independent variables. Odds ratios (ORs) are reported as effect sizes and classified according to the recommendations by Rosenthal,^[@bibr38-0046958018775570]^ where OR \< 1.5 indicates a weak association, 1.5 to 2.5 a moderate association, 2.5 to 4.0 a strong association, and \>4.0 a very strong association. The analysis contained 6 different categories of variables, the background variables, relationship indicators, roles of caregivers, physical and mental health status, and coping variables. All the analyses were performed with SPSS IBM statistics version 21. Results {#section6-0046958018775570} ======= Quantitative Analysis {#section7-0046958018775570} --------------------- ### Description of the sample {#section8-0046958018775570} From all participants, adults and seniors, who filled in the questionnaire of the Limburg Health Monitor, 3067 (12.8%) adult and 1936 (13.6%) senior participants provided informal care and could be included in this study ([Table 1](#table1-0046958018775570){ref-type="table"}). The majority of the adults were women over the age of 40. Most participants were married or were living together with a partner, had an intermediate or high educational level, a paid job, and an annual income over 24 201 euros. Most of the senior participants were aged between 65 and 74, and a small majority of the participants were women. Most were married or were living together with a partner, had an intermediate educational level, an annual income below 24 200 euros, and were not in work. About two-thirds of the senior caregivers were retired. Almost 15% of all caregivers perceived a high burden due to their caregiving tasks. ###### Description of Informal Caregivers Participating in the Study.  Adult caregivers (n = 3067, 12.8%) Senior caregivers (n = 1936, 13.6%) ------------------------------- ------------------------------------ ------------------------------------- Age, % - 17-24 177 (5.8) --- - 25-39 322 (10.5) --- - 40-54 1431 (46.7) --- - 55-64 1137 (37.1) --- - 65-74 --- 1116 (57.6) - 75-79 --- 482 (24.9) - 80+ --- 338 (17.5) Gender, n (%) - Male 935 (30.5) 941 (48.6) - Female 2132 (69.5) 995 (51.4) Education, n (%) - Low 87 (2.8) 263 (13.6) - Intermediate I 882 (28.8) 892 (46.1) - Intermediate II 1178 (38.4) 397 (20.5) - High 920 (30.0) 384 (19.8) Household income, n (%) - Maximum of €15 200 314 (10.2) 160 (8.3) - €15 201-€19 400 452 (14.8) 503 (26.0) - €19 401-€24 200 621 (20.2) 528 (27.3) - €24 201-€31 100 846 (27.6) 424 (21.9) - Minimum of €31 101 834 (27.2) 321 (16.6) Employment status, n (%) - Not currently employed 902 (29.4) 1839 (95.0) - Employed 2165 (70.6) 97 (5.0) Marital status, n (%) - Married/living with partner 2441 (77.5) 1573 (81.3) - Never been married 370 (11.8) 61 (3.2) - Widowed 71 (2.3) 65 (3.4) - Divorced 266 (8.4) 237 (12.2) Burden perceived, n (%) - Low 2632 (85.8) 1650 (85.2) - High 435 (14.2) 286 (14.8) ### Associations between the burden perceived and the other characteristics {#section9-0046958018775570} As regards the background variables, informal caregivers who had a lower educational level, a lower household income, a smaller social network centered on family, and privacy restricted, and caregivers who were widowed or divorced, perceived a significantly higher burden. Female gender was associated with perceiving a high burden among the adult caregivers ([Table 2](#table2-0046958018775570){ref-type="table"}). ###### Associations Between the Burden Perceived by Informal Caregivers and Their Background Variables.  Background variables Adults (n = 3067) Chi-square Senior (n = 1936) Chi-square ------------------------------------------------- ------------------- -------------------------------------------------------------- ------------------- -------------------------------------------------------------- Age, %  17-24 5.4 853.30[\*](#table-fn1-0046958018775570){ref-type="table-fn"}  25-39 12.8  40-54 18.1  55-64 11.5  65-74 12.3 64.76[\*](#table-fn1-0046958018775570){ref-type="table-fn"}  75-79 17.1  80+ 14.5 Gender, %  Male 11.8 273.62[\*](#table-fn1-0046958018775570){ref-type="table-fn"} 13.8 0.95  Female 16.1 13.3 Education, %  Low 24.6 236.30[\*](#table-fn1-0046958018775570){ref-type="table-fn"} 21.6 196.90[\*](#table-fn1-0046958018775570){ref-type="table-fn"}  Intermediate I 14.0 13.7  Intermediate II 13.6 11.5  High 15.4 10.6 Household income, %  Maximum of €15 200 21.4 558.46[\*](#table-fn1-0046958018775570){ref-type="table-fn"} 16.2 99.57[\*](#table-fn1-0046958018775570){ref-type="table-fn"}  €15 201-€19 400 15.1 16.6  €19 401-€24 200 13.8 12.8  €24 201-€31 100 15.0 11.9  Minimum of €31 101 11.3 10.2 Marital status, %  Married/living with partner 13.5 567.54[\*](#table-fn1-0046958018775570){ref-type="table-fn"} 14.7 107.59[\*](#table-fn1-0046958018775570){ref-type="table-fn"}  Never been married 13.7 10.7  Widowed 22.3 7.5  Divorced 23.5 11.8 Social contacts, %  Contacts with family: at least once a week 14.3 22.15[\*](#table-fn1-0046958018775570){ref-type="table-fn"}  Contacts with family: less than once a week 16.3  Contacts with friends: at least once a week 13.3 277.55[\*](#table-fn1-0046958018775570){ref-type="table-fn"}  Contacts with friends: less than once a week 17.9  Contacts with neighbors: at least once a week 14.2 13.67[\*](#table-fn1-0046958018775570){ref-type="table-fn"}  Contacts with neighbors: less than once a week 15.1 Network type, %  Locally integrated 11.0 156.33[\*](#table-fn1-0046958018775570){ref-type="table-fn"}  Family-dependent 18.2  Local self-contained 12.9  Focused on wider community 10.2  Private restricted 16.3 *P* \< .05. As regards the relationship factors, there was a positive association between perceiving a high burden and all caregiving activities, especially when more activities were provided. Providing care to a partner or children was associated with significantly higher burden, while care for parents (or in-laws), other family members, neighbors, and friends was associated with a low burden. The duration of the care provision also turned out to be a significant factor, as a larger number of hours per week and a longer duration were associated with a higher burden. Role variables of the informal caregivers showed positive associations with a high burden being perceived by adult caregivers with children living at home, but this association was not found among the senior caregivers. Associations with social roles revealed a lower burden among employed caregivers. Finally, a high burden was found among adult and senior caregivers who had financial difficulties. As regards the physical health of adult and senior informal caregivers, a lower perceived health status, a higher number of chronic conditions, and the presence of a long-lasting disease in the last 12 months were associated with a high burden. Among all caregivers, experiencing depressive symptoms and loneliness was associated with perceiving a high burden. Caregivers who expressed a need for support, both practical and emotional, perceived a high burden. The strongest association was found when caregivers expressed a need for replacement and advocacy. Finally, a low sense of mastery (all caregivers) and low self-management (senior caregivers) were associated with a high perceived burden. An overview of the bivariate analysis of the characteristics and needs is provided in [Table 3](#table3-0046958018775570){ref-type="table"}. ###### Associations Between the Burden Perceived by Informal Caregivers and Their Characteristics and Needs.  Adults (n = 3067) Chi-square Senior (n = 1936) Chi-square ------------------------------------------------------------- ------------------- ----------------------------------------------------------------- ------------------- --------------------------------------------------------------- Relationship variables between caregiver and care recipient  Activities   Number of activities, %    1 10.8 9173.69[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 7.5 2241.10[\*](#table-fn2-0046958018775570){ref-type="table-fn"}    2 6.6 3.9    3 9.3 6.8    4 16.5 20.8    5 32.6 23.0    6 50.6 31.1    7 48.1 51.2  Practical support, %   Yes 14.7 9.66[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 15.1 134.28[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   No 13.6 7.8  Personal and medical care support, %   Yes 29.9 3894.87[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 25.9 997.22[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   No 10.6 8.9  Emotional support, %   Yes 15.5 146.13[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 16.6 238.55[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   No 12.2 8.8  Recipient   Partner, %    Yes 27.4 1306.50[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 21.0 748.68[\*](#table-fn2-0046958018775570){ref-type="table-fn"}    No 13.0 7.9   Child, %    Yes 32.9 2576.80[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 15.5 10.62[\*](#table-fn2-0046958018775570){ref-type="table-fn"}    No 12.4 13.2   Parents (-in-law), %    Yes 13.5 122.48[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 11.1 14.77[\*](#table-fn2-0046958018775570){ref-type="table-fn"}    No 16.3 13.8   Other family members, %    Yes 9.8 294.14[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 7.3 172.97[\*](#table-fn2-0046958018775570){ref-type="table-fn"}    No 15.5 15.1   Neighbors, friends, %    Yes 8.1 373.41[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 3.2 504.97[\*](#table-fn2-0046958018775570){ref-type="table-fn"}    No 15.5 16.3  Hours per week, %   1-5 hours 6.6 7346.92[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 4.2 1935.21[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   6-15 hours 19.0 11.6   \>16 hours 36.5 19.1  Duration, %   \<3 months 12.1 21.47[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 6.5 39.26[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   \>3 months 14.7 13.7 Roles of informal caregivers  Living situation, %   Families with children living at home    Yes 16.8 297.16[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 13.4 0.01    No 12.6 13.5  Employment status, %   Not currently employed 17.6 256.52[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 13.6 2.59   Employed 13.3 11.8  Financial difficulties, %   Yes 22.3 1231.26[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 20.0 147.61[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   No 12.1 12.1 Physical health of the informal caregiver  Perceived health status, %   (Very) poor 11.6 1888.41[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 7.7 879.52[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   (Very) bad 24.2 22.0  Number of chronic conditions, %   None 9.1 1626.13[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 5.3 531.69[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   1 13.1 9.5   2 18.1 14.2   3 19.3 17.8   4 24.6 21.8 Mental health  Experiencing depressive symptoms, %   None or low symptoms 8.5 5418.80[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 5.5 2134.79[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   Moderate symptoms 19.0 21.0   High symptoms 44.5 48.5  Loneliness, %   None 9.2 4203.01[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 8.2 1085.37[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   Moderate 19.8 16.4   (Very) severe 35.6 34.8 Needs of informal caregivers  No need for support, %   Yes 9.3 10,584.34[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 6.7 3902.44[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   No 46.6 46.8  Need for information and advice, %   Yes 43.6 4413.62[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 40.0 1142.34[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   No 12.1 10.9  Need for replacement, %   Yes 64.9 5027.75[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 60.6 2394.21[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   No 12.9 10.5  Need for emotional support, %   Yes 50.7 3147.69[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 48.6 887.20[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   No 13.1 11.9  Need for relaxing activities, %   Yes 50.4 2602.38[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 54.9 1145.88[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   No 13.3 11.8  Need for advocacy, %   Yes 60.0 2383.41[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 58.3 1073.06[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   No 13.6 12.0 Coping variables  Sense of mastery, %   Low 33.0 2033.79[\*](#table-fn2-0046958018775570){ref-type="table-fn"} 35.3 1403.70[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   High 12.9 9.7  Sense of self-management, %   Low 26.1 619.28[\*](#table-fn2-0046958018775570){ref-type="table-fn"}   Moderate 11.7   High 10.0 *P* \< .05. ### Logistic regression analysis {#section10-0046958018775570} For the adult informal caregivers, the final model accounted for 30.2% of the explained variance in perceiving low or high burden. With regard to the background variables, being a female caregiver was moderately associated with perceiving a high burden, while being 50 to 54 years of age and being widowed were strongly associated with perceiving a high burden. Having an intermediate I and II educational level seemed to be associated with a low burden. The relationship variables of providing personal and medical care were moderately associated with a high burden, while more hours of care provision per week was strongly associated with perceiving a high burden. Although role factors turned out to be nonsignificant, experiencing many depressive symptoms was strongly associated with a high burden. A moderate/severe level of loneliness was moderately associated with a high burden among caregivers, while perceived health status and mastery were not significantly associated with perceiving a high burden. Among the senior informal caregivers, the model explained 36.1% of the variance. Of the background variables assessed, only an intermediate II educational level was moderately associated with perceiving a low burden. The relationship variables of providing emotional support and personal/medical care support were moderately associated with perceiving a high burden. When more time was spent on providing care, this was strongly associated with a high burden. Although the role factors and perceived health status were not significantly associated with a high burden, a moderate/high number of depressive symptoms (strongly associated) and a severe loneliness level (moderately associated) were. Finally, a high sense of mastery was found to have a strong favorable association with the burden being perceived. An overview can be found in [Table 4](#table4-0046958018775570){ref-type="table"}. ###### Logistic Regression Analysis of Characteristics of Informal Caregivers With Regard to Perceiving a High Burden.  Independent variables High burden among adults (n = 3067) High burden among senior caregivers (n = 1936) -------------------------------------------------------- ------------------------------------- ------------------------------------------------------------------------- ------ -------------------------------------------------------------------------- Background variables  Age   25-39 vs 17-24 1.33 (0.57-3.11)   40-54 vs 17-24 3.03 (1.31-7.01)[\*\*\*](#table-fn4-0046958018775570){ref-type="table-fn"}   55-64 vs 17-24 1.90 (0.75-4.81)   75-79 vs 65-74 1.08 (0.70-1.69)   80+ vs 65-74 0.69 (0.15-3.08)  Gender   Female vs male 1.50 (1.02-2.22)[\*\*](#table-fn4-0046958018775570){ref-type="table-fn"} 0.90 (0.58-1.40)  Education   Intermediate II vs high 0.63 (0.42-0.97)[\*\*](#table-fn4-0046958018775570){ref-type="table-fn"} 0.49 (0.26-0.93)[\*\*](#table-fn4-0046958018775570){ref-type="table-fn"}   Intermediate I vs high 0.49 (0.31-0.77)[\*\*](#table-fn4-0046958018775570){ref-type="table-fn"} 0.55 (0.30-1.02)   Low vs high 0.59 (0.14-2.46) 0.65 (0.33-1.29)  Financial difficulties   Yes vs no 1.10 (0.77-1.56) 1.04 (0.60-1.79)  Marital status   Never married vs married 1.41 (0.72-2.75) 1.20 (0.33-4.43)   Divorced vs married 1.10 (0.65-1.87) 1.14 (0.35-3.66)   Widowed vs married 2.57 (1.10-6.04)[\*\*\*](#table-fn4-0046958018775570){ref-type="table-fn"} 0.69 (0.26-1.84)  Contacts with family   \<Once a week vs ≥once a week 0.97 (0.51-1.82)  Contacts with friends   \<Once a week vs ≥once a week 1.06 (0.75-1.51)  Contacts with neighbors   \<Once a week vs ≥once a week 0.76 (0.53-1.09)  Network type   Family-dependent vs locally integrated 1.27 (0.75-2.15)   Local self-contained vs locally integrated 0.84 (0.47-1.50)   Focused on wider community vs locally integrated 0.67 (0.21-2.12)   Private restricted vs locally integrated 0.73 (0.29-1.86) Relationship variables  Practical support   Yes vs no 1.23 (0.83-1.82) 1.10 (0.52-2.32)  Personal and medical care support   Yes vs no 2.13 (1.43-3.18)[\*\*](#table-fn4-0046958018775570){ref-type="table-fn"} 1.88 (1.22-2.90)[\*\*](#table-fn4-0046958018775570){ref-type="table-fn"}  Emotional support   Yes vs no 1.01 (0.56-1.83) 1.73 (1.13-2.76)[\*\*](#table-fn4-0046958018775570){ref-type="table-fn"}  Recipient 1   Partner: yes vs no 1.28 (0.73-2.25) 1.29 (0.54-3.08)  Recipient 2   Child: yes vs no 1.45 (0.88-2.37) 1.57 (0.66-3.77)  Recipient 3   Parents (-in-law): yes vs no 1.28 (0.83-1.99) 2.72 (0.96-7.72)  Recipient 4   Other family members: yes vs no 0.86 (0.51-1.44) 1.72 (0.72-4.11)  Recipient 5   Neighbors, friends: yes vs no 0.54 (0.28-1.02) 0.53 (0.25-1.11)  Hours per week   6-15 hours vs 1-5 hours 2.60 (1.75-3.86)[\*\*\*](#table-fn4-0046958018775570){ref-type="table-fn"} 2.02 (1.04-3.93)[\*\*](#table-fn4-0046958018775570){ref-type="table-fn"}   \>16 hours vs 1-5 hours 5.41 (3.29-8.89)[\*\*\*\*](#table-fn4-0046958018775570){ref-type="table-fn"} 6.27 (3.13-12.55)[\*\*\*\*](#table-fn4-0046958018775570){ref-type="table-fn"}  Duration   \>3 months vs \<3 months 0.98 (0.49-1.95) 2.16 (0.72-6.50) Roles, physical health, mental health of the caregiver  Families with children at home   Yes vs no 1.30 (0.88-1.91) 0.67 (0.24-1.86)  Employment status   Employed vs not currently employed 1.01 (0.67-1.52) 1.68 (0.74-3.84)  Perceived health status   (very) bad vs (very) good 1.41 (0.98-2.02) 1.34 (0.80-2.23)  Experience of depressive symptoms   Moderate vs none or low 1.45 (0.93-2.26) 3.24 (1.94-5.41)[\*\*\*](#table-fn4-0046958018775570){ref-type="table-fn"}   High vs none or low 3.15 (1.66-5.97)[\*\*\*](#table-fn4-0046958018775570){ref-type="table-fn"} 5.79 (2.40-14.00)[\*\*\*\*](#table-fn4-0046958018775570){ref-type="table-fn"}  Loneliness  Moderate vs none 1.76 (1.11-2.80)[\*\*](#table-fn4-0046958018775570){ref-type="table-fn"} 1.38 (0.85-2.22)   (very) severe vs none 2.21 (1.23-3.96)[\*\*](#table-fn4-0046958018775570){ref-type="table-fn"} 2.28 (1.23-4.21)[\*\*](#table-fn4-0046958018775570){ref-type="table-fn"} Coping variables  Sense of mastery   Yes vs no 0.73 (0.25-2.15) 0.36 (0.15-0.86)[\*\*\*](#table-fn4-0046958018775570){ref-type="table-fn"}  Sense of self-management   Moderate vs low 0.95 (0.41-1.28)   High vs low 1.18 (0.62-2.23) *Note.* OR = odds ratio; CI = confidence interval. Small (OR \< 1.5). \*\*Moderate (OR 1.5 ≤ 2.5). \*\*\*Strong (OR 2.5 ≤ 4.0). \*\*\*\*Very strong (OR \> 4.0). Discussion {#section11-0046958018775570} ========== This study explored the characteristics and needs of informal caregivers regarding their perceived burden. This section summarizes the main findings based on the variables included in the statistical analysis, that is the personal, contextual, relational and coping variables, and the needs of informal caregivers, and compares them with the findings of previous studies. As regards the personal factors, a moderate association with perceiving a high burden was found for female gender, while strong associations were found for caregivers aged 40 to 54 years and those who were widowed. A moderate association with perceiving a low burden and intermediate levels of education was determined. This is in line with what was reported by others.^[@bibr22-0046958018775570],[@bibr39-0046958018775570],[@bibr40-0046958018775570]^ Female caregivers from the "sandwich generation," ie, those who need to take care of their own children while supporting their parents,^[@bibr40-0046958018775570]^ were found to perceive a high burden. Caregivers with an intermediate educational level perceived a low burden. They are likely to be better able to handle the caregiving tasks, while feeling that their intellectual capacities are left unused.^[@bibr41-0046958018775570]^ Widowed caregivers perceived a high burden, probably because of their lower level of coping resources to address the current caregiving tasks. They might be unable to share their tasks and thoughts or were still experiencing grief and bereavement because of a spouse who had died.^[@bibr39-0046958018775570],[@bibr42-0046958018775570]^ Among the senior informal caregivers (aged 65 and above), only a moderate association was found between an intermediate educational level and perceiving a low burden, as was also reported.^[@bibr41-0046958018775570]^ A smaller social network mainly dependent on family members, eg, close family ties with few neighbors and peripheral friends, or focused on privacy, eg, absence of relatives and friends nearby and low levels of community involvement,^[@bibr43-0046958018775570]^ and feelings of loneliness were associated with perceiving a high burden. This is in line with the results of previous studies,^[@bibr20-0046958018775570],[@bibr44-0046958018775570],[@bibr45-0046958018775570]^ which indicated a higher burden when no support was received from the social network, probably because no back-up is available from family or friends. All relationship factors included in our study were associated with perceiving burden. This study, like the previous work by some,^[@bibr26-0046958018775570],[@bibr46-0046958018775570]^ showed that if more tasks were carried out by informal caregivers, especially personal and medical care tasks in both samples and emotional support in the senior sample, a moderate association was found with perceiving high burden. It was shown that people providing personal/medical care can feel uncomfortable or unable to provide this type of care,^[@bibr47-0046958018775570]^ and it was indicated that senior people who provide emotional support might not feel highly valued and respected by their care recipient, which increases their burden.^[@bibr48-0046958018775570]^ It was found that more hours of care and a longer duration of the caregiving relation were associated with a high burden.^[@bibr27-0046958018775570],[@bibr48-0046958018775570]^ In the current study, only the number of caring hours had a significant effect, which is in agreement with the findings in the study by Kenny, King, and Hall.^[@bibr49-0046958018775570]^ Although a poor perceived physical health status showed no significant association in our logistic regression analysis, this factor should be included in future assessments, as the presence of health problems may impede caregivers in providing care.^[@bibr7-0046958018775570],[@bibr15-0046958018775570],[@bibr20-0046958018775570]^ A higher number of depressive symptoms was strongly associated with perceiving a high burden. Caregivers have to deal with psychosocial strains because of their caregiving situation and people who are less able to deal with this tend to perceive a higher burden.^[@bibr20-0046958018775570]^ Favorable coping variables, that is, mastery and self-management, were suggested to mediate the studied relationship by reducing the perceived burden among caregivers. This was confirmed, where being able to manage your situation, as an informal caregiver, was associated with perceiving a low burden. In particular, caregivers with a high sense of mastery notice a positive influence. This indicates that when caregivers feel able to handle their care provision tasks, show a confrontational coping style, and have a good personal balance between their role as informal caregiver and their personal life, they are more likely to deal successfully with challenges associated with the provision of informal care.^[@bibr19-0046958018775570],[@bibr20-0046958018775570],[@bibr50-0046958018775570]^ Significant associations were found between the need for support among informal caregivers and perceiving a high burden. Expressed needs are typically a consequence of the intensity of the care provided, but might also reflect the burden perceived.^[@bibr51-0046958018775570]^ Although needs for information and advice and relaxing activities are frequently mentioned, this study showed that reporting a need for replacement was associated with perceiving a high burden. The stronger association may imply that the caregiver is not able to handle the caregiving tasks.^[@bibr52-0046958018775570]^ Finally, reporting a need for advocacy was strongly associated with perceiving a high burden, probably because informal caregiving is unrecognized and unsupported by society, in particular caregivers need more support from national and local government, the medical sector, and their employers.^[@bibr53-0046958018775570],[@bibr54-0046958018775570]^ By focusing on the needs replacement and advocacy, informal caregivers perceiving a high burden can be supported to be able to provide informal care in the future. The factors studied explained 30.2% of the variance in the burden perceived by the adult informal caregivers and 36.1% of the variance in the burden perceived by senior caregivers, indicating a small to moderate contribution to a high burden.^[@bibr55-0046958018775570]^ Even though the model that was used in the regression analysis was not complete, a relatively large percentage of the variance was explained. Variables such as perceived appreciation for the caregiving tasks, uplifts of caregiving, and the possibility to engage in activities distracting from the caregiving tasks were not included in this study. If it were possible to include all these variables as well, the percentage of explained variance could increase even further. Strengths and Limitations {#section12-0046958018775570} ------------------------- The major strength of this study is the large sample size and the inclusion of a large number of potentially important explanatory variables. The study was sufficiently powered to test multiple associations^[@bibr56-0046958018775570]^ and provides a comprehensive overview of the current knowledge and relevant concepts regarding the characteristics and needs of informal caregivers.^[@bibr57-0046958018775570]^ Besides these strengths, some limitations should be acknowledged. The cross-sectional design made it impossible to draw conclusions about causality, as it is unknown whether the selected characteristics actually preceded the occurrence of burden.^[@bibr58-0046958018775570]^ Second, there was a risk of selection bias in view of the voluntary participation, where some characteristics of the participants may have differed from those of nonparticipants. For example, people with a poorer general health status and with a low level of literacy are less likely to participate in studies like this, and participants who do not perceive themselves to be informal caregivers will not be included either.^[@bibr59-0046958018775570]^ Third, the data source consisted of self-reports, which is attended by increased risks of incorrect answers due to information bias. Caregivers might not be willing to share the burden they perceive, which may lead to inaccurate or socially desirable answers, and probably an underestimation of the actual burden on caregivers.^[@bibr60-0046958018775570]^ Fourth, although most characteristics were measured by a validated scale, some were not, eg, quality of life and perceived burden. These characteristics were measured on a Likert-type scale and eventually dichotomized. This was necessary to have sufficient numbers of participants in each subgroup, but might also lose some details. Fifth, the data used were gathered in 2012, which means that this study might underestimate the current level of burden being perceived and the associations, because the role of informal caregivers in the Dutch society has increased in recent years. Finally, there is the issue of generalizability, as only caregivers from the Dutch province of Limburg were included. The characteristics of informal caregivers can vary between areas. For instance, it is known that the general health status of the Limburg population is poorer and incomes are lower than in other areas of the Netherlands.^[@bibr61-0046958018775570]^ Recommendations for Future Research and Practice {#section13-0046958018775570} ------------------------------------------------ Based on our findings, 2 main recommendations for future research can be offered. First, longitudinal research is warranted to establish the causal directions of the found associations. This also provides better opportunities to develop a model, which can explain the risk factors for perceiving a high burden and also gives a better insight in the consequence of this burden. Furthermore, explorative research is needed to assess the importance of factors such as perceived appreciation for the caregiving tasks, uplifts of caregiving, and the possibility to engage in activities distracting from the caregiving tasks, as these were not included in the present study. If the importance of these characteristics can be confirmed, they may serve as screening criteria for selecting caregivers who might be at risk of overburdening. In consultation with caregivers who perceive a low or high burden, activities could then be developed to provide support tailored to their needs. For now, it seems that it is especially those informal caregivers spending more than 16 hours a week providing care who perceive the highest burden. In practice, possibilities should be developed to share the care tasks by involving other family members, while neighbors and friends might also take over some tasks, which might relieve the burden perceived by informal caregivers. Finally, the options should be searched to have a better cooperation with and support from the concerning municipalities and formal care providers, who might also be able to relieve or reduce the burden of informal caregivers. Conclusion {#section14-0046958018775570} ========== This cross-sectional study explored the population characteristics and needs of informal caregivers reporting a low or high burden. In less intensive stages of caregiving, informal care is often perceived to be a source of positive influence, while as the intensity of the care increases, more caregivers report to perceive high burdens. Different risk factors for perceiving a high burden could be identified: female gender, being aged 40 to 54 years, being widowed, providing emotional support, providing many hours of care, low sense of mastery, presence of depressive symptoms, and severe loneliness. Furthermore, caregivers reporting a need for replacement and advocacy may be the ones most at risk for perceiving a high burden. Although longitudinal research is warranted to establish the causal directions of these associations, focusing on these characteristics and needs is useful to relieve the perceived burden of informal caregivers, where a better cooperation with and support from the concerning partners in the municipalities is a necessity. **Declaration of Conflicting Interests:** The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. **Funding:** The author(s) received no financial support for the research, authorship, and/or publication of this article. **ORCID iD:** Lotte Prevo  <https://orcid.org/0000-0001-5469-5133>

Data were obtained from a third party, and are freely available from the following URL: <http://www.occ.gov/topics/capital-markets/financial-markets/trading/derivatives/derivatives-quarterly-report.html>. Introduction {#sec001} ============ After the financial crisis of 2008 the systemic risk resulting from OTC (over-the-counter) derivatives has become an important topic of public debate and scientific research. Different from exchange-traded derivatives, OTC derivatives are traded on non-regulated markets which have grown both in size and importance during the last decade. In December 2008 the Bank for International Settlements (BIS) reported (see Semiannual OTC derivatives statistics at <http://www.bis.org>) that total notional amount on outstanding OTC derivatives grew up from 370,178 bn USD in June 2006 to 683,725 bn USD in June 2008, i.e., it almost doubled in size in only two years. A particular worrying feature of this development results from the increasing concentration of the counterparty risk of OTC derivatives in the hands of only a few institutions. This trend has not changed after the financial crisis of 2008, on the contrary the concentration increased. Taking the example of the US alone, in the 4th quarter of 1998 contracts totaling 331 bn USD were signed by 422 commercial banks and trust companies which where not listed in the top 25 institutions dealing with OTC derivatives. This numbers have to be compared against the contracts totaling 32,668 bn USD (i.e., a hundred times more) signed by only the top 25 institutions in the OTC derivatives market. Comparing this to the time after the financial crisis, the difference became much bigger. In the 1st quarter of 2012 the 25 top ranked US institutions held contracts totaling 227,486 bn USD (i.e., almost ten times more than in 1998), whereas all other institutions held contracts totaling only 496 bn USD (which is almost comparable to what was held in 1998). Hence, we observe an extreme concentration of derivatives market where the share of derivative contracts held by the top 25 institutions was almost 99% in 1998 and increased to more than 99.5% in 2012. This increasing concentration may also increase the vulnerability of the institutions involved and can lead to cascades in case of default. Until now, no concentration of exposure against a particular counterparty is reported by banks. The Basel Committee on Banking Supervision referred to this issue for the first time only in its report of March 2013 \[[@pone.0136638.ref001]\]. In our paper, we address the problem in a twofold way. Based on a dataset of the 25 most active players in the U.S. derivative market, over a period of 14 years, we reconstruct the network of counterparty risk. We show that this risk generates an almost fully connected network of interdependence among these players, however it is skewly distributed, i.e., most of the counterparty risk is concentrated in only 10 mayor institutions. This implies two problems: in a fully connected network, it becomes much more difficult to hedge the risk of default, because every player is a counterparty of any other. This may increase the risk of default cascades, which can be amplified by the particularly active counterparties. Additionally, the concentration of counterparty risk in a few institutions may exacerbate the problem of contagion and financial distress in the whole network if those institutions become distressed. OTC Derivatives {#sec002} =============== The role of derivatives {#sec003} ----------------------- Derivatives are financial instruments, i.e., they are tradable assets. Importantly, they have no intrinsic value. Instead, their value depends on, or is *derived* at least partly from, the value of other entities, denoted as the "underlying". These can be other assets such as commodities, stocks, bonds, interest rates and currencies, but, dependent on the complexity of the financial product, the underlying can be almost anything that deemed to have an intrinsic value. This implies that socio-psychological issues such as "confidence", "faith" or "trust" play an important role in defining those values. Formally, derivatives are specified as contracts between two parties. Such contracts define how the value of the underlying is estimated at particular future dates and what conditions have to be fulfilled for payments between these parties. Because parties do not need to own the underlying, derivatives make for an ideal instrument to speculate about the future rising or falling value of underlyings or to hedge against the risk associated with it, provided that a counterparty is willing to bet on this. Trading derivatives basically means to find a counterparty for the contract. Importantly, parties can trade derivatives in two different ways, in regulated markets specialized in trading derivatives (ETD, exchange-traded derivatives) or privately, without involving an exchange or other institutions (OTC, over-the-counter derivatives). Although OTC markets are usually well organized, they are less formal. In particular, there is no central authority which would regulate the conditions of the derivative contracts or would control the fulfillment of these conditions. OTC derivatives are usually preferred over the exchange traded ones because taxes and other expenses are lower and they are much more flexible, meaning that the counterparties can agree on very specific or unusual conditions as opposed to the limited set of derivative types designed and operated by an exchange. As a trade-off for flexibility and the possibility of higher earnings OTC derivatives bear significant additional risks as compared to the exchange traded ones. Risk involved in OTC derivatives {#sec004} -------------------------------- Derivatives are generally used to hedge risks, but derivatives themselves are a source of risk. These are credit risk and market risk, along with liquidity, operational and legal risks \[[@pone.0136638.ref002]\]. In case of OTC derivatives, credit risk is the main source of risk because of the usual absence of a clearing house that guarantees the fulfillment of obligations between parties. Thus, the two contracting parties are exposed to *counterparty default risk*, i.e., the risk that a counterparty will undergo distress, or even default prior to expiration of the contract and thus will not make the current and future payments. In contrast to lending risk, to which only the party which lends is exposed, both sides involved in OTC contract are exposed to counterparty risk. To have some sort of mitigation, the parties involved in OTC derivatives are usually banks which act on their own behalf or on behalf of their clients. There are different ways to mitigate counterparty risk in case of default. For example, using close-out netting agreements allows that all contracts are netted, eliminating the possibility of selective execution of contracts \[[@pone.0136638.ref003]\]. For *bilateral close-out netting*, which mostly applies to OTC derivatives markets, the two parties agree to net with one another, i.e., to set off gains and losses from *all* of their bilateral contracts. This differs from the case of multilateral close-out netting which mostly applies to ETD, i.e., to markets where all parties' obligations are netted together. In both cases, netting is only a procedure to follow after a default and thus does not address the *emergence* of counterparty risk. It is obvious that netting decreases credit exposure, as it takes into account only the *net* obligations, thus reducing both operational and settlement risk and operational costs. In order to know the risk, the *present value* of contracts, prior to their contracted termination, has to be determined. Outstanding contracts are marked to market, taking into account the *replacement costs*, i.e., the loss suffered by the non-defaulting party in replacing the relevant contract. This assessment of credit exposure at a single point in time is denoted as *current credit exposure* (CCE). However, derivative contracts usually have considerable lifetimes and are very often characterized by fast and large changes in credit exposure. Therefore, the *potential future exposure* (PFE) is used to estimate the possible CCE increase over a fixed time frame. These estimates are, of course, predictions that depend on the choice of financial models and corresponding confidence level. The *total credit exposure* (TCE) is then measured as the sum of CCE and PFE, following the Basel I framework. In Section Correlations in risk we will use the TCE values reported by financial institutions to estimate *correlations* in their risk. Whereas netting agreements work in the absence of clearing houses, recent developments try to mitigate counterparty risk by means of *central counterparty clearing houses* (CCPs) \[[@pone.0136638.ref004]\]. In the presence of a CCP a bilateral contract between two counterparties is substituted by *two* contracts, so that the CCP stands between the two contracting parties. This allows for more transparency and for multilateral netting, which can facilitate the reduction of both counterparty and systemic risk. Although involvement of a CCP was previously required in contracts for *credit default swaps* (CDS) \[[@pone.0136638.ref005]\], a special class of derivatives, its broader utilization can be seen as a reaction to the financial crisis of 2008. However, regulations requiring CCPs in all standardized types of OTC derivatives are either new, e.g. the US Dodd-Frank Act from 2010, or are still being developed. Therefore, their impact on OTC derivatives markets is not well known yet, both empirically and theoretically. \[[@pone.0136638.ref006]\] recently attempted to shed some light on the possible systemic effects from CCPs. They performed a theoretical investigation of cascading effects and systemic risk in different financial networks with one or two CCPs. One may argue that not considering the role of CCPs in OTC derivatives networks is a limitation of this paper. But one should bear in mind that we analyse data ranging from 1998 to 2012, i.e., most of the time CCP were not required, and not reflected, in the OTC data. To keep our methods consistent for the whole time period, we neglect the possible (but not documented) presence of CCPs. Moreover, even today it is not known whether the wide adoption of CCPs will succeed in making the OTC derivatives network entirely transparent. So our methods to infer undiscovered and potentially dangerous links of the network may still be needed in the future. Clustering of counterparty risk {#sec005} ------------------------------- In this paper, we discuss a particular risk involved in OTC derivatives, namely the *clustering of counterparty risk*. While counterparty risk itself is already difficult to estimate, it becomes even more tedious for a party to find out about the *additional* risk that a counterparty bears because of it's involvement in *other* OTC derivatives. The problem is illustrated in [Fig 1](#pone.0136638.g001){ref-type="fig"}. It shows nine institutions that have in total ten different OTC contracts. The width of the links shall indicate the volume of these contracts, i.e., the three institutions 1, 2, 3 in the center (indicated by the dashed line) form a fully connected cluster of strongly engaged institutions. What is their implicit impact on those institutions outside the center? Each of these has only one contract with one of the major institutions in the center and is likely not aware of the whole structure of the network of OTC derivatives. {#pone.0136638.g001} There is a two-step scenario to increase the risk of the different institutions: (i) *Transfer of risk from the outer institutions to the central counterparty:* Institution 4 is probably not aware that its counterparty 1 also has contracts with institutions 5 and 6. If one of these outer institutions defaults, this puts an additional risk for institution 1 to default, which is likely not accounted for in the OTC contract between 4 and 1. Additionally, institutions 4 and 5 also have a contract which is likely not known to institution 1. Thus, the default of *either* 4 or 5 increases the risk for the remaining one, which indirectly increases the risk for institution 1 \[[@pone.0136638.ref007]\]. (ii) *Increase of risk between central institutions:* Because the center institutions form a fully connected cluster, if one of these undergoes distress or even defaults this immediately affects the other two core institutions. This in turn affects the outer institutions. In conclusion, because of the strong coupling of the center institutions, which we call *clustering of counterparty risk* here, all institutions indirectly have to bear (part of) the counterparty risk of all other institutions in the network. This should be priced in their OTC derivatives, but effectively it is not because that would imply to know (a) all the links and (b) all their weights or, in plain words, all the OTC contracts made. But, as explained above, the existence of OTC derivatives is precisely because such information should *not* be made publicly available. As we will see from the data, all public information only refers to the total amount of OTC derivatives for each institution, but not to their counterparty network. This sets the stage for our paper. Even in the absence of official information about the network of counterparty risk, we want to derive some insights into its structure, from a dataset described in the following. Specifically, we want to derive a proxy for the *structure* of this weighted, and time dependent, network. Further, we want to estimate correlations between OTC derivatives, i.e., infer on possible counterparties from the co-movement of the engagement of institutions. The Network of OTC Derivatives {#sec006} ============================== Activities and Ranks {#sec007} -------------------- In order to reconstruct the network of counterparty risk from the available dataset, we need to introduce a few variables that are later to be mapped to specific data. First of all, we identify each institution in the dataset by an index *i* = 1, ..., *N*, where *N* = 61, i.e., the total number of distinct institutions. Note that the dataset for each quarter only lists the 25 best ranked institutions, which are not necessarily the same for each quarter (see also [Fig 2](#pone.0136638.g002){ref-type="fig"}). Thus, during the whole period of 14 years, 61 different institutions appeared in the dataset. {#pone.0136638.g002} At each time step *t*, where *t* is discrete and measured in quarters, up to *T* = 57, institutions *i* and *j* can act as counterparties, i.e., they have contracts of total volume *x* ~*ij*~(*t*). Importantly, the dataset neither lists the counterparties *j* nor the volume of their contracts, *x* ~*ij*~(*t*). It lists, however, the quarterly activity of each institution, $a_{i}(t) = \sum_{j = 1}^{N}x_{ij}(t)$, i.e., the *aggregated volume*, given in column 5 of Table A in [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"} Thus, the aim of our paper is to reconstruct the network of dependencies from this aggregated data. Note that, if an institution was not active in a particular quarter, i.e., not listed in the dataset for that period, its activity is set to zero. To give an example, [Fig 3](#pone.0136638.g003){ref-type="fig"} shows the activity of two banks that are consistently engaged in OTC derivatives in every quarter. Impressively enough, their activities differ in about *two orders of magnitude* and further show a different business strategy over time. While the quarterly activity of *Keybank* remains almost constant over 12 years, the activity of *Bank of America* grew *exponentially* during the same period of time, clearly shown in the linear slope in the logarithmic plot. Only in 2012, after the financial crisis, this involvement was slightly reduced. {#pone.0136638.g003} Based on the quarterly activities, *a* ~*i*~(*t*), we can assign each institution *i* a rank *r* ~*i*~(*t*) ← *r*\[*a* ~*i*~(*t*)\] with *r* discrete and *r* ∈ {1, 2, ...*N*} such that *r*\[*a* ~*i*~(*t*)\] \< *r*\[*a* ~*j*~(*t*)\] if *a* ~*i*~(*t*) \> *a* ~*j*~(*t*) for any pair *i*, *j* ∈ *N*. I.e., rank 1 corresponds to the institution with the highest activity value at time *t*, rank 2 to the one with the second highest activity, and so forth. If an institution was not active in a given period, its rank is set to zero. Because the rank *r* ~*i*~ considers the position *relative* to other institutions, it can change even if the activity of an institution remained constant over a certain period. [Fig 2](#pone.0136638.g002){ref-type="fig"} gives an overview of how often the institutions were present in the ranking up to 25 in any of the quarters, with their ranks color coded. This matrix already indicates that there are remarkable fluctuations in the ranks of most of the institutions, except for a group of about 10 institutions. [Fig 4](#pone.0136638.g004){ref-type="fig"} gives a more detailed picture by plotting the ranks of this group over time. We observe that there exists a smaller core group (of about 7 members) with consistently low ranks, which can be well separated from a second group with higher, and more fluctuating, ranks. {#pone.0136638.g004} This can be also observed by looking at the ranks *R* ~*i*~ ← *r*\[*A* ~*i*~\] resulting from the aggregated activities $A_{i} = \sum_{t = 1}^{T}a_{i}(t)$. Plotting the inverse function *A*(*R*) shown in [Fig 5](#pone.0136638.g005){ref-type="fig"}, we observe a rather skew distribution of the aggregated activities with respect to the rank, with a skewness value *γ* = 4.637150 and a Gini coefficient \[[@pone.0136638.ref008]\]*g* = 0.9558996. Moreover, the plot suggests that the aggregated activity *A* follows a log-normal distribution with respect to the rank *R*: $$A\left( R \right) = \frac{1}{R\sigma\sqrt{2\pi}} \cdot \exp\left\lbrack - \frac{\left( \ln R - \mu \right)^{2}}{2\sigma^{2}} \right\rbrack\;;\;\; R \geq 1$$ where *μ* = 14.54116 is the mean value and *σ* = 2.865165 the standard deviation of the distribution. To further compare the empirical with the log-normal distribution, Fig A in [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"} shows the *Q* − *Q* plot and gives the results of the two-sample Kolmogorov-Smirnov test. {#pone.0136638.g005} The inset of [Fig 5](#pone.0136638.g005){ref-type="fig"} presents the cumulative distribution $P(R < Y) = \sum_{R = 1}^{Y}A(R)$. It indicates that about 95% of the total activity results from the seven first ranked institutions, while the 15 first ranked institutions cover more than 99% of the total activity. It may be tempting to restrict the analysis to only these 15 institutions. However, the aggregated activities do not allow to draw conclusions about the concentration of activities in certain time periods or a change of strategy in choosing counterparties, before and after the financial crisis. Therefore, we will present more details on the temporal activities in Section Temporal and aggregated networks. The available data also allows us to analyse the composition of the activities *a* ~*i*~(*t*) with respect to exchange traded derivatives (ETD) and OTC derivatives. I.e., the value of the total derivatives is split into $a_{i}(t) = a_{i}^{\text{ETD}}(t) + a_{i}^{\text{OTC}}(t)$ and $A_{i} = A_{i}^{\text{ETD}} + A_{i}^{\text{OTC}}$, respectively. Already the sheer numbers of the *a* ~*i*~(*t*) and $a_{i}^{\text{OTC}}(t)$ tell that OTC derivatives make up for the vast amount of contracts. I.e., we should not assume that the ranks *r* ~*i*~(*t*) or *R* ~*i*~ obtained from both ETD and OTC derivatives are different from those ranks that would result from only considering the values of $a_{i}^{\text{OTC}}(t)$ or $A_{i}^{\text{OTC}}$. To test this hypothesis, Fig B in the [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"} provides a *Q* − *Q* plot to compare both values. We see that up to rank 15 there is no difference in the ranks obtained by these two measures, whereas between ranks 15 and 50 the difference in ranks would be 1 or 2. Only for ranks above 50, the differences become remarkable. So it is reasonable to use the ranks *r* ~*i*~(*t*) and *R* ~*i*~ in the further evaluation. However, when analysing the counterparty risk in derivative contracts, we will make a distinction between the (less risky) ETD and the more risky OTC derivatives. In fact, as [Fig 6](#pone.0136638.g006){ref-type="fig"} indicates, the importance of OTC derivatives as compared to the ETD vastly differs across institutions. The ratio $A_{i}^{\text{OTC}}/A_{i}^{\text{ETD}}$ is below 10 for about 1/3 of all institutions, which implies that 10% or more of the activities is in ETD. However, looking at the 15 best ranked institutions, we see for most of them the ETD business accounts for only 2%-5% of their activity. So again, it is reasonable to proxy activities related to OTC derivatives by the total activities---but whenever possible, we will take into account the real values for OTC derivatives. {#pone.0136638.g006} Temporal and aggregated networks {#sec008} -------------------------------- In order to estimate the link structure of the network of counterparty risk, we first look into the co-occurrence of any two institutions among the 25 best ranked institutions in each given quarter. I.e., we define a link as *l* ~*ij*~(*t*) = 1 if for both institutions 1 ≤ {*r* ~*i*~(*t*), *r* ~*j*~(*t*)} ≤ 25 and *l* ~*ij*~(*t*) = 0, otherwise. Their co-occurrence does not necessarily imply that the two institutions are counterparties of an OTC derivative. A ranked institution *i* could do all its OTC contracts with the many institutions that have ranks too high (i.e., activities too low), to be listed in this dataset. Practically, however, this cannot be the case because, as the OCC reports verify, already 99% of all OTC derivatives are held by the 25 best ranked institutions. So, the not listed ones would make only for 1%, which cannot explain the large activities of any of the 25 best ranked institutions. Consequently, it is reasonable to assume that *i* has at least one contract with any of the other 24 institutions, and the best ranked institutions have likely more than one. The co-occurrence network certainly overestimates the business relations based on OTC contracts because it is basically a fully connected network between the 25 best ranked institutions. Further, the co-occurrence may change in each quarter. Therefore, as the next step, it is reasonable to assign weights for the links between any two institutions based on the number of quarters, they co-appear in the dataset. I.e., we define weights as $$w_{ij} = \frac{1}{T}\sum\limits_{t = 1}^{T}l_{ij}\left( t \right)$$ to normalize them to the available time period. A node that has links with high weights to its neighbors certainly represents an important institution in the OTC derivatives market. We use the weights to define the importance of an institution as $W_{i} = \sum_{j = 1}^{N}w_{ij}$. In the following network figures, the size of the nodes is scaled to the *normalized* importance, *W* ~*i*~/∑~*i*~ *W* ~*i*~. This allows us now, based on the aggregated values, to draw in [Fig 7](#pone.0136638.g007){ref-type="fig"} a first approximation of the network of counterparty risk. While this figure clearly shows the important institutions with respect to their *co-occurrence*, it neglects another important information, namely their *ranking* which is a proxy of their relative *activity*. Imagine institution *i* with a steady but relatively low activity over time, just enough for frequently appearing in the network, while institution *j* may have a much higher activity, but during a shorter period of time, resulting in a better, but less frequent ranking. As a result, institution *i* will be over-presented in the network drawn in [Fig 7](#pone.0136638.g007){ref-type="fig"}, while institution *j* will be under-represented. Such activity differences are prevalent in the dataset as the investigations in Section Activities and Ranks show. In the example shown in [Fig 3](#pone.0136638.g003){ref-type="fig"}, the activity of *Keybank* was two to three orders of magnitude lower than the activity of *Bank of America*. But because *KeyBank* was present in the top 25 list during the whole time period, it gained a similar position in the network in [Fig 7](#pone.0136638.g007){ref-type="fig"} as giants such as *Bank of America* or *Citibank*. {#pone.0136638.g007} Therefore, to further improve our estimation of the network of counterparty risk, we take into account the overall activity of an institution by using their ranks to assign weights to the links of co-occurrence. I.e., instead of *l* ~*ij*~(*t*) = 1, we use $$l_{ij}\left( t \right) = \min\left\{ \frac{1}{r_{i}\left( t \right)},\frac{1}{r_{j}\left( t \right)} \right\}$$ The rationale behind is to bind the weight of a link to the activity of the *less active* institution. To elucidate this, let us assume that institution *i* is a big player with rank *r* ~*i*~(*t*) = 2 at time *t*, while *j* is a less important institution with rank *r* ~*j*~(*t*) = 21. Because both institutions co-appear in the same quarter, each of them has links to all other institutions listed in the same time period, i.e., 24 links. For the less important institution *j*, 20 of these links get assigned a weight of 1/20, namely those links to institutions with better ranks. But there are 4 links to institutions with an activity less than *j* and therefore with higher ranks. Those links get assigned the weights 1/22, 1/23, 1/24, 1/25. I.e., for each institution, links to less active counterparties have less weight, while links to more active counterparties have the maximum weight that could occur given the rank of that institution. Likewise, for institution *i* only one link, namely the link to the highest ranked institution, gets a weight 1/2, whereas the 23 links to all other institutions become less and less important as 1/3, 1/4, ..., 1/25. The resulting network is shown as an animation (at the time of writing only supported in Adobe products) in Fig D in [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"}. At each time step this is a fully connected network, but the weights of the links, as well as the importance of the institutions, change during every timestep. The animation nicely elucidates the emergence of new key players in the OTC derivatives markets before and after the crisis, as well as the changed preferences in choosing counterparties. To allow a comparison with [Fig 7](#pone.0136638.g007){ref-type="fig"}, we aggregate the weights of the links over time according to [Eq (2)](#pone.0136638.e013){ref-type="disp-formula"}, to take into account both co-occurrence and activity, and calculate the importance of an institution as before, $W_{i} = \sum_{j = 1}^{N}w_{ij}$. The resulting weighted network is then shown in [Fig 8](#pone.0136638.g008){ref-type="fig"}, which should be compared to [Fig 7](#pone.0136638.g007){ref-type="fig"}. The most obvious difference is a less dense core, built up by a smaller number of important institutions, in [Fig 8](#pone.0136638.g008){ref-type="fig"}. Tracing particular institutions, e.g. *Union Bank*, we see that their position becomes less influential. But the core of the network, i.e., the set of the ten most important institutions, remains the same and shall be investigated in the following. {ref-type="fig"}, but the *w* ~*ij*~ and *W* ~*i*~ are calculated from the *l* ~*ij*~ as given by [Eq (3)](#pone.0136638.e015){ref-type="disp-formula"} i.e., dependent on the ranks. The time resolved network is shown in Fig D in [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"}. The aggregated network should be compared with [Fig 7](#pone.0136638.g007){ref-type="fig"} where activities are not taken into account.](pone.0136638.g008){#pone.0136638.g008} Core-periphery structure {#sec009} ------------------------ So far, we have used the following information to describe counterparty relations: (i) *Aggregated measures* derived from the aggregated *co-occurrence* *l* ~*ij*~ in the ranking of the 25 top players in the OTC market, in particular the *weights* *w* ~*ij*~ and the *importance* *W* ~*i*~. The results are concluded in the network of [Fig 7](#pone.0136638.g007){ref-type="fig"}. (ii) *Temporal measures* derived from the *ranking* *r* ~*i*~(*t*), in particular the temporal *co-occurrence* *l* ~*ij*~(*t*). The results are concluded in the animated network of Fig D in [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"}, with the time-aggregated network shown in [Fig 8](#pone.0136638.g008){ref-type="fig"}. While the latter can be seen as the most refined network of counterparty risk, the characterization of both nodes and links is still based on the *activity* *a* ~*i*~(*t*) of the respective institution, i.e., it is derived from a single scalar measure. So, the question is whether the reconstruction of the aggregated temporal network would allow us to add another dimension to characterize institutions, based on *topological* information. Already a visual inspection of Figs [7](#pone.0136638.g007){ref-type="fig"} and [8](#pone.0136638.g008){ref-type="fig"} verifies that the network is rather heterogeneous with respect to its density. We can easily detect a *core* of larger (i.e., more active) and more densely connected nodes which can be distinguished from a *periphery* of nodes that are smaller (i.e., less active) and less densely connected. In fact, peripheral nodes are mostly connected towards the core and much less to other peripheral nodes. The core of the network is depicted in [Fig 9](#pone.0136638.g009){ref-type="fig"} and gives a good impression of the fully connected network, albeit with links of different weights. {ref-type="fig"}.\ For the coding see the legend in [Fig 8](#pone.0136638.g008){ref-type="fig"}.](pone.0136638.g009){#pone.0136638.g009} Whether institutions can be found in the core or in the periphery of the network certainly relates to their importance in the OTC market. In order to quantify the topological information encoded in the network structure, we use the *weighted *K* core analysis*, which is an established method to assign an importance value to nodes. In the first step, for the time aggregated network shown in [Fig 8](#pone.0136638.g008){ref-type="fig"}, each node gets assigned a *weighted degree* ${\hat{k}}_{i}$\[[@pone.0136638.ref009]\]: $${\hat{k}}_{i} = \left\lbrack k_{i}^{\alpha}\mspace{180mu}\left( \sum\limits_{j}^{k_{i}}w_{ij} \right)^{\beta} \right\rbrack^{\frac{1}{\alpha + \beta}},$$ where *k* ~*i*~ is the degree of node *i*, i.e., its number of links to neighboring nodes, and $\sum_{j}^{k_{i}}w_{ij}$ is the sum over all its link weights as defined in [Eq (2)](#pone.0136638.e013){ref-type="disp-formula"} with the weighted *l* ~*ij*~ given by [Eq (3)](#pone.0136638.e015){ref-type="disp-formula"}. The exponents *α* and *β* are used to weight the two different contributions, i.e., *number* of links versus *weight* of links. In our analysis we used *α* = 0 and *β* = 1, i.e., we focused only on the weights since the network is almost fully connected and the node degree does not give us any information. In the second step, we follow a pruning procedure to recursively remove all nodes with degree $\hat{k} \leq K$ from the network, where *K* = 1, 2, ... I.e., first all nodes with $\hat{k} \leq 1$ are removed, which may leave the network with other nodes that now have $\hat{k} \leq 1$ simply because some of their neighbors were removed. So the procedure continues with removing these nodes, too, unless *no* nodes with $\hat{k} \leq 1$ are left. Then all nodes removed during this step get assigned to a *core* *K* = 1, and the procedure continues to successively remove all nodes with degree $\hat{k} \leq 2$ and assign them to a core *K* = 2, etc. The procedure stops at a certain high core value, *K*, when all nodes are removed. The higher the *K*-core a node is assigned to, the more it belongs to the "core" of the network and the more important it is, from a topological perspective. Evidently, nodes assigned to a core with low *K* value are much less *integrated* in the network. This does not refer simply to the number of neighbors, but also to non-local properties such as the number of neighbors of their neighbors, because the *K*-core decomposition also takes these into account. That means, the *K*-core a node is assigned to reflects is position in the network much better than simple measures such as the degree (i.e., the number of neighbors), alone. The results of the weighted *K*-core analysis are shown in the left side of Fig C in [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"}, where the *K* value is normalized to 1. Based on their *K* value, institutions can be ranked such that the higher the *K* value (i.e., the better the integration in the network), the better the rank. This *topological* ranking does not necessarily coincides with the ranking *R* ~*i*~ obtained from the aggregated *activity* *A* ~*i*~ which is shown on the right side of Fig C in [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"}, for comparison. This indicates that *structural* measures based on the network *topology* indeed provide information different from the *temporal* measures based on the market activities of the institutions. But, comparing the left and the right sides with respect to the color coding, we observe that only in a few cases institutions have considerably different levels of importance dependent on the measurement. It would be worth looking at these in a case-by-case study, to find out which importance measure better reflects their overall performance in the financial market. We would like to note that, for consistency, we have used the ranking obtained from the weighted *K*-core analysis to sort the different institutions in all the figures. Correlations {#sec010} ============ Correlation measures {#sec011} -------------------- So far, we have analysed the *co-occurrence* of financial institutions in the set of the 25 best ranked institutions, weighted by their ranks. These ranks were based on their activities, i.e., *total derivatives*. As a result, we could reconstruct the weighted network of counterparty risk which also reflects the importance of the nodes. This network was reconstructed (a) on a time resolution of one quarter year, to show the dynamics of the network (Fig D in [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"}), and (b) on the time aggregated level ([Fig 8](#pone.0136638.g008){ref-type="fig"}). To further analyse the mutual dependence between the best ranked institutions, we now calculate different correlations. The network of counterparty risk has revealed how the co-occurrence changes over time. But will the OTC derivatives of institution *i* increase, or decrease, if the same measure of institution *j* increases? Answering this question allows some more refined conclusions about the dependence between these institutions. The simplest measure is the *Pearson correlation coefficient* *ρ*, which points to a *linear* dependence between two variables. As explained above, for each institution *i* we have a dataset **a** ~*i*~ = {*a* ~*i*~(1), *a* ~*i*~(2), ..., *a* ~*i*~(*T*)} available which contains up to *T* entries about its quarterly activity *a* ~*i*~(*t*) measured by means of its total derivatives. We recall that some of these entries are zero whenever institution *i* was not listed among the best 25 ranked. Let us define the mean value and the standard deviation of each of these samples as: $${\overline{a}}_{i} = \frac{1}{T}\sum\limits_{t = 1}^{T}a_{i}\left( t \right)\;; s_{i}^{a} = \sqrt{\frac{1}{T - 1}\sum_{t = 1}^{T}\left\lbrack a_{i}\left( t \right) - {\overline{a}}_{i} \right\rbrack^{2}}.$$ The Pearson correlation coefficient with respect to the variable *a* is then defined as $$\rho_{ij}^{a} = \frac{1}{T - 1}\sum\limits_{t = 1}^{T}\mspace{180mu}\left\lbrack \frac{a_{i}\left( t \right) - {\overline{a}}_{i}}{s_{i}^{a}} \right\rbrack\mspace{180mu}\left\lbrack \frac{a_{j}\left( t \right) - {\overline{a}}_{j}}{s_{j}^{a}} \right\rbrack.$$ Values of *ρ* can be between -1 and +1. The latter indicates that the relation between activities *a* ~*i*~ and *a* ~*j*~ can be perfectly described by a linear relationship, where *a* ~*i*~ increases as *a* ~*j*~ increases. -1, on the other hand, indicates a perfect linear relationship where *a* ~*i*~ decreases as *a* ~*j*~ increases, and vice versa. Zero would indicate that there are no linear dependencies detected in the data. [Eq (5)](#pone.0136638.e026){ref-type="disp-formula"} also shows that, in case of a positive correlation, if $a_{i}(k) > {\hat{a}}_{i}$ then also $a_{j}(k) > {\hat{a}}_{j}$ for most of the time, and if $a_{i}(k) > {\hat{a}}_{i}$ then also $a_{j}(k) > {\hat{a}}_{j}$ for most of the time, i.e., the activities of both institutions are mostly above (or below) their respective average, at the same time. Correlations in activities {#sec012} -------------------------- We first discuss the results for the most active institutions, i.e., those appearing among the 25 best ranked institutions with respect to their total derivatives in every quarter. Interestingly, this applies only to 8 out of the 61 listed institutions. [Fig 10](#pone.0136638.g010){ref-type="fig"} shows the correlation matrix for these institutions, their activities proxied by the total notional amount of derivative contracts as listed in column 5 of Table A in [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"}. {#pone.0136638.g010} There are two observations to be made: (i) the correlations between any two of these institutions are always positive and often even close to 1, (ii) *Keybank* is a noticeable exception. This can be explained by the combination of two effects: The first one is the vastly growing market in OTC derivative during the observation period which resulted in the growth of OTC derivatives for these core institutions. Thus, the observed correlations could, in principle, be caused by the underlying market dynamics rather than by the mutual interaction. However, taking into account that the 10 best ranked institutions already account for 95% of the OTC derivatives market, there is little room for the assumption that their growth is based on OTC derivative contracts with institutions that do not belong to the core of 10, or to the 25 best ranked institutions. In conclusion, these eight institutions increased their OTC derivatives activities by repeatedly choosing the same core institutions as counterparties. The low correlations for *Keybank* could result both from the absence of growth (see [Fig 3](#pone.0136638.g003){ref-type="fig"}), while all others were growing, and from choosing counterparties from outside the set of core banks. If we wish to extend this correlation analysis to the whole set of 61 institutions, it would generate a number of artifacts which should be avoided. We discuss them here, first, to motivate our own approach presented afterward. As already shown in [Fig 2](#pone.0136638.g002){ref-type="fig"}, most of these institutions were not present in the ranking of the best 25, for some longer or shorter period. So, one could limit the correlation analysis to those quarters where the two institutions were indeed present in the ranking. I.e., if institution *i* appeared at times *t* ~1~, *t* ~2~, *t* ~3~, *t* ~4~ and while institution *j* appeared at times *t* ~2~, *t* ~4~, *t* ~5~, *t* ~6~, the correlation coefficient for them is computed using only the observations at times *t* ~2~ and *t* ~4~ where both were present. The Pearson correlation coefficient based on pairwise available observations with respect to the variable *a* is then defined as $$\rho_{ij}^{a} = \frac{1}{\#\left\lbrack \mathcal{T}_{i} \cap \mathcal{T}_{j} \right\rbrack - 1}\sum\limits_{t \in \mathcal{T}_{i} \cap \mathcal{T}_{j}}\left\lbrack \frac{a_{i}\left( t \right) - {\overline{a}}_{i}}{s_{i}^{a}} \right\rbrack\mspace{180mu}\left\lbrack \frac{a_{j}\left( t \right) - {\overline{a}}_{j}}{s_{j}^{a}} \right\rbrack$$ 𝒯~*i*~ and 𝒯~*j*~ are subsets of {1, 2, ..., *T*}, comprising the time steps when the institutions *i* and *j* appeared in the ranking among the top 25, and \#\[𝒯~*i*~\] and \#\[𝒯~*i*~\] are the numbers of these time steps. 𝒯~*i*~∩𝒯~*j*~ then defines the subset of timesteps where *both* institutions *i* and *j* appeared together, and \#\[𝒯~*i*~∩𝒯~*j*~\] gives the respective number of those time steps. Consequently, the average activity ${\bar{a}}_{i}$ and the standard deviation $s_{i}^{a}$ are also calculated only for the subset 𝒯~*i*~: $${\overline{a}}_{i} = \frac{1}{\#\left\lbrack \mathcal{T}_{i} \right\rbrack}\sum\limits_{t \in \mathcal{T}_{i}}a_{i}\left( t \right)\;; s_{i}^{a} = \sqrt{\frac{1}{\#\left\lbrack \mathcal{T}_{i} \right\rbrack - 1}\sum_{t \in \mathcal{T}_{i}}\left\lbrack a_{i}\left( t \right) - {\overline{a}}_{i} \right\rbrack^{2}}$$ The results of this analysis are shown in Fig E in [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"}. We observe that, in addition to the strong correlations in the core of those institutions always present, there are a lot of strongly *anti-correlated* activities (indicated by rich red) among the low ranked institutions which need to be interpreted, both with respect to the correlation and to the magnitude. We start with the latter. Defining the Pearson correlation coefficient according to [Eq (6)](#pone.0136638.e031){ref-type="disp-formula"} has the drawback that the correlation coefficients for different institutions are no longer normalized to the same number of observations, *T*, as in [Eq (5)](#pone.0136638.e026){ref-type="disp-formula"} and thus cannot be compared. Precisely, the correlations between *Bank of America* and *Citibank*, which were both present in the ranking for *T* = 57 quarters will get the same weight as the correlations between *Citibank Nevada* and *Chase Manhattan Bank USA* which were present together only two times. The second drawback results from the time lapse between the co-appearance. While the times *t* ~4~ and *t* ~6~ in the above example may still be relatively close, the interval between *t* ~4~ and *t* ~56~ would be much longer and, because of the unknown intermediate values, interpretations about the correlated move of both institutions become highly speculative. In contrast to the above example, in which the two intermediaries appear only in a few quarters, but yet co-appear twice, some pairs of intermediaries which are important both by means of long term presence and good rankings, never appeared *together*, for example *Goldman Sachs* and *Bank of New York*, and, as a consequence, the Pearson correlation coefficient is not even defined for them, which is yet another drawback. One could argue that these drawbacks disappear if we simply keep the normalization *T*, as in Eqs ([4](#pone.0136638.e025){ref-type="disp-formula"}) and ([5](#pone.0136638.e026){ref-type="disp-formula"}), and instead assign an activity *a* ~*i*~(*t*) = 0 whenever an institution *i* is not present in the ranking. While there is no evidence that the activity was indeed zero, the error produced this way is certainly small because of the very skew distribution of activities shown in [Fig 5](#pone.0136638.g005){ref-type="fig"}, and both the mean and the standard deviation of the activity are not substantially affected. But it becomes a problem when there is indeed no data because the institution does not exist in certain quarters, e.g. because of mergers and acquisitions, as in the case of *Chase Manhattan Bank* and *JPMorgan Chase Bank*. Additionally, by proceeding like this we would generate another artifact, namely generating artificial correlations between those institutions that are often not in the rankings and, in the worst case, never co-appear. It is in fact the absence of data that generates their correlations, artificially. Taking again the example of *Goldman Sachs* and *Bank of New York*, these two institutions would then appear *anti-correlated* while, in fact, no correlation was defined for them. Thus, solving the above mentioned drawbacks this way would generate yet a different one. Consequently, we will go with the correlations defined on the pairwise co-appearance, [Eq (6)](#pone.0136638.e031){ref-type="disp-formula"}, but we compensate for the different normalization by multiplying the correlation coefficients $\rho_{ij}^{a}$ with the weights *w* ~*ij*~ defined in [Eq (2)](#pone.0136638.e013){ref-type="disp-formula"} with *l* ~*ij*~ = 1, which is the relative number of co-appearances. This implies that the correlations between two institutions that rarely co-appeared in the ranking are scaled down. Precisely, after this correction, the weights *w* ~*ij*~ define the bounds of the values of the correlation coefficients, which are different for each pair of institution, namely \[−*w* ~*ij*~, + *w* ~*ij*~\] instead of \[−1, +1\]. These weighted correlation coefficients shall be interpreted differently from the conventional correlation coefficients in that a close-to-zero coefficient no longer means that the variables are uncorrelated, but that there is no *significant* correlation because of the low weight. The resulting correlation matrix is shown in [Fig 11](#pone.0136638.g011){ref-type="fig"}. Compared to the non-scaled Fig E in [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"}, both the correlated and the anti-correlated activities loose importance for institutions with higher ranks, because the co-appearance in the ranking is rather sparse. But still, it is obvious that the correlated activities are concentrated in the core, while the anti-correlated activities can be mostly found in the periphery. Keeping in mind the exponential growth of the derivative volume of some of the key players, as shown in [Fig 3](#pone.0136638.g003){ref-type="fig"}, it means that the OTC market acted rather *heterogeneous*. Most banks with high ranks, i.e., key players, increased their activities in a growing market. Banks with lower ranks, such as *First National Bank of Chicago* or *RBS Citizens*, have either reduced their overall OTC exposure or have concentrated their activities towards only mayor institutions, avoiding other low ranked institutions. {#pone.0136638.g011} Correlations in risk {#sec013} -------------------- So far, we have only analysed correlations in *activities*, i.e., the correlated increase or decrease in OTC derivatives volumes between any two institutions. We found that the correlated behavior was the dominating one which, together with a vastly growing OTC market, implies that most institutions increased their involvement. The question remains what this would mean for the *risk* of the counterparties, We already mentioned in Section Risk involved in OTC derivatives that credit risk is the main source of risk for banking institutions. To estimate the *total credit exposure* (TCE), we sum up their *current credit exposure* (CCE) and their *potential future exposure* (PFE) as explained in Section Risk involved in OTC derivatives. This data has been made available in "Table 4" of the OCC reports for each quarter year (see Table B in [S1 Appendix](#pone.0136638.s001){ref-type="supplementary-material"}) and is used for our subsequent correlation analysis. "Table 4" lists, for each of the 25 first ranked institutions, the *bilaterally netted current credit exposure* and the *bilaterally netted potential future exposure* and the sum of both, TCE = CCE+PFE, as reported by the institutions themselves. Looking at Q1 of 2012, we first notice that, for the high ranked institutions (according to their activity in OTC derivatives), the potential future exposure exceeds considerably the current exposure, which is generally not the case for the lower ranked institutions. The question whether this observation is related to the financial crisis of 2008 is addressed further below. We can now define a correlation coefficient $\rho_{ij}^{\text{TCE}}$ based on [Eq (6)](#pone.0136638.e031){ref-type="disp-formula"} by just replacing the values of the activities *a* ~*i*~(*t*) by TCE~*i*~(*t*), and for $\rho_{ij}^{\text{CCE}}$ accordingly. Following the argumentation above, we weight these correlations again by the weights *w* ~*ij*~. The results are shown in [Fig 12](#pone.0136638.g012){ref-type="fig"}. Both figure parts indicate that, at least for the subset of banks which are the closest to the core according to the core-periphery decomposition, the credit exposures are highly positively correlated. This indicates that the core of the network consists of institutions which are very strongly interdependent. This can become a reason for systemic instability, as the credit exposures and the connected risks cannot be well diversified. {#pone.0136638.g012} The correlation pattern for the risk resembles the one found for the activities, [Fig 11](#pone.0136638.g011){ref-type="fig"}. We have to note, however, that a large correlation coefficient $\rho^{a_{ij}}$ is a good indicator of a long-term activity between institutions *i* and *j*, but a large correlation coefficient $\rho_{ij}^{\text{TCE}}$ does not allow us to derive such a conclusion. Up to this point the analysis was based on the whole available period of time (1998---2012). It is interesting to repeat the correlation analysis of risk for the time before and after the financial crisis, separately. We avoid to discuss the precise mapping of "before" and "after" and have chosen Q4 of 2008 to divide the time series into two periods. In Q4 of 2008 *Goldman Sachs* entered the ranking of the OCC, for the first time, right after the collapse of *Lehman Brothers* on 15 September, 2008. The results of our analysis before and after Q4 of 2008 are shown in [Fig 13](#pone.0136638.g013){ref-type="fig"}. Comparing the two parts of the figure, we make two observations: (i) All listed banks follow a similar behavior before and after the crisis. But after the crisis the correlations became more homogeneous and non-negative even between low-to-low ranked and low-to-high ranked institutions. (ii) Except only few banks, the key players in the core did not change. Therefore, the OTC derivatives market structurally remained the same despite its vast growth. {#pone.0136638.g013} Conclusions {#sec014} =========== Our investigation reveals the *hidden* network structure behind the OTC market in the United States, and the network evolution from 1998 to 2012. For this, we use publicly available data from the \[[@pone.0136638.ref010]\] reports, which contains aggregated numbers about the activities of financial institutions, measured by the volume of their different derivatives. We focus on two different aspects: (i) *co-occurrence patterns* of institutions, which take into account their ranks and activities to reconstruct the network of counterparty risk. This network was further analysed using of a weighted k-core method, to reveal its core-periphery structure. This allowed us to compare the topology-based ranking with the activity-based ranking, and to identify the most important institutions and their mutual relations. (ii) *correlation patters*, to reveal *dependencies* in activities, and the subsequent counterparty risks of any two institutions. Our findings, namely an emergence of a pronounced core and the higher correlations in credit exposure associated with it, hint at *increasing* counterparty and systemic risk in OTC derivatives market. One could argue that the list of the few top institutions with the highest counterparty risk is not really surprising and financial experts would have known this anyway. But the point of our investigation is to present a formal, yet simple approach, to *decompose* their *known aggregated activities* into *unknown bilateral exposures*. Only this allows us to reveal the hidden *network*, and to estimate the *systemic risk*. Counterparty risk is not just the sum of individual risks, but can be amplified over the network of dependencies. Precisely, the failure of single institutions, even in the periphery of the OTC network, can lead to the collapse of the whole system because of distress and load distributed over the network \[[@pone.0136638.ref011]\]. Such considerations do not only enhance our understanding of systemic risk, they also allow to develop more refined risk measures, and a more realistic pricing of OTC contracts. This network perspective is missing in existing investigations \[[@pone.0136638.ref012], [@pone.0136638.ref013]\] on systemic risk in OTC derivatives markets. It moves the focus from discussing netting procedures *after* a default to the more important question of how systemic risk *emerges*, i.e., what happens *before* a default. To conclude, our investigations contribute to the ongoing debate about the impact of the OTC derivatives market on the stability of the financial markets. We support the position that OTC derivatives increase financial instability, because they generate a hidden network of dependencies that at the end increase the chance of failure cascades. This has not become obvious because of the bilateral nature of counterparty risk and the lack of transparency in OTC markets. But our simple and practical method allows to at least estimate this hidden network of additional dependencies and to better estimate, and price, the risk resulting from these. It particularly points to the limitations in diversifying risk in such markets and the need to implement further regulations, as already proposed by the European Market Infrastructure Regulation (EMIR) under the Basel III umbrella BCBS (2011). Supporting Information {#sec015} ====================== ###### Supporting figures, animation and tables. (PDF) ###### Click here for additional data file. The authors thank Stefano Battiston and Paolo Tasca for discussions, and Michelangelo Puliga for pointing us to the dataset. We acknowledge financial support from the Project CR12I1_127000 *OTC Derivatives and Systemic Risk in Financial Networks* financed by the Swiss National Science Foundation. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: VN AG FS. Performed the experiments: VN. Analyzed the data: VN AG. Contributed reagents/materials/analysis tools: AG VN FS. Wrote the paper: FS VN.

The authors confirm that all data underlying the findings are fully available without restriction. All relevant data are within the paper and its Supporting Information. Introduction {#s1} ============ Human noroviruses (HuNoV) are the most common cause of acute viral gastroenteritis worldwide [@pone.0106805-Glass1]. Members of the *Caliciviridae* family, these viruses are transmitted by a variety of routes, but frequently cause outbreaks in closed settings such as schools, nursing homes, and cruise ships. Contamination of foods and water is another common transmission mode, as HuNoV are the leading cause of foodborne disease in the U.S. [@pone.0106805-Scallan1] and perhaps worldwide [@pone.0106805-Patel1] [@pone.0106805-Ahmed1]. Low infectious dose, high virus concentrations in the feces and vomitus of infected individuals, lengthy environmental persistence, and resistance to many commonly used sanitizers and disinfectants all contribute to the high degree of transmissibility of HuNoV [@pone.0106805-Hall1]. Despite their public health significance, routine detection of HuNoV in community settings or in food and environmental samples, is limited. Firstly, there is no cell culture model to propagate these viruses. Secondly, HuNoV have tremendous antigenic diversity which has complicated the development of broadly reactive antibodies, meaning that enzyme immunoassays have poor sensitivity [@pone.0106805-Costantini1] [@pone.0106805-Kele1]. While molecular amplification methods (specifically reverse transcriptase quantitative PCR or RT-qPCR) are commonly used in public health settings, these methods are rarely used in clinical diagnostic laboratories in the U.S. Detection of HuNoV in food and environmental samples is even more complicated because virus concentrations are so low in these samples that it is necessary to perform labor intensive and relatively inefficient pre-concentration step(s) prior to detection [@pone.0106805-Knight1]. There is a need to develop alternative HuNoV diagnostic reagents to complement existing ones. Nucleic acid aptamers are short ssDNA or ssRNA sequences having binding affinity for a target molecule, like bacteria, viruses, or cells. Once identified, they offer advantages over other binding ligands such as ease of production, regeneration and stability [@pone.0106805-SuhSH1]. From a diagnostic perspective, they have been used for both target capture and detection purposes [@pone.0106805-SuhSH1]. In this study, we describe the selection and characterization of ssDNA aptamers with binding affinity to HuNoV. The utility of these aptamers was demonstrated in their use for capture and detection of HuNoV in outbreak-derived stool samples and a representative food matrix. Materials and Methods {#s2} ===================== Viruses and Virus-Like Particles (VLPs) {#s2a} --------------------------------------- ### Viruses {#s2a1} Snow Mountain virus (SMV), the prototype genogroup II, genotype 2 (GII.2) HuNoV and the target for aptamer selection, and Norwalk (NV) the prototype genogroup I, genotype 1 (GI.1) strain were obtained as stool specimens originating from a human challenge study (courtesy of C. L. Moe, Emory University, Atlanta, GA). The SMV human challenge study was conducted at the University of North Carolina at Chapel Hill (UNC-CH) and was approved by the UNC-CH Biomedical IRB. The NV study was conducted at Emory University and approved by the Emory University IRB, Biomedical Committee. Both studies had written consent, with each participant signing an informed consent witnessed by trained study staff. The informed consent documents were approved by the IRBs. The Emory University group also supplied pre-challenge stool samples confirmed (by RT-qPCR) as negative for HuNoV. These were used for counter selection and as negative controls in some studies. Additional fecal specimens associated with previously confirmed HuNoV outbreaks (representing strains GI.6, GII.1, GII.3, GII.4, GII.7 and an untypable GII) were also used in detection assays (courtesy of S.R. Greene, North Carolina Department of Health and Human Services, Raleigh, NC). All stool samples were suspended 20% in phosphate buffered saline (PBS). In some cases, these suspensions were used without further purification, designated as crude suspensions. In other cases, the suspensions were partially purified using chloroform extraction [@pone.0106805-Shin1]. Hepatitis A virus HM175 (cell culture adapted) and poliovirus 1, routinely used in our laboratory, were partially purified (chloroform extracted) from cell culture lysates and used in exclusivity studies. All virus suspensions were stored at −80°C until use. ### Virus-Like Particles (VLPs) {#s2a2} Self-assembled non-infectious Virus-Like Particles (VLPs), produced using the recombinant expression of HuNoV capsid proteins were used as purified candidate proteins for the characterization of aptamer binding affinity. The VLP panel consisted of representatives of genogroup I \[GI.1 (Norwalk virus), GI.4, GI.6, GI.7 and GI.8)\] and genogroup II \[GII.1, GII.2 (SMV), GII.3, GII.4 (Houston and Grimsby virus), GII.6, GII.7, GII.12 and GII.17\] HuNoV (kindly provided by R. Atmar, Baylor College of Medicine, Houston, TX). VLPs were stored at 4°C until use. Selection of Aptamers using SELEX (Systematic Evolution of Ligands by EXponential Enrichment) {#s2b} --------------------------------------------------------------------------------------------- ### Preparation of the DNA Library {#s2b1} All primer and probe sequences used in this study were obtained from Integrated DNA Technologies (Coralville, IA). The 81-base combinatorial DNA library consisted of 40-mer random regions flanked by forward and reverse constant regions \[5′-AGTATACGTATTACCTGCAGC-(N)~40~-CGATATCTCGGAGATCTTGC-3′\]. Preparation of the dsDNA library for SELEX was done in accordance with the method of Dwivedi et al. (2010) [@pone.0106805-Dwivedi1]. Specifically, the library was amplified using a fluorescein (FAM)-labeled forward constant region primer \[5′-56-FAM/-AGTATACGTATTACCTGCAGC-3′\] and a biotinylated reverse constant region primer \[5′-/5Bios/-GCAAGATCTCCGAGATATCG-3′\]. Briefly, a 50 µl reaction master mix containing 5 µl of aptamer library (10 µM), 1x GoTaq buffer (Promega Corp., Madison, WI), 0.2 mM dNTPmix (Applied Biosystems, Foster City, CA), 5 U Go Taq DNA polymerase (Promega), and 500 nM of each primer were amplified as follows: 95°C for 5 min, followed by 30 cycles of 95°C (1 min), 55° (1 min), 72°C (1 min) and a final extension at 72°C for 10 min, using a DNA Engine (PTC-200) Peltier Thermal Cycler-200 (MJ Research/Bio-Rad Laboratories, Hercules, CA). The labeled dsDNA library was conjugated to Streptavidin MagneSphere Paramagnetic particles (SA-PMPs) (Promega) according to manufacturer instructions and harvested with an MPC-M magnetic particle concentrator (Invitrogen Dynal AS, Oslo, Norway). The library-magnetic bead conjugate was washed three times in 0.1X SSC. The FAM-labeled ssDNA moieties were then separated from the immobilized biotinylated strands by alkaline denaturation using 50 µl of 0.15 M NaOH for 7 min at room temperature (RT), followed by magnetic separation of the beads. The supernatant obtained was mixed with 1 ml of nuclease free water followed by filtration to remove residual NaOH using a YM-30 filter device (Millipore, Billerica, MA). It was then purified by ethanol precipitation with resuspension in 50 µl of nuclease free water. The presence of FAM-labeled DNA was confirmed using a fluorescent plate reader (Tecan Safire, Tecan Group Ltd., Männedorf, Switzerland) and its concentration determined using a NanoPhotometer Pearl (Implen GMbH, Munich, Germany). ### Preparation of the target for SELEX {#s2b2} The target for SELEX was produced by immobilizing SMV to antibody-bead conjugates. Briefly, M-280 Tosylactivated Dynabeads (Invitrogen Dynal) were cross-linked to mouse monoclonal antibodies against SMV (Abcam, Cambridge, MA) as per manufacturer instructions. One hundred µl of the partially purified SMV stock (consisting of approximately 10^5^--10^6^ RT-qPCR amplifiable units/ml) was mixed with 10 µl of the antibody-bead conjugate suspended in 500 µl of binding buffer \[consisting of phosphate buffered saline (pH 7.1) supplemented with 100 mg/L CaCl~2~, 100 mg/L MgCl~2~ and 0.05% Tween 20\] followed by RT incubation for 2 h. After washing thrice with PBS supplemented with 0.05% Tween 20 (PBST), the conjugate was resuspended in 20 µl of PBS and used for SELEX. ### SELEX {#s2b3} An aliquot of 300--500 pmoles of FAM-ssDNA pool suspended in 500 µl of binding buffer was added to 10 µl immobilized SMV target followed by gentle rotation for 45 min at RT. Aptamer-bound SMV was recovered by magnetic capture, washed thrice in PBST, and the aptamers eluted from the bead-bound virus particles with 200 µl of nuclease free water followed by heating at 90°C for 5 min. The resulting aptamer pool was purified by ethanol precipitation, re-amplified using the labeled constant region primers, and the FAM-ssDNA aptamers were recovered as described above. This constituted one selection round and a total of nine such iterations of the selection process were performed. To avoid non-specific amplification of the recovered aptamer pool, single stranded binding protein (Promega) at a concentration of 0.1 µg/µl was added to the PCR reactions and the pool was amplified using the appropriate annealing temperatures (obtained by running a temperature gradient within a range of 55°C to 65°C on the recovered aptamer pool). ### Counter-SELEX {#s2b4} Two sequential counter-SELEX rounds were done after each round of SELEX round to impart specificity to the aptamer candidates against (1) the components of a 20% HuNoV-negative human stool suspension; and (2) the bead-antibody complex (without SMV). In counter-SELEX, exposure of the aptamer pools to the negative stool specimens or bead-antibody complex was done as described above, but in this case, the aptamers bound to the complex were discarded, while the unbound aptamer pool (supernatant) was recovered. This was purified by ethanol precipitation, re-amplified by PCR, and used in another round of SELEX. ### Identification of Aptamer Sequences {#s2b5} After the 4^th^, 7^th^, and 9^th^ rounds of sequential SELEX and counter-SELEX, final PCR product was cleaned using the QIAquick PCR purification kit (Qiagen, Valencia, CA) and cloned into TOPO vector (Invitrogen, Carlsbad, CA) according to manufacturer instructions. The DNA insertion in each clone was sequenced by Genewiz (South Plainfield, NJ). The unique aptamers were identified by sequence analysis. Characterization of Aptamer Candidates {#s2c} -------------------------------------- ### Dissociation Constant and Structural Analysis of Aptamers {#s2c1} Consistent with the approach of others [@pone.0106805-Friguet1] [@pone.0106805-Fuch1], an ELISA-like assay (Enzyme-Linked Aptamer Sorbant Assay, or ELASA, described below) was used to determine equilibrium dissociation constants (K~d~) for the candidate aptamers. This was done using SMV VLPs (3 µg/ml) and different concentrations (10 nM, 100 nM, 500 nM, 1 µM, 2 µM) of each aptamer. To estimate K~d~, plots of the ratio between absorbance of Test samples/absorbance of Negative controls (T/N ratios, Y axis) as a function of the aptamer concentration (X axis) were generated using a non-interacting binding sites model in Sigma Plot (Jandel, San Rafael, CA). Common sequence motifs were identified using the online MEME server (<http://meme.sdsc.edu>). Structural folding analyses of the selected aptamers were done using the DNA Mfold online server (<http://mfold.rna.albany.edu/?q=mfold/DNA-Folding-Form>) [@pone.0106805-Zuker1]. To identify potential binding epitopes, the major capsid protein (VP1) sequences of each of the VLPs tested were aligned using Clustal W alignment in the Molecular Evolutionary Genetics Analysis program (MEGA 6.0) (<http://www.megasoftware.net/>). Aligned sequences were matched with the aptamers that showed the strongest (+++) binding affinity to each VLP. ### Binding Affinity Studies Using Enzyme-Linked Aptamer Sorbent Assay (ELASA) {#s2c2} Binding affinity studies were performed with candidate aptamers and VLPs using a protocol adapted from a previously reported ELISA-based antigen detection assay [@pone.0106805-Rogers1]. In this case, the antibody was replaced with an aptamer and the resulting procedure termed Enzyme-Linked Aptamer Sorbent Assay (ELASA). For this assay, the selected aptamers were labeled with a 5′ biotin moiety. One hundred µl of pure VLP suspension (3 µg/ml) was placed in each well of a covered, flat-bottomed polystyrene 96-well plate (Costar 3591, Fisher, Pittsburg, PA) and incubated overnight at 4°C. After coating the wells with VLPs, the plates were blocked with 200 µl of 5% skim milk in PBST containing non-related DNA sequences (i.e., *L. monocytogenes* primers hlyQF/R and L23SQF/R) [@pone.0106805-RodrguezLzaro1], followed by overnight incubation at 4°C. After washing three times with PBST, 100 µl of biotinylated aptamer (1 µM) was added to each well and the plate was incubated for an hour at RT with gentle mixing. After removing excess aptamers, the plates were washed three times with PBST. One-hundred µl of ELISA-grade streptavidin-horse radish peroxidase conjugate (1 mg/ml, 1∶5000, Invitrogen) was added to the plate and incubated at RT for 15 min. The unbound conjugate was removed and the plate was washed five times with PBST before applying 100 µl of 3,3′5,5′tetramethylbenzidine (TMB) peroxidase substrate (KPL, Gaithersburg, MD) to each well. The plate was incubated for 5 min at RT after which 100 µl of 1 M phosphoric acid was added to stop the reaction. Absorbance at 450 nm was recorded using a microplate reader (Tecan Infiniti M200pro). Negative controls consisted of no VLPs. As per convention [@pone.0106805-Ebel1], binding affinity was interpreted based on the ratio between the absorbance readings for the test samples (to which labeled aptamer had been added) versus those for the negative control (PBS alone), (T/N ratios). Ratios ≤2.0 were considered negative [@pone.0106805-Ebel1]. Ratios between 2.0 and 5.0; \>5.0 and 10.0; and \>10.0 were interpreted as low, medium or strong binding, respectively. Ratios obtained for the negative control were in the range of 0.1--0.3. To evaluate binding inclusivity, ELASA was performed using 1 µM (100 µl) of each candidate aptamer as applied to a panel of virus-like particles (VLPs) corresponding to genogroup I \[GI.1 (Norwalk virus), GI.4, GI.6, GI.7 and GI.8)\] and genogroup II \[GII.1, GII.2 (SMV), GII.3, GII.4 (Houston and Grimsby), GII.6, GII.7, GII.12 and GII.17\] HuNoV. Exclusivity analyses were done by ELASA using hepatitis A virus (HAV) and poliovirus. A scrambled random sequence for aptamer 25 (25 S) was generated by the Random Nucleic Acid Sequence Generator server on line <http://molbiol.ru/eng/scripts/01_16.html> and used to evaluate background binding in the ELASA. In all cases, PBS and SMV-VLPs were included as negative and positive controls, respectively. ### Binding affinity Studies Using Enzyme-Linked Immunosorbent Assay (ELISA) {#s2c3} The original ELISA protocol adapted for ELASA was used in these experiments [@pone.0106805-Rogers1]. Briefly, SMV-VLPs were used to coat 96-well polystyrene plates as described above. The wells were blocked with 200 µl of 5% skim milk in PBST and incubated for 2 h at RT. After washing 3 times with PBST, 100 µl of diluted (1∶5000) GII.2 antibody (Abcam, Cambridge, MA) was added and incubated for 1 h at 37°C. After washing 3 times with PBST, 100 µl of the secondary antibody (anti-mouse conjugated with horseradish peroxidase; Abcam, Cambridge, MA) at a 1∶5000 dilution was added to each well and incubated for 1 h at 37°C. The plate was washed 3 times with PBST with subsequent development using TMB peroxidase substrate and absorbance reading was recorded as described above for ELASA. Application of Aptamers for Detection of HuNoV in Clinical and Food Samples {#s2d} --------------------------------------------------------------------------- ### Detection of HuNoV in Outbreak-Derived Stool Samples {#s2d1} The ELASA assay was used to assess the performance of select aptamer candidates for detection of HuNoV in outbreak-derived fecal suspensions. Specimens evaluated included those representing strains GI.1 (Norwalk), GI.6, GII.1, GII.2 (SMV), GII.3, GII.4, GII.7 and an untypable GII. Briefly, 100 µl of ten-fold serially diluted partially purified fecal suspensions were used to coat plates followed by incubation overnight at 4°C. After three washes with PBST, the ELASA assay was done as described above. PBS alone and fecal suspensions derived from uninfected individuals were used as negative controls. Due to limited availability of SMV VLPs, we used GII.4 (Houston) VLPs, which were also highly reactive to aptamer 25, as the positive control. ### Detection of HuNoV in Artificially Contaminated Lettuce Samples Using a Combined Pre-Concentration-Aptamer Magnetic Capture (AMC)-RT-qPCR Assay {#s2d2} For virus inoculation and pre-concentration, 3 g of lettuce (approximately 3×3 cm square) was disinfected by UV light and inoculated with 200 µl of serially diluted crude GII.4 virus stock suspension (due to limited availability of SMV) at inoculum concentrations ranging from 1--5 log~10~ RNA copies per lettuce sample. The inoculum was allowed to dry for 30 min. Virus pre-concentration was done using a previously reported elution-concentration method [@pone.0106805-Leggitt1]. Briefly, the samples were mixed with 25 ml of 0.5 M glycine-0.14 M NaCl buffer (pH 9.0), placed in sterile Whirl-Pak-filter bags (Nasco, Fort Atkinson, WI) and stomached (Stomacher 400 Circulator, Seward, Norfolk, UK) for one min. The recovered filtrate (containing the eluted viruses) was adjusted to 0.9 M NaCl and supplemented with 1% bovine serum albumin (Sigma Aldrich, St. Louis, MO), after which 12% polyethylene glycol (PEG) MW 8,000 (Sigma Aldrich) was added. After incubation for 2 h at 4°C, samples were centrifuged at 10,000×*g* for 20 min at 8°C and the recovered pellet was resuspended in 1 ml of PBST. For aptamer magnetic capture (AMC), the resuspended pellet was incubated with 1 µM of biotinylated aptamer for an hour at RT with rotation. This was followed by the addition of 10 µl of Streptavidin (SA)-C1 magnetic beads (from 10 mg/ml stock solution) (Invitrogen Dynal AS, Oslo, Norway) previously blocked overnight with 5% skim milk in PBST followed by incubation for 25 min at RT. The bead-aptamer-virus conjugates were recovered using the magnetic particle concentrator. The conjugates were washed twice with PBST, suspended in 100 µl PBS, and the RNA extracted using the easyMAG automated system (bioMerieux SA, Marcy l\'Etoile, France) according to manufacturer\'s instructions. The viral RNA was eluted in 40 µl of proprietary elution buffer and stored at −80°C until used for amplification. RT-qPCR was carried out using the Superscript III Platinum One-Step RT-qPCR system (Invitrogen) according to manufacturer instructions. Specifically, 25 µl RT-qPCR reactions consisted of 12.5 µl of 2X reaction mix, 5.5 µl DNase-RNase free water, 200 nM of each primer \[JJV2F (5′-CAAGAGTCAATGTTTAGGTGGATGAG-3′) and COG2R (5′-TCGACGCCATCTTCATTCACA-3′)\], and probe RING2-TP \[5′-56-FAM TGGGAGGGCGATCGCAATCT-1BHQ-3′\] [@pone.0106805-Jothikumar1] [@pone.0106805-Kageyama1], 0.5 µl of the enzyme mix (SuperScript III RT/Platinum Taq Mix) and 5 µl of the RNA template. Amplification was done under the following conditions: 50°C for 15 min, 95°C for 2 min followed by 45 cycles of 95°C for 15 s, 54°C for 30 s, and 72°C for 30 s in a SmartCycler (Cepheid, Sunnyvale, CA). The RNA copy number was extrapolated from a standard curve based on Ct values obtained by RT-qPCR amplification of serially diluted synthetic RNA as previously reported [@pone.0106805-Escudero1]. Capture efficiency, expressed as a percentage, was estimated from the standard curve and calculated as the ratio of the extrapolated RNA copies (after capture and detection by RT-qPCR) to the total input RNA copies per sample, multiplied by 100 [@pone.0106805-Joshi1]. Negative controls consisted of capture by blocked beads in the absence of aptamer. Statistical analysis {#s2e} -------------------- Data were expressed as mean ± standard deviation of three replicates of each experiment. The data were analyzed by one-way analysis of variance (ANOVA) with the Tukey\'s multiple comparison test using GraphPad Prism ver. 5.0d (San Diego, CA). Values of *p*\<0.05 were considered statistically significant. Results {#s3} ======= ssDNA Aptamer Selection {#s3a} ----------------------- After 4, 7, and 9 rounds of sequential SELEX and counter-SELEX, 34 unique aptamer candidates were identified from a total of 80 clones sequenced. All sequences are provided in [Table 1](#pone-0106805-t001){ref-type="table"}. Candidates designated as 19, 21, 25 and 26 were selected for further analysis as they were the most abundant in the identified aptamer pool (identified between 5--10 times) and showed strong preliminary binding affinity for SMV, genogroup I.1 (Norwalk), and genogroup II.4 (Houston) VLPs using the ELASA assay. 10.1371/journal.pone.0106805.t001 ###### Aptamer sequences obtained from 4^th^, 7^th^ and 9^th^ round of SELEX for SMV. {#pone-0106805-t001-1} ROUND RANDOM REGION SEQUENCE \# OF REPEATS APTAMER SELECTED ------- ---------------------------------------------- --------------- ------------------ 4 TGTTGGATTTTACGAAAAACGTGCTTACTTCATAGCGGCC 1 4 GGTTGGGTAAGGGGGTCTGGTCAGGTAGGGCGGGGGGGGG 1 4 TCGTAAACCCCTTATCCGTGAACCTTCAGCGGTAGACGCT 2 4 CTCCCTCCAGCCTGCCTATTTTGCTTGGTTACGCATCTGT 1 7 CCAGCGAAGGAAAGTCTTGGTTGGTCTAGTTTTTCGTGTG 2 7 CTACGTGTGCGTTCCGATTGTTTAAATTGCTCAATGTATG 2 7 CACACCACCTGAATTCCAGCACACTGGCGGCCGTTACTAG 2 9 CACTCGACCTTCAGGGCGGCTTCTCAGCGTGTAGTGGTGA 1 9 CTCGACTGATAGACCTAGCGTCAATCCTCATTGTTCGCTG 1 9 CCAGTATTAGAGTCCTACTTTACACCGCTCTTGGCATCGT 1 9 CACATGATAAGGTCGCGTGACTGTGAGTTAGTTGTTACAC 2 9 TCGGCATAGGTCAAGTCGCTTCATTTGGATTAAGTTGAGG 1 9 CACATACCAAAGTATTGGTCGCTAACTTTCGCCCAATTGA 1 9 CTACGAGGTGGTTATAAGAGAACTTATCCGTGTTGGTTGC 1 9 TGGTAGTGGGATATAGTTTTTCCAAGCGTACCCAGTTCTG 2 9 CTATCAGCCATGAATTGCATTACCTTTGTTCTCCCCTTGC 1 9 CCCCTCGGAAGATAGATTTTGCGAGAGTCTTGGGTTGAGG 1 9 CCAGATAGCAGCACCTAATCTTATCCCTTTTATTTTTGGT 2 9 TCGGGGGGAGGAGGGGGAATGGGAAGAAGGAGGTCGAGGG 1 9 TGGATTACACGGCTAACTTCCCTGGTTCTTTTCTTTGATG 2 9 TGGACGTTATTTGCACTCGTCGAACCCTATCATGCCTCCT 1 9 CCTCATGCACAAAGGCTTATTACGGTCTAATTCTTTATAA 1 9 TCGACATTATGTTTGACATCGATTGTTAATGTTTCTTTGC 2 9 CCCCTACACAGTAAAATTCTTTAACACCTAGATCTTCGAC 2 7, 9 **CACCAGTGTGTTGAGGTTTGAGCACACTGATAGAGTGTCA** 9 SMV-19 9 TGAGCCTCCGTTTTAGTGATCAGAAGGGATGTGTGGCGTA 1 9 **CCATGTTTTGTAGGTGTAATAGGTCATGTTAGGGTTTCTG** 9 SMV-21 9 CGAGGGATACATGCTGACTATGGAATTATTTGAATTCCCA 4 9 CTACAGGAGTTCATCTGGGAGAGTGTAAAGGATGAGGTGG 2 7, 9 **CATCTGTGTGAAGACTATATGGCGCTCACATATTTCTTTC** 10 SMV-25 9 **TGACCGAGTGTCTGGTCATTTTCGATGTCTGTTGTTAGGC** 7 SMV-26 9 CCCTCCTTATCTCTGCTAATGGTTGATCCGTGTCCCGTAC 1 9 CCCTGTTATCCTTATCCAACGAGCTTAATGTAACTTGGAC 2 9 TGGGGGAGTGGTAGGTGTGCTGTGAAGGGGAGGGTTGGGG 1 Bolded sequences were used in characterization studies and applications. Characterization of ssDNA Aptamer Candidates {#s3b} -------------------------------------------- The structural folding for all four aptamers demonstrated a dominant loop and two protruding hairpins ([Figure 1](#pone-0106805-g001){ref-type="fig"}). Three motifs were observed when comparing the random sequence regions of the four aptamers. Motif 1 was found in aptamers 19, 21, and 26. Motif 2 was found in all four aptamers and was in most cases, involved in stem-loop formation. Motif 3, immediately downstream of motif 2, was found in aptamers 19 and 21 and was also involved in stem-loop structures. Equilibrium dissociation constant (K~d~) values approximated for the aptamers were 191 nM for aptamer 19; 101 nM for 21; 232 nM for 25; and 281 nM for 26. [Figure 2](#pone-0106805-g002){ref-type="fig"} shows the corresponding K~d~ curves generated using the one site binding model. The regression coefficients (R^2^) associated with this model ranged from 0.95 to 0.99. {#pone-0106805-g001} {#pone-0106805-g002} Signal intensity ratios (T/N) in ELASA as evaluated for aptamers 19, 21, 25 and 26 using a panel of VLPs ranged from a low of 1.3 to a high of 18.1 ([Table 2](#pone-0106805-t002){ref-type="table"}). Aptamer 21 demonstrated medium to high binding affinity with VLPs corresponding to GI.7, GII.1, GII.2, GII.3, GII.4 (both VLPs), GII.7, GII.12, and GII.17 and T/N ratios ranging from 6.4 to 18.1. Aptamer 25 reacted positively with VLPs corresponding to GI.4, GI.8, GII.1, GII.2, GII.3, GII.4 (both VLPs), GII.6, and GII.7 (T/N ratios ranged from 5.4 to 12.4). Aptamers 19 and 26 were less broadly reactive. In general, T/N ratios were higher for GII VLPs than for GI VLPs. Although all four aptamers showed variable binding affinity to different genotypes in genogroups I and II, binding affinity was highest for the GII.2 VLP, which is represented by SMV, the target used for the aptamer selection. Binding affinity was also quite high for GII.4 Houston VLPs. 10.1371/journal.pone.0106805.t002 ###### Binding affinity of selected aptamers against a broad panel of HuNoV VLPs. {#pone-0106805-t002-2} VLPs Aptamers[1](#nt102){ref-type="table-fn"} ----------------- ------------------------------------------ ------- ---------- ------- ---------- ------- --------- ------- GI.1 (Norwalk) 5.4±0.9 (+) 2.4±0.1 (+/−) 3.0±0.4 (+/−) 2.6±0.5 (+/−) GI.4 1.7±0.3 (−) 4.3±0.2 (+/−) 5.6±0.2 (+) 1.1±0.2 (−) GI.6 3.7±1.3 (+/−) 3.8±0.1 (+/−) 2.4±0.1 (+/−) 1.5±0.4 (−) GI.7 10.3±0.6 (++) 8.1±0.1 (+) 3.5±0.4 (+/−) 1.8±0.8 (−) GI.8 1.7±0.2 (−) 4.6±0.3 (+/−) 6.2±0.1 (+) 1.2±0.1 (−) GII.1 7.1±0.3 (+) 8.6±0.3 (+) 9.4±0.1 (+) 2.1±0.3 (+/−) GII.2 (SMV) 12.9±5.1 (++) 18.1±3.2 (++) 12.4±1.1 (++) 4.1±0.8 (+/−) GII.3 1.7±0.9 (−) 11.8±2.7 (++) 5.4±0.1 (+) 2.4±0.5 (+/−) GII.4 (Grimsby) 9.6±4.8 (+) 6.4±1.9 (+) 10.4±0.8 (++) 3.2±0.3 (+/−) GII.4 (Houston) 12.5±4.2 (++) 11.3±1.2 (++) 11.0±1.1 (++) 2.8±0.4 (+/−) GII.6 1.8±1.0 (−) 3.0±0.2 (+/−) 7.3±0.7 (+) 2.8±0.3 (+/−) GII.7 4.0±1.9 (+/−) 13.5±1.4 (++) 9.1±2.1 (+) 2.8±0.8 (+/−) GII.12 1.5±0.4 (−) 6.6±0.1 (+) 1.9±0.1 (−) 2.4±0.5 (+/−) GII.17 3.0±1.6 (+/−) 12.2±0.3 (++) 2.0±0.1 (+/−) 1.3±0.1 (−) Values indicate the ratio of absorbance readings for the test sample (T) versus negative control (N) using ELASA. Per convention [@pone.0106805-Ebel1], results less than 2.0 are considered negative (−). Low (+/−), medium (+) or strong (++) binding were interpreted for ratios between 2 and 5; \>5 and 10; and \>10, respectively. All experiments were done in triplicate. Relative to exclusivity analysis, binding affinity (T/N) of the four aptamers to the non-target virus poliovirus was in the range of 3.2--4.4, and for HAV, from 2.8--3.4. These values were statistically significantly lower (*p*\<0.05) as compared to the positive control \[GII.2 (SMV) VLPs\] ([Figure 3](#pone-0106805-g003){ref-type="fig"}). {#pone-0106805-g003} Aptamer 25 and a scrambled derivative (aptamer 25 S) were used in ELASA to demonstrate the potential for non-specific binding. The T/N ratios obtained by ELASA using GII.2 SMV VLPs and biotinylated aptamer 25 S were statistically significantly lower (*p*\<0.05) when compared to those obtained using biotinylated aptamer 25. In competition studies, the binding of biotinylated aptamer 25 was not negatively impacted by the presence (in excess) of aptamer 25 S (*p*\<0.05), demonstrating that non-specific binding of an irrelevant scrambled aptamer did not negatively impact the binding specificity of aptamer 25 to SMV VLPs ([Figure S1](#pone.0106805.s001){ref-type="supplementary-material"}). For comparison purposes, ELASA and ELISA assays were performed using biotinylated aptamer 25 or a GII.2 commercial antibody, respectively, as applied to SMV VLPs suspended in partially purified HuNoV-negative stool. Based on comparison of T/N ratios, both assays performed equivalently within the aptamer/antibody concentration ranges of 1 to 5 µg/ml. ([Figure 4](#pone-0106805-g004){ref-type="fig"}). {#pone-0106805-g004} Application of Aptamers for Capture and Detection of HuNoV in Human Stool and Lettuce Samples {#s3c} --------------------------------------------------------------------------------------------- ELASA assays using aptamer 25 were performed on serially diluted partially purified outbreak-derived stool specimens. The T/N ratios corresponding to GI.1 (Norwalk), GII.1, GII.2 (SMV), GII.3, GII.4, and GII untypable outbreak specimens were all statistically significantly higher (*p*\<0.05) when compared to the ratios obtained for either negative control samples (i.e., virus-free stool suspensions and no aptamer controls) ([Figure 5](#pone-0106805-g005){ref-type="fig"}). T/N ratios were higher for GII.2 (SMV), GII.1 and GII.4 outbreak specimens relative to GI.1 (Norwalk), GII.3, and GII untypable. Positive ELASA signals were not obtained for outbreak specimens corresponding to GI.6 and GII.7 HuNoV strains. {#pone-0106805-g005} When virus on artificially contaminated lettuce samples (inoculated with GII.4 at levels ranging from 1--5 log~10~ RNA copies per sample) were pre-concentrated, used in AMC with aptamer 25, and detected after RNA extraction using RT-qPCR, a detection limit of about 1 log~10~ RNA copies per 3 g lettuce was obtained ([Figure 6](#pone-0106805-g006){ref-type="fig"}). Over this inoculum range, the capture efficiency of the combined pre-concentration-AMC-RT-qPCR assay ranged from 2.5--36%. Capture efficiency was significantly higher (p\<0.05) than the negative controls which consisted of blocked beads in the absence of aptamer. Capture efficiency increased with decreasing virus concentration. {#pone-0106805-g006} Discussion {#s4} ========== Nucleic acid aptamers are emerging molecules increasingly being used as diagnostic tools for detection of pathogens, including those transmitted by foodborne routes [@pone.0106805-SuhSH1]. Two recent papers report on the development of aptamers directed against noroviruses [@pone.0106805-Giamberardino1] [@pone.0106805-Beier1]. Specifically, Giamberardino et al. (2013) [@pone.0106805-Giamberardino1] produced a DNA aptamer against the cultivable surrogate murine norovirus (MNV) that displayed cross-reactivity to a GII.3 HuNoV VLP. This aptamer was used as a component of an electrochemical sensor that was able to detect as few as 180 MNV particles. Common to our approach, these investigators directed the SELEX process against the whole virus and used a similar number of selection rounds. Beier et al. (2014) [@pone.0106805-Beier1] used a SELEX process directed against the capsid protein (VP1) of a GII.4 HuNoV to produce an aptamer for which they simulated the binding characteristics using predictive modeling. They did not evaluate the performance of the aptamer for virus capture or detection in clinical or food samples. To our knowledge, the data presented here is the first time in which an investigative team has identified several different ssDNA aptamers that were extensively characterized for binding affinity using a wide range of HuNoV strains, both as VLPs and as intact viruses. This study also provides the first demonstration of the utility of aptamers in candidate capture-detection platforms, including those relevant to the fecal matrix and foods. Methodologically, it is novel because we employed extensive counter-SELEX to ameliorate potential non-specific binding. These features have led to some unique results. Despite the targeting of a single virus (SMV) in SELEX, two of the identified aptamers (candidates 21 and 25) showed binding affinity to a panel of HuNoV VLPs, demonstrating the efficacy of SELEX in identifying aptamers with binding inclusivity to HuNoV strains. The binding inclusivity of aptamer 25 was further confirmed by the detection of HuNoV strains in outbreak-derived fecal samples by ELASA. The performance of aptamer 25 with outbreak stool specimens containing GII.1, GII.2, GII.3, and GII.4 HuNoV is consistent with the VLP binding data, all of which gave positive signals with ELASA. Likewise, poor performance of aptamer 25 with an outbreak specimen corresponding to GI.6 is consistent with the low degree of binding to that VLP. Binding of stool specimens containing GI.1 virus was inconsistent with VLP data, as was lack of binding to GII.7 specimens. Such inconsistency between VLP binding assays and application to outbreak-derived stool specimens could be a function of residual matrix associated aptamer binding (or interference with aptamer binding) or potential differences between the behavior of VLPs and native virus, which has been discussed amongst experts in the field (R. Atmar, personal communication). Nonetheless, in the absence of a means by which to cultivate HuNoV and/or the availability of broad strain panels, VLPs remain a valuable reagent in these types of studies. In general, the ELASA T/N ratios were higher for GII strains than for GI strains. This is not unexpected as major capsid protein (VP1) amino acid sequences for GII strains differ from those for GI strains by over 61.4% [@pone.0106805-Zheng1], and our SELEX target (SMV) was a GII strain. Nonetheless, different VLP binding patterns for the four characterized aptamers suggest that they may bind to slightly different regions of the viral capsid. It could also be argued that the more broadly reactive aptamers bind to more highly conserved areas of the virus, such as the shell or P1 domains [@pone.0106805-Lindesmith1]. Of the three common motifs identified in this study, motif 2 was found in all four aptamers and this may imply a conserved interaction site with HuNoV. Secondary structures have been considered as possible putative binding epitopes on aptamers [@pone.0106805-Kato1]. In an effort better predict aptamer binding epitopes, cluster analysis of the major capsid protein (VP1) sequences of each of the VLPs was undertaken, with aligned sequences matched with aptamers showing strong binding affinity ([Figure S2](#pone.0106805.s002){ref-type="supplementary-material"}). Aptamers 21 and 25 bound strongly to a group of closely related GII VLPs. Further amino acid sequence analysis demonstrated that SMV residue 406 (part of a fairly conserved area of the VPI that consists of residues 405--408) was more highly conserved (a V or K) among VLPs with stronger binding to aptamers 19 and 25. Aptamer 21 appeared to have multiple candidate epitope residues, including residue 65, for which the change to a Q from other residues (N, G, E, or A) seemed to eliminate binding. This residue is part of a sequence of amino acids that is quite conserved (NFVQAPQGEFT), which could help explain the relatively broad reactivity of aptamer 21. Additionally, the presence of an amino acid different from V or A at residue 394 was correlated with lower ELASA T/N ratios for aptamer 21. Finally, a sequence starting at residue 428 \[FPGEQ(I/L)(I/L)\] was also conserved among the six VLPs to which aptamer 21 bound most strongly. Further characterization of the putative aptamer binding sites will require mutagenic, structural, bioinformatics, and/or other analyses beyond the scope of this current work. Antibodies, the most frequently used ligands for HuNoV capture and detection [@pone.0106805-Yao1] [@pone.0106805-Park1] [@pone.0106805-Lee1], tend to lack broad reactivity, meaning that subsequent assays developed with these antibodies lack analytical sensitivity [@pone.0106805-Costantini1]. Having an alternative ligand type showing broad reactivity to multiple HuNoV strains provides another tool upon which capture and detection assays may be based. Theoretically, aptamers having different specificities could be used as a polyvalent cocktail to impart a high level of inclusivity for virus capture and detection. They could also be used in combination with other ligands such as antibodies or peptides [@pone.0106805-Rogers2]. These sorts of applications have recently been proposed by others [@pone.0106805-Kim1] and are the subject of future investigations. Using the ELASA assay, we were able to approximate the SMV aptamer K~d~ values to be in the 100--200 nM range. This is similar to most commercial antibodies which have K~d~ values in the low µM to nM range ([www.abcam.com](http://www.abcam.com)). In binding studies specifically with Norwalk virus VLPs, monoclonal antibodies have been shown to have K~d~ values in the low nM range [@pone.0106805-Chen1] and K~d~ values for enteric virus binding protein as applied to HuNoV VLPs was similarly in the range of 210--240 nM [@pone.0106805-Imai1]. Although we did not perform K~d~ studies on the commercial antibody used in comparative studies with aptamer 25, the antibody and aptamer did perform equivalently in ELISA and ELASA, respectively. Some non-specific binding using a scrambled sequence of aptamer 25 was observed, and this is consistent with others [@pone.0106805-Kaur1]. However, we note that few studies use scrambled sequences to evaluate non-specific binding of nucleic acid aptamers. The degree of non-specific binding observed here may be a function of the concentration of the aptamer used in the assay, as at 1 µM we approached saturation of the signal. It could also be a feature specific to aptamer 25, and could therefore differ were the other aptamers to be screened in this manner. Future studies using a range of aptamer concentrations and screening a wider variety of VLPs and aptamers would be a logical next step. Aptamer 25 performed quite well when applied to a virus concentrate derived from an artificially contaminated model food product using a combined virus pre-concentration followed by AMC-qPCR assay format. Use of aptamers for pre-concentration of microbes has been reported recently by others [@pone.0106805-Ozalp1]. Comparatively speaking, the detection limit and capture efficiency of this method, at 10 RNA copies and 36%, respectively, were comparable to those for HuNoV immunomagnetic separation-RT-qPCR assays applied to artificially contaminated fresh produce items [@pone.0106805-Park1] [@pone.0106805-Lee1] and clinical specimens [@pone.0106805-Yao1]. Our detection limits were also similar if not better than those for other capture approaches that use more non-specific ligands such as histo-blood group antigens [@pone.0106805-Tian1] [@pone.0106805-Morton1] [@pone.0106805-Pan1] and porcine gastric mucin [@pone.0106805-Tian2]. In conclusion, we report the identification of ssDNA aptamers having binding affinity and inclusivity to various HuNoV strains. We also confirmed the performance of these aptamers in capture and detection assays for HuNoV in outbreak associated human stool suspensions and in an artificially contaminated food matrix. Consistent with recent literature [@pone.0106805-Ozalp1], nucleic acid aptamers are promising ligands to facilitate pathogen capture and detection in complex sample matrices. As demonstrated in this study, aptamers may be good alternatives to HuNoV antibodies due to their relatively high inclusivity, specificity, ease of synthesis, and lower cost. Their use in detection assays is obvious, but they may also have utility in other sorts of assays, such as those that seek to discriminate between infectious and non-infectious virus particles [@pone.0106805-Knight1]. Future work will continue to focus on these sorts of applications. Supporting Information {#s5} ====================== ###### **Scrambled aptamer analysis.** The binding of biotinylated aptamer 25 was compared to that of a labeled scrambled aptamer (25 S) in competitive and non-competitive ELASA. For the latter, combinations of biotinylated labeled aptamer 25 with unlabelded 25 S, and vice versa, were used. Labeled aptamers were added at a concentration of 1 µM; unlabeled aptamers were used in 4-fold excess. Experiments were done by triplicate. Different letters indicate statistically significant differences between treatment groups (*p*\<0.05). (TIFF) ###### Click here for additional data file. ###### **Maximum likelihood tree for VLPs.** Full-length protein sequences of the VP1 protein of each of the VLPs tested were aligned using Clustal W alignment in the Molecular Evolutionary Genetics Analysis program (MEGA 6.0) (<http://www.megasoftware.net/>). Aligned sequences were matched with the aptamers that showed the strongest (+++) binding affinity to each VLP. (TIFF) ###### Click here for additional data file. The authors thank Dr. C.L Moe (Emory University, Atlanta GA) for providing SMV, Norwalk, and the pre-challenge stool samples negative for HuNoV; Dr. S.R. Greene (North Carolina Department of Health and Human Services, Raleigh, NC) for providing the fecal specimens associated with HuNoV outbreaks; and Dr. R. Atmar (Baylor College of Medicine, Houston, TX) for providing the VLPs. We also thank Dr. Elizabeth Bradshaw for editorial assistance. [^1]: **Competing Interests:**The authors have the following interests: This study was funded in part by the Merieux Research Grants Program. A provisional United States patent application entitled "Aptamers with binding affinity to norovirus" (PATENT APPLICATION NUMBER 62/011,880) was filed through the Office of Technology Transfer at North Carolina State University (authors BEA, LAJ, MM). This does not alter the authors\' adherence to all the PLOS ONE policies on sharing data and materials, as detailed online in the guide for authors. [^2]: Conceived and designed the experiments: LAJ BEA SHS MM HPD. Performed the experiments: BEA SHS MM. Analyzed the data: LAJ BEA SHS MM HPD. Contributed reagents/materials/analysis tools: LAJ. Contributed to the writing of the manuscript: LAJ MM HPD BEA SHS. [^3]: Current address: Food Microbiology Division, Ministry of Food and Drug safety, Osong-eup, South Korea [^4]: Current address: bioMérieux, Inc., Hazelwood, Missouri, United States of America

Introduction ============ Post hysterectomy complications at the vault site such as a bleeding incident can be commonly observed at a short term postoperative period. Other delayed complications often occur as a hematoma, granuloma, keloid, incisional hernia, and or vascular formation at the vault \[[@B1][@B2]\]. Many of these complications may be accompanied with bleeding symptoms. However a sudden massive bleeding occurring after more than a year postoperatively is rare. Here we report a case of sudden vaginal vault bleeding in a post-hysterectomy state patient with no known history of endometriosis, hysterectomy undergone prior 13 months ago, proven to be a rare case of vault endometriosis. Hence, we would like to present how such a rare complication might be misdiagnosed and moreover increase the awareness of such complication of the vault in a post-hysterectomy patient. Case report =========== Early on October 2014, a 45-year-old woman was presented at our outpatient gynecology clinic with sudden lower pelvic discomfort and vaginal bleeding symptoms. The patient had a history of hysterectomy 13 months ago due to symptomatic multiple leiomyomas and adenomyosis. The previous surgery was conducted as a single-port approach laparoscopic-assisted vaginal hysterectomy in July 2013. Surgical findings showed an enlarged uterus of about a 14 weeks gestational age size. Both adnexa were grossly normal in appearance and the peritoneum was clear with no signs of endometriosis. The vaginal vault was sutured vaginally using a 1-0 vicryl. No complications were observed during the postoperative period and the patient was discharged as scheduled. The histology of the uterus was confirmed as adenomyosis with leiomyoma with a secretory phase endometrium. Follow up examinations at 3 and 6 months showed no complications and thus no additional follow-ups were required afterwards. However on October 2014, after more than a year from her last check up, the patient appeared at our clinic complaining of sudden pelvic discomfort and vaginal bleeding symptoms. Pelvic examinations showed no signs of active bleeding. Yet, a dark wine colored papule suggestive of a small hematoma or ulcerative lesion was observed upon the previous operative vaginal vault site ([Fig. 1.](#F1){ref-type="fig"}). Her vitals were stable and she showed no signs of fever. The pain was tolerable during manipulation of the vault site. The ulcerative lesion was suspected as an old hematoma or granulation formation of the previous hysterectomy vault or possibly due to an unknown malignancy or cancerous change. For further investigation, a quick excision biopsy using sharp scissors was performed under local anesthesia at the outpatient operation room. Bleeding control was done with sutures and tampon gauzes were inserted intravaginally. Nevertheless, the patient returned to the clinic within a few hours with excessive bleeding. Upon re-examination, active bleeding was visible at the vault site, and additional maneuvers including sutures were performed for hemostasis. The pathology reports showed normal vaginal tissue with non-specific loss of surface epithelium and subepithelial hemorrhage. The diagnosis was uncertain but any kind of malignancy could be ruled out. Ten days later, the patient revisited our clinic with another incidence of massive vaginal bleeding with large clots. She looked pale and anemic in general with vital signs showing an increased pulse with a decreased blood pressure at an initial 130/70 to 100/60. The patient also complained of symptoms of dizziness. Her lab results showed a hemoglobin level of 6.9 g/dL and hematocrit of 20.2%. Further attempts to achieve hemostasis with ball electrode cauterization were insufficient. The lesion healed only for a short while and the fragile tissue continued to bleed. Because of the massive bleeding, a clinical diagnosis of a possible vascular malformation or uterine artery pseudoaneurysm was considered. An emergent exploratory laparoscopic operation was performed to control vault bleeding and to evaluate any possible vascular complication or collateral arterial bleeding. Due to anemic conditions, transfusion with 2 pints of packed red blood cells were necessary prior to the operation. Upon surgery, other peritoneal structures including both ovaries were grossly normal and showed no signs of endometriosis. The pelvic side vault site was securely sealed with normal peritoneum covering the vault with no dehiscence or any other complication. Under conventional laparoscopy, after dissecting the anterior rectal wall and bladder peritoneum, a full thickness excision of approximately 2 cm in diameter was performed at the vault site and resealed with sutures. Pathology results of the excised lesion was confirmed to be consistent with endometriosis ([Fig. 2.](#F2){ref-type="fig"}). Hence, our difficult diagnosis of delayed vault site bleeding turned out to be a rare case of vault site iatrogenic endometriosis. Discussion ========== Endometriosis is defined as the presence of endometrial tissue lesions or nodules that are similar to the endometrium but are present at sites outside the uterus \[[@B3]\]. Common sites of endometriosis include the ovaries, pelvic viscera and the peritoneum and the presentations can vary from a few minimal lesions on pelvic organs to massive endometric ovary cysts or even extensive adhesions involving subperitoneal spaces, intestinal system, and urinary system \[[@B4][@B5]\]. Rare cases of extrapelvic endometriosis may result from vascular or lymphatic dissemination of endometrial cells to many gynecologic and non gynecologic sites. It has been reported that endometriosis of the skin and soft tissue makes up about 3.5% of cases of extrapelvic endometriosis with a majority of such cases occurring in surgical scars following operations of the uterus \[[@B5][@B6][@B7]\]. Although rare, there have been few reports of vaginal vault endometriosis with patients presenting with irregular or cyclic menstrual bleeding several months or years after hysterectomy \[[@B8]\]. However, those cases had a history of a functional endometriosis at the ovaries with adhesions or a fistulous tract to the vault or even some endometriotic spots left behind near the vault site \[[@B9]\]. Usually in cases of sudden bleeding with a history of prior hysterectomy, pathologic causes such as granulation tissue formation or malignancies such as cervical stump cancer must be excluded \[[@B10]\]. Among the various complications of hysterectomy, here we report an incidence of a patient who developed endometriosis at the vaginal vault with no previous history of the disease. There have been some reports where endometriosis found on scar tissues with no prior history of the disease. Cases such as scar endometriosis developing after a cesarean section are one of them \[[@B11]\]. However vault endometriosis after hysterectomy is an extremely rare complication and therefore difficult to diagnose. In our case, the previous hysterectomy had no evidence of endometriosis on other sites. Even the pathologist, although having suspicions for endometriosis, could not give a clear diagnosis due to such a rare occurrence. However after examining the final vault excision specimen and correlating with the symptoms with our patient, the pathologist confirmed both biopsy results as a typical histology for endometriosis. The possible pathophysiology in our particular case is suspected to be endometrial fragment implantation to the vaginal vault site at the time of surgery. We believe a risk of development into vault site endometriosis could be due to a uterus showing adenomyosis and or maybe due to a secretory phase endometrium. Our patient was on the 25th to 26th day of her menstrual cycle during the previous surgery, where the endometrium thickness was about 9.2 mm and the pathology proved the endometrium to be at secretory phase. Implantation during morcellation of the adenomyotic uterus or the secretory phase endometrium probably had an effect on the living tissues of the vagina, with transplantations to the site through cell adhesion mechanisms and then progression to iatrogenic endometriosis of the vaginal vault \[[@B3][@B8]\]. This theory is highly possible although further investigation and studies are required. Although such complications are extremely rare, iatrogenic vault site endometriosis must be considered in delayed bleeding occurring in a post-hysterectomy patient when other diagnoses have been excluded. Due to the nature of these lesions being highly likely to recur, the principle of treatment as in any other extra pelvic endometriosis, is total surgical excision. **Conflict of interest:** No potential conflict of interest relevant to this article was reported. {#F1} {#F2}

INTRODUCTION {#s1} ============ Postoperative adjuvant systemic therapy is considered to be an integral component of the management of primary breast cancer \[[@R1], [@R2]\]. The decision to give adjuvant chemotherapy is based on prognostic and predictive factors, such as age, axillary lymph node status, histologic grade, tumor size, and hormone receptor (HR) status \[[@R1], [@R2]\]. Several multiple gene assays have been demonstrated to predict the survival of HR-positive patients, and help physicians and patients to decide whether to administer adjuvant chemotherapy \[[@R3]--[@R6]\]. However, these assays only test the alteration of gene expression from tumor tissues but do not test the underlying genetic variations of the patient \[[@R3]--[@R6]\]. Our and others' studies have previously demonstrated that patients with different genotypes of single nucleotide polymorphisms (SNPs) of estrogen and tamoxifen metabolizing genes, such as *CYP19*, *COMT, CPY2D6*, and *SULT1A1*, may carry different responses to anti-estrogen treatment and hence have different outcomes \[[@R7]--[@R11]\]. In addition to these candidate genes, SNPs identified from genome-wide association studies (GWAS) have been found to be associated with breast cancer risk \[[@R12]--[@R16]\] and survival \[[@R17], [@R18]\]. Taken together, we hypothesized that host factors, as shown by SNPs identified from GWAS and SNPs of genes involved in the metabolism of estrogen, tamoxifen, and chemotherapeutic agents (Figure [1](#F1){ref-type="fig"}), may influence the effect of adjuvant treatment, and thus the survival of breast cancer patients. {#F1} In the present study, we aimed to investigate whether these 34 SNPs, which included GWAS-identified genes, such as, *MAP3K1* rs889312, *FGFR2* rs2981582, *TNRC9* (or *TOX3*) rs3803662, *HCN1* rs981782, and *5p12* rs10941679 and rs4415084; candidate genes involved in the metabolism of estrogen, such as, *CYP19* (TTTA)n, rs4646, rs1065779, rs1870050, and rs700519, *COMT* rs4680, *ESR1* rs3020314, rs3020396, rs2982684, rs1801132, rs2234693, and rs2046210, and *UGT1A1* rs4148323; the metabolism of tamoxifen, such as *CYP2C9* rs1057910, *CYP2C19* rs4244285 and rs4986893*, CYP3A5* rs776746, and *CYP2D6* rs16947, rs1065852, rs28371725 and rs3892097; and the metabolism of chemotherapeutic agents, such as *ABCB1* rs1128503, rs2032582, and rs1045642*, ALDH3A1* rs2231142 and rs2228100, and *CYP2B6* rs4802101 and rs3211371 are associated with the prognoses, including the distant disease-free survival (DDFS), disease-free survival (DFS) and overall survival (OS), of early-stage HR-positive breast cancers with negative or 1 to 3 nodal metastases. RESULTS {#s2} ======= Clinicopathologic features of hormone receptor-positive patients {#s2_1} ---------------------------------------------------------------- Four hundred and fourteen patients were included in the study. As shown in Table [1](#T1){ref-type="table"}, the median age was 48 years (range 23-81 years) and 251 subjects were premenopausal and 163 were postmenopausal. The clinicopathologic characteristics and treatments are also listed in Table [1](#T1){ref-type="table"}. As shown in Table [1](#T1){ref-type="table"}, histologic subtypes of our breast cancer included infiltrating ductal carcinoma (IDC) (n=349, 84.3%), infiltrating lobular carcinoma (ILC) (n=16, 3.9%), medullary carcinoma (n=3, 0.7%), ductal carcinoma *in situ* (DCIS) with microinvsion (n=14, 3.4%), tubular carcinoma (n=4, 0.9%), mucinous carcinoma (n=23, 5.6%), and papillary carcinioma (n=5, 1.2%). Briefly, 384 (92.8%) of 414 patients received adjuvant hormonal therapy with tamoxifen, and 30 (7.2%) patients received ovarian ablation or a luteinizing hormone-releasing hormone agonist with or without tamoxifen. Because aromatase inhibitor was not reimbursed by the National Health Insurance for our patients treated with multimodality treatment between January 1, 1994 and June 30, 2006, none of them received aromatase inhibitor. Two hundred and ninety-six patients (71.5%) were LN-negative, whereas 118 patients (28.5%) had 1 to 3 LN metastases. One hundred and sixty-three (39.4%) did not receive chemotherapy, and 251 (60.6%) received standard adjuvant chemotherapy (Table [1](#T1){ref-type="table"}). Furthermore, 308 (74.4%) of 414 patients were positive for both ER and PR. ###### Demographics and clinical characteristics of 414 hormone receptor (HR)-positive breast cancer patients Characteristic HR(+) (N=414) --------------------------------- --------------- Age (years)  Median (range) 48 (23-81) LN  0 296 (71.5)  1-3 118 (28.5) Menopausal status  Premenopausal 251 (60.6)  Postmenopausal 163 (39.4) Pathology  Infiltrating ductal carcinoma 349 (84.3)  Infiltrating lobular carcinoma 16 (3.9)  Medullary carcinoma 3 (0.7)  DCIS+Microinvasion 14(3.4)  Tubular carcinoma 4 (0.9)  Mucinous carcinoma 23 (5.6)  Papillary carcinoma 5 (1.2) Grade  I 171 (41.3)  II 187 (45.2)  III 56 (13.5) Tumor size (cm)  \<=2 219 (53.2)  \>2-5 193 (46.8)  Missing 2 Hormone receptor status  ER (+) PR (+) 308 (74.4)  ER (+) PR (-) 64 (15.5)  ER (-) PR (+) 42 (10.1) Adjuvant hormone therapy  Tamoxifen 384 (92.8)  Others\* 30 (7.2) Adjuvant chemotherapy  CT\*\* 251 (60.6)  No CT 163 (39.4) Abbreviations: LN, lymph node; DCIS, ductal carcinoma *in situ*; ER, estrogen receptor; PR, progesterone receptor; CT, chemotherapy \*: Ovarian ablation or luteinizing hormone-releasing hormone \*\* CT regimen: CEF (117 cases, 46.6%), CMF (97 cases, 38.6%), AC (26 cases, 10.4%), and AC+paclitaxel (11 cases, 4.4%) C, cyclophosphamide; E, epirubicin; F., 5-FU; M, methotrexate; A, adriamycin. The median follow-up period was 10.6 years (7.2% of patients were followed-up for more than 15 years, and 5.1% for less than 5 years); by the end of the follow-up period, 51 (12.3%) patients had died (43 \[84.3%\] due to breast cancer, and 8 \[15.7%\] due to causes not related to breast cancer), and 363 remained alive. Among the patients who had died due to causes that were not related to breast cancer, 1 experienced senility without the presence of psychosis; 1 had diabetes; 2 had malignancies other than breast cancers; 1 had lymphoma; 1 had a urinary tract infection that was accompanied by sepsis; 1 had a malignant neoplasm of the urethra; and 1 had coronary atherosclerosis. Due to the limited proportion of deaths that were not related to breast cancers, we believe that these were not confounding factors in our results. SNPs associated with good survival {#s2_2} ---------------------------------- Using the stepwise selection multiple Cox model analyses (adjusted multiple SNPs and clinicopathologic features), we revealed that *ESR1* codon325 rs1801132 (G/G/+G/C vs. C/C) was the only SNP significantly associated with good survival in all women (DDFS, *P* = 0.05) (Table [2](#T2){ref-type="table"}). ###### Multiple stepwise selection cox model of the predictors of survival in hormone receptor-positive early breast cancer patients Total patients DDFS DFS OS -------------------------------------------------------------------------- --------------------------------- ---------- --------------------- ---------- --------------------------------- ---------- **aHR (95%CI)** ***P*** **aHR (95%CI)** ***P*** **aHR (95%CI)** **P** ESR1 codon325 rs1801132 (G/G/+G/C vs. C/C) 0.6 (0.3-1.0) 0.05 UGT1A1 rs4148323 (A/A+A/G vs. G/G) 1.9 (1.1-3.1) 0.02 CYP2B6 rs3211371 (T/C vs. C/C) 322.2 (25.2-4113.7) \<0.0001 140.0(14.3-1375.2) \<0.0001 129.1(14.0-1190.1) \<0.0001 MAP3K1 rs889312 (C/C vs. C/A+A/A) 2.3 (1.4-3.8) 0.002 2.1 (1.3-3.4) 0.001 2.1 (1.1-3.8) 0.02 HCN1 rs981782 (A/A+A/C vs. C/C) 4.6 (1.1-19.1) 0.04 ER (-) PR (+) vs. ER (+) PR (-)/ER (+) PR (+) 2.0 (1.1-3.8) 0.02 2.3 (1.1-5.0) 0.03 **Premenopausal patients** CYP2B6 rs4802101 (T/T vs. C/C+C/T) 3.3 (1.4-6.9) 0.004 CYP2B6 rs3211371 (T/C vs. C/C) 18.0 (2.0-165.2) 0.01 118.0 (10.3-1349.6) 0.0001 70.5 (6.4-779.6) 0.0005 MAP3K1 rs889312 (C/C vs. C/A+A/A) 2.4 (1.3-4.4) 0.007 2.0 (1.1-3.4) 0.02 Pathologic status of grade III vs. grade I+II 1.8 (1.1-3.0) 0.03 ABCB1 rs2032582 (C/C vs. C/T+T/T) 3.4 (1.0-11.3) 0.05 ALDH3A1 rs2231142 (G/G vs. G/T+T/T) 0.6 (0.3-1.0) 0.05 ER (-) PR (+) vs. ER (+)PR (-)/ER (+) PR (+) 2.2 (1.0-4.5) 0.04 **Postmenopausal patients** without any significant markers without any significant markers **Patients receiving adjuvant chemotherapy (total women)** MAP3K1 rs889312 (C/C vs. C/A+A/A) 2.0 (1.0-3.8) 0.04 2.3 (1.2-4.1) 0.008 **Patients receiving** **adjuvant hormonal therapy alone (total women)** UGT1A1 rs4148323 (A/A+A/G vs. G/G) 2.9 (1.2-6.7) 0.01 CYP2B6 rs3211371 (T/C vs. C/C) 68.6 (6.7-697.4) 0.0004 126.5 (7.9-2022.4) 0.0006 297.3 (16.3-5420.9) 0.0001 Pathologic status of grade III vs. grade I+II 2.2 (1.1-4.6) 0.03 ESR1_pvuII rs2234693 (C/C+C/T vs. T/T) 0.3 (0.1-0.8) 0.01 MAP3K1 rs889312 (C/C vs. C/A+A/A) 3.0 (1.2-7.8) 0.02 Abbreviation: DDFS, distant disease-free survival; DFS, disease-free survival; OS, overall survival; aHR, adjusted hazard ratios. SNPs associated with poor survival {#s2_3} ---------------------------------- In multiple stepwise selection Cox model analyses, SNPs including *UGT1A1* rs4148323*, CYP2B6* rs3211371*, MAP3K1* rs889312*, HCN1* rs981782*, CYP2B6* rs4802101 and *ABCB1* rs2032582 were associated with poor survival (Table [2](#T2){ref-type="table"}). Among them, *CYP2B6* rs3211371 (*P* \< 0.0001 for DDFS, DFS, and OS) and *MAP3K1* rs889312 (*P* = 0.002 for DDFS, *P* = 0.001 for DFS, and *P* = 0.02 for OS) were associated with poor survival for all women; and these two SNPs were predominantly associated with premenopausal women (*CYP2B6* rs3211371, *P* = 0.01 for DDFS, *P* = 0.0001 for DDFS, and *P* = 0.0005 for OS; *MAP3K1* rs889312, *P* = 0.007 for DDFS and *P* = 0.02 for DFS), but not associated with postmenopausal women. As shown in Table [3](#T3){ref-type="table"}, patients with *CYP2B6* rs3211371 (T/C) had significantly poorer DDFS, DFS and OS than those with *CYP2B6* rs3211371 (C/C). Furthermore, patients with *MAP3K1* rs889312 (C/C) had significantly poorer DDFS and DFS, and a poorer OS than those with *MAP3K1* rs889312 (C/A+A/A) (Table [3](#T3){ref-type="table"} and Figure [2](#F2){ref-type="fig"}). ###### Proportion of 5-year and 10-year survival according to SNPs of *CYP2B6* rs3211371 and *MAP3K1* rs889312 Survival rate Genotype DDFS DFS OS ---------------------------- --------------- ---------------- ----------------- ---------------- ----------------- ---------------- ----------------- **5-year (%)** **10-year (%)** **5-year (%)** **10-year (%)** **5-year (%)** **10-year (%)** ***CYP2B6*** **rs3211371** **T/C** 0 0 0 0 0 0 **C/C** 91.91. 82.1 90.4 77.9 95.1 88.8 ***P*-value** \< 0.01 \< 0.01 \< 0.01 ***MAP3K1*** **rs889312** **C/C** 88.9 77.8 85.3 69.3 94.3 84.1 **C/A+A/A** 94.7 89.2 92.9 82.2 95.2 90.7 ***P*-value** 0.029 0.014 0.07 Abbreviation: SNP, single nucleotide polymorphisms; DDFS, distant disease-free survival; DFS, disease-free survival; OS, overall survival. {#F2} Three other SNPs (*CYP2B6* rs4802101, *P* = 0.004 for DDFS; *ABCB1* rs2032582, *P* = 0.05 for OS; *ALDH3A1* rs2231142, *P* = 0.05 for DFS) were only associated with the survival of premenopausal women, whereas another 2 SNPs (*UGT1A1* rs4148323, *P* = 0.02 for DDFS; *HCN1* rs981782, *P* = 0.04 for DDFS) were associated with all women but not associated with premenopausal women or postmenopausal women. In addition to the aforementioned SNPs, multiple Cox model analyses of the associations of prognosis with individual genotypes, adjusted by the clinicopathologic characteristics listed in Table [1](#T1){ref-type="table"} but not by other SNPs ([Supplementary Table 1](#SD1){ref-type="supplementary-material"}) showed that 7 other SNPs (*CYP19* rs4646*, CYP19* rs1870050, *CYP19* rs700519, *COMT* rs4680, *CYP2D6*\*10, *FGFR2* rs2981582, and *ABCB1* rs1128503) and one *CYP19* (TTTA)n were associated with poor survival. Among them, *CYP19* (TTTA)n, *CYP19* rs1870050, *CYP19* rs700519, and *FGFR2* rs2981582) were associated with all women ([Supplementary Table 1](#SD1){ref-type="supplementary-material"}). Among these, 4 SNPs (*CYP19* rs4646, *CYP19* rs1870050, *COMT* rs4680, and *ABCB1* rs1128503) were predominantly associated with premenopausal women. The *CYP2D6*\*10 (intermediate metabolizer phenotype) was the only SNP associated with the survival of postmenopausal women, but the significance of the *CYP2D6*\*10 (intermediate metabolizer phenotype) was lost when all other SNPs were analyzed in the multiple stepwise selection Cox model (Table [2](#T2){ref-type="table"}). SNPs associated with poor or good survival in patients with adjuvant hormonal therapy alone {#s2_4} ------------------------------------------------------------------------------------------- Among women with adjuvant hormonal therapy alone (without adjuvant chemotherapy), indicating their good prognosis, one SNP (*ESR1*\_pvuII rs2234693, *P* = 0.01 for OS) was associated with good survival by multiple stepwise selection Cox model. Furthermore, three SNPs (*UGT1A1* rs4148323, *P* = 0.01 for DDFS; *CYP2B6* rs3211371, *P* = 0.0004 for DDFS, *P* = 0.0006 for DFS, and *P* = 0.0001 for OS; *MAP3K1* rs889312, *P* = 0.02 for OS) were associated with poor prognosis for women without adjuvant chemotherapy (Table [2](#T2){ref-type="table"}). Also shown in [Supplementary Table 1](#SD1){ref-type="supplementary-material"} (the associations of prognosis with individual genotypes, adjusted by conventional prognostic factors but not by other SNPs, using the multiple Cox model), *ESR1* codon325 rs1801132 were significantly associated with a better DDFS (*P* = 0.03) for patients not receiving adjuvant chemotherapy. Four SNPs (CYP19\_(TTTA)n, *COMT* rs4680, *ABCB1* rs1128503 and *FGFR2* rs2981582) were significantly associated with poor prognosis. To delineate whether SNPs are closely associated with the prognoses of patients who received adjuvant endocrine therapy alone, we again utilized the stepwise selection Cox model to analyze the SNPs identified by GWAS, and the candidate genes involved in estrogen or tamoxifen metabolisms in patients who received endocrine therapy alone, but excluded the candidate genes involved in the metabolism of chemotherapeutic agents. As shown in the [Supplementary Table 2](#SD1){ref-type="supplementary-material"}, we found that *ESR1* codon325 rs1801132 (G/G/+G/C) (involved in estrogen metabolism) was closely associated with better DDFS and *FGFR2* rs2981582 (A/A+A/G) was closely associated with poor DFS in patients who received endocrine therapy alone. SNPs associated with poor survival in patients with adjuvant chemotherapy {#s2_5} ------------------------------------------------------------------------- In a multiple stepwise selection Cox model, *MAP3K1* rs889312 was significantly associated with patients receiving adjuvant chemotherapy and thus poor prognosis (*P* = 0.04 for DDFS and *P* = 0.008 for DFS) (Table [2](#T2){ref-type="table"}). As shown in [Supplementary Table 1](#SD1){ref-type="supplementary-material"} (multiple Cox model analyses adjusting conventional prognostic factors but not other SNPs), *CYP19* rs1870050 was associated with patients receiving adjuvant chemotherapy and thus poor prognosis. DISCUSSION {#s3} ========== In the present study, we demonstrated that genetic variants of the host, such as SNPs of *MAP3K1*, *CYP2B6*, *UGT1A1, HCN1*, *ABCB1*, and *ALDH3A1*, may worse the prognosis of HR-positive breast cancer patients, predominantly for premenopausal women. Of them, 92.8% patients received adjuvant hormonal therapy with tamoxifen. Whether new treatment, such as the GnRh analogue plus aromatase inhibitor, improves the survival of the SNP-poor prognostic group compared to treatment with tamoxifen deserves further study. In addition, several studies have revealed that a longer duration of adjuvant hormonal therapy improves the survival of HR+ patients \[[@R19]--[@R21]\], and host factors may be helpful in the selection of patients who may benefit more from longer duration of hormonal therapy. In the present study, using the multiple stepwise selection Cox model (adjusted multiple SNPs and clinicopathologic features), we found that *ESR1* codon325 rs1801132 (G/G/+G/C vs. C/C) was the only SNP significantly associated with a good DDFS, whereas *MAP3K1* and *CYP2B6* were significantly associated with poor DDFS, DFS, and OS in all women. Interestingly, *CYP2B6* rs3211371 (T/C) and *MAP3K1* rs889312 (C/C) were associated with poor prognosis in patients who receive adjuvant hormonal therapy alone, whereas *MAP3K1* rs889312 (C/C) was significantly associated with poor DDFS and DFS in patients receiving adjuvant chemotherapy. After excluding the candidate genes involved in the metabolism of chemotherapeutic agents in the multiple stepwise selection Cox model, we found that *ESR1* codon325 rs1801132 (G/G/+G/C) and *FGFR2* rs2981582 (A/A+A/G) were closely associated with better DDFS, and poor DFS, respectively, in patients who received adjuvant hormonal therapy alone ([Supplementary Table 2](#SD1){ref-type="supplementary-material"}). Although our sample size is limited (patients treated with endocrine therapy alone, n=163), these findings indicate that the variations in the genes that participate in the cell proliferation pathways (e.g. *FGFR2*) and in the metabolism of anti-hormone drugs may influence the anti-endocrine effect of the therapy, and thus determine the prognoses of this subgroup of patients. Further validation of the prognostic value of the SNPs identified in our study in a larger cohort of hormone receptor (HR)-positive patients who receive anti-hormone therapy alone is merited. Several *CYP2B6* genotypes were associated with the metabolism of *CYP2B6* substrate drugs, including cyclophosphamide and tamoxifen, frequently used in adjuvant therapy for breast cancer \[[@R22]\]. In breast cancer patients receiving adjuvant chemotherapy with cyclophosphamide and doxorubicin, *CYP2B6* rs3745274 (*CYP2B6\*9*) was reported to be associated with a poor OS \[[@R23]\]. Our findings showed that *CYP2B6* rs4802101 (T/T), and *CYP2B6* rs3211371 (T/C) were associated with a poor DDFS in premenopausal women. The association between certain *CYP2B6* SNPs and the outcome of breast cancer patients receiving tamoxifen alone has not yet been reported. Our study also showed that the minor allele (T) of *CYP2B6* rs3211371 was associated with poor DDF, DFS, and OS in all women, and in patients not receiving adjuvant chemotherapy supposedly with a good prognosis, but receiving tamoxifen or ovarian ablation. The C to T substitution of *CYP2B6* rs3211371 results in the substitution of arginine for cysteine; thus, it is speculated that the presence of this polymorphism may decrease the production of *CYP2B6* and further hamper the metabolism of anti-hormone agents \[[@R24]--[@R26]\]. However, the estimated HRs were relatively imprecise because of less frequent SNPs in the T allele. Bochud et al. recently reported that the rare G allele of rs8099917 near the *IL28B* gene was associated with poor responses to interferon therapy in patients with chronic hepatitis C who were infected with non-1 HCV genotypes \[[@R27]\]. Chen et al. also described a rare germline polymorphism, *YAP1* R331W, which is associated with an increasing risk of lung adenocarcinomas \[[@R28]\]. Pathogenic rare variants of *BRCA2* have been found to be associated with hereditary breast and ovarian cancers by the 1000 Genomes dataset \[[@R29]\]. Our current study has also identified a very low minor allele frequency of 0.04 at *CYP2B6* rs3211371 (T/C), and this rare allele was found to be associated with a poor prognosis. Further exploration of this rare variant SNP, *CYP2B6* rs3211371, through a rapid growth sequencing technology and a high-density SNP genotyping array \[[@R30], [@R31]\] will enable us to have increasing opportunities to swiftly detect rare genetic alleles, and to further investigate whether these rare variants could determine the responses to treatments and the subsequent prognoses of breast cancers. In the GWAS study, *MAP3K1* rs889312 was found to be associated with breast cancer risk \[[@R12], [@R32], [@R33]\]. *MAP3K1* participates in the MAPK signal transduction pathway, responding to a number of mitogenic and metabolic stimuli, including estrogen, which may influence breast cancer susceptibility by cell proliferation \[[@R32]\]. Growing evidence has demonstrated that MAPKs and their endogenous negative regulator, MAPK phosphatase-1 (MKP-1), may involve in the development of resistance to tamoxifen and chemotherapeutic agents \[[@R34], [@R35]\]. These mechanisms may explain why our patients with the C/C allele of *MAP3K1* rs889312 had a poor prognosis, even in patients receiving adjuvant chemotherapy. In addition to SNPs of *CYP2B6* rs3211371 and *MAP3K1* rs889312, some SNPs of candidate genes or genes identified from GWAS were associated with poor survival, which showed (1) GWAS-identified SNP; *HCN1* rs981782, poor DDFS for all women; (2) Estrogen metabolism-associated SNP; *UGT1A1* rs4148323, poor DDFS for all women and for patients without chemotherapy; and (3) Chemotherapeutic agents for metabolism-associated SNPs; *ABCB1* rs2032582, poor OS for premenopausal women, and *ALDH3A1*, poor DFS for premenopausal women; *CYP2B6* rs4802101, and poor DDFS for premenopausal women. In contrast to the aforementioned SNPs, SNPs of estrogen metabolism, *ESR1* codon325 rs1801132 (G/G/+G/C vs. C/C), and *ESR1 pvuII* rs2234693 (C/C+C/T vs. T/T) were associated with a better DDFS in all women, and a better OS in patients without adjuvant chemotherapy, respectively. Further study of the underlying mechanisms for the better prognosis of patients with genetic variants of *ESR1* codon325 rs1801132 and *ESR1 pvuII* rs2234693 is warranted. Although the aforementioned SNPs did not show consistent associations between OS, DFS, and DDFS, we cannot rule out potential confounding factors resulting from the relatively small frequency of minor alleles or a proportion of local recurrence and distant metastases that were not reported but death was noted in the death registry used in this study. However, in the present study, the aforementioned SNPs were not associated with prognosis in postmenopausal women. Previously, we had reported that *CYP19* (TTTA)n and *CYP19* genetic polymorphisms haplotype *AASA* were closely associated with poor survival in premenopausal patients with LN-negative and HR-positive breast cancers \[[@R10], [@R36]\]. In this study, we found that SNPs identified by GWAS (*MAP3K1* rs889312), and SNPs involved in the metabolism of chemotherapeutic agents (*ABCB1* rs2032582*, ALDH3A1* rs2231142, and *CYP2B6* rs4802101 and rs3211372) were associated with the prognoses in premenopausal women, but not with the prognoses in postmenopausal woman. Although we cannot rule out potential confounding effects resulting from a relatively smaller sample of postmenopausal patients, the possible reasons for the aforementioned SNPs affecting the prognoses of our premenopausal female patients are (1) the proliferation of HR-positive breast cancer cells is more estrogen-dependent in premenopausal women than in postmenopausal woman, and anti-hormone therapy (mostly with tamoxifen) or chemotherapy (partial anti-hormone effect) might cause greater decreases in the estrogen synthesized by the ovaries to support the growth of breast cancers in premenopausal women \[[@R37], [@R38]\] (2) the premenopausal women harboring the aforementioned SNPs may have higher levels of estrogen despite the anti-hormone therapy and anti-chemotherapy effects, and the existing estrogen may activate hitherto quiescent tumor cells, and may thus promote the proliferations, migrations, and distant metastases of breast cancers \[[@R36]--[@R38]\]. Previous studies have demonstrated that MAP3K1 could trigger the transcriptional activities of the ERs in endometrial and ovarian cancer cells \[[@R39]\]. In the TCGA data on breast cancers, *MAP3K1* alterations were more frequently found in the luminal A subtype than in other subtypes of breast cancers \[[@R40]\]. Although the relationship between estrogen levels and the SNP, *MAP3K1* rs889312, remains unclear, we speculated that the C/C allele of *MAP3K1* rs889312 may alter estrogen metabolism, and thus contribute to the progression of estrogen-dependent breast cancers, especially in premenopausal women. In a recent study assessing the relationship between 11 GWAS-identified breast risk-associated SNPs, including *CASP8* rs17468277, *TGFB1* rs1982073, *FGFR2* rs2981582, 8q24 rs13281615, *LSP1* rs3817198, *MAP3K1* rs889312, *TOX3* rs3803662, 2q35 rs13387042, *SLC4A7* rs4973768, *COX11* rs6504950, and rs10941679 (5p12), and 62 candidate/GWAS SNPs and prognosis of 25853 breast cancer patients (with a median follow-up of 6.4 years, 15.8% died), the authors showed that only *TOX3* rs3803662 (T/T) was significantly associated with a poorer OS (HR=1.21, *P*= 0.0002, after adjusting age, tumor size, nodal status and grade) \[[@R17]\]. Further analyses showed that *TOX3* rs3803662 (T/T) remained a poor prognostic factor in ER-positive patients, but lost significance in ER-negative patients \[[@R17]\]. However, Riaz et al. showed that *TOX3* rs3803662 was not associated with a short metastasis-free survival in 1290 LN-negative breast cancer patients without adjuvant chemotherapy \[[@R41]\]. Our results also showed that *TOX3* (*TNRC9*) rs3803662 was not associated with the DDFS, DFS, and OS in HR-positive early breast cancer patients (71.5% are LN-negative). Another recent study evaluating 8 risk SNPs, including *FGFR2* rs1219468 and *TOX3* rs8051542, which were different from our studies of *FGFR2* rs2981582 and *TOX3* (*TNRC9*) rs3803662, showed that only two SNPs, 16q12 rs12443621 and 17q23 rs6504950, influenced OS after adjusting for age, clinical stage, and treatment \[[@R18]\]. The different composition of study populations may explain the different findings of our results from their studies \[[@R17], [@R18]\]. For example, our patients were HR-positive, LN node-negative, or had up to 3 positive LNs; they were also Taiwanese, had detailed information concerning their adjuvant chemotherapy regimen, and underwent long-term follow-up with a median of 10.6 years (7.2% followed for more than 15 years, 5.1% for less than 5 years). Further studies exploring the influence of GWAS-identified genes, such as 16q12 rs12443621 and 17q23 rs6504950, on the survival of HR-positive and LN node-negative breast cancers or those with up to 3 positive LNs are merited because these SNPs were reported after we genotyped our GWAS-identified genes \[[@R17], [@R18]\]. In this study, *CYP2D6\*10* was the only genotype associated with worse survival of postmenopausal women after adjustment for the conventional prognostic factors listed in Table [1](#T1){ref-type="table"}. *CYP2D6 \*10* lost its significance when all the other SNPs were adjusted together in the multiple stepwise selection COX model, which may explain why the associations of *CYP2D6* and the survival of tamoxifen-treated breast cancer patients conflict in different reports. In Asians, *CYP2D6\*10* is the predominant polymorphism that accompanies the intermediate metabolizer phenotype, in which 2 metabolites of tamoxifen, 4-hydroxytamoxifen (4OHtam) and 4-hydroxy-N-desmethyl tamoxifen (endoxifen) exhibit greater ER affinity and are predominantly catalyzed by cytochrome *CYP2D6* \[[@R42]--[@R44]\]. Previous studies suggested that *CYP2D6\*10* alleles decreased CYP2D6 activity; thus, a shorter recurrence-free survival period was observed in Asian patients with adjuvant tamoxifen \[[@R8], [@R45]\]. Two studies reported that the poor or intermediate metabolizer of *CYP2D6* was not associated with the clinical outcome of postmenopausal Caucasian women patients with HR-positive operable invasive breast cancer receiving adjuvant tamoxifen \[[@R46], [@R47]\]. However, these studies did not include premenopausal patients and did not analyze *CYP2D6\*10* alleles. In this study, 251 (60.6%) patients received different standard adjuvant chemotherapy agents, including cyclophosphamide, epirubicin, 5-fluorouracil, methotrexate, doxorubicin, and paclitaxel. In clinical practice, the choices of different standard chemotherapeutic agents and regimens made by physicians depend upon their assessments of the clinicopathological characteristics of patients, including tumor sizes, tumor grades, estrogen receptor (ER), progesterone receptor (PR), lymph nodes (LNs), underlying comorbidities in patients, and the potential toxicities of the different chemotherapy regimens. Therefore, as shown in Table [1](#T1){ref-type="table"}, various chemotherapeutic agents were inevitably included in this study. However, in the current study, the standard adjuvant chemotherapy that was administered to LN-positive and LN-negative patients with high-risk factors after undergoing breast surgeries was based on the indications and the adjuvant chemotherapeutic regimens and doses described in previously published literature, or those recommended by the NCCN guidelines, the NIH consensus, and the St. Gallen consensus \[[@R2], [@R48], [@R49]\]. As shown in Table [4](#T4){ref-type="table"}, LN-positivity, larger tumor sizes, and higher histologic grades were determining factors for patients to receive adjuvant chemotherapy. However, because only a limited number of patients in our study received adjuvant chemotherapy and heterogeneous chemotherapy regimens, the interpretations of the associations between the SNP, *MAP3K1* rs889312 (C/C), and the DDFS and DFS of patients who received adjuvant chemotherapy should be cautious. Further validation of our identified prognostic SNPs in a larger cohort of HR-positive patients with LN 1--3 who receive the same chemotherapy regimens is warranted. ###### Multiple stepwise selection logistic regression model analyses of the predictors of patients whether receiving adjuvant chemotherapy Covariate aOR (95%CI) *P* ------------------------------------------------------------------------------------------------ -------------------- --------- Infiltrating ductal carcinoma + Infiltrating Lobular carcinoma+Medullary carcinoma. vs. others 23.4 (3.5-156.6) 0.001 LN 1-3 vs. 0 154.8 (19.7-999.9) \<.0001 Size 1.9 (1.4-2.7) 0.0001 Grade 1.7 (1.0-2.8) 0.03 aOR: adjusted odds ratio In summary, our findings suggested that genetic variations in genes participating in the cell proliferation pathways and in the metabolism of anti-hormone drugs and anti-chemotherapy agents are likely to synergistically influence the outcome of HR-positive breast cancer patients. These findings provide additional evidence that the genetic variants may affect the prognosis of breast cancer. Functional analysis and validation of the biologic significances of SNPs of *CYP2B6* rs3211371 and *MAP3K1* rs889312 in this subtype of breast cancer patients are warranted. In addition, patients with *MAP3K1* rs889312 (C/C) might need different or more aggressive treatments. PATIENTS AND METHODS {#s4} ==================== Study cohort and sources of information {#s4_1} --------------------------------------- Eligible women were newly diagnosed patients with stage I or II (AJCC 2007) HR-positive early breast cancers diagnosed at the National Taiwan University Hospital between January 1, 1994 and June 30, 2006. One pathologist (Dr. Lien) reviewed the histological grade and hormone receptor status of the primary tumor of each patient. Patients were considered HR-positive if the percentage of estrogen receptor (ER)- or progesterone receptor (PR)-positive epithelial cells was ≥ 10% \[[@R2], [@R50]\]. Genomic DNA and detailed demographic information were obtained from the patients and their medical charts with their written informed consent. The pathologic review, blood samples, and genetic studies were approved by the National Taiwan University Hospital (NTUH) ethics committee (200906075R). Pathologic and clinical information about treatment (including type of surgery, receipt or non-receipt of adjuvant systemic therapy, and type and dose of adjuvant systemic therapy) and follow-up information (including recurrence and distant metastasis) were obtained from pathology reports and clinical records. Patients with high-risk factors, such as grade III cancers, large tumors, and lymph node (LN) positivity (N1), all received standard adjuvant chemotherapy, such as CMF, CEF, CAF, AC/EC, or AC/EC followed by paclitaxel/docetaxel regimens as defined in our previous study \[[@R2]\]. In the present study, the definition of menopausal status was based on our previous study: (1) If menstruation had taken place within one year, the woman was considered to be premenopausal, and, if not, postmenopausal (2) Women who had undergone hysterectomy without bilateral oophorectomy were considered to be premenopausal if they were younger than 52 and postmenopausal if older \[[@R2], [@R36]\]. As shown in Table [4](#T4){ref-type="table"}, poor prognosis factors of pathologic status, such as LN-positivity (adjusted odds ratio \[aOR\], 154.8; 95% CI, 19.7-999.9, *P* \< 0.0001), larger tumor size (aOR, 1.9; 95% CI, 1.4-2.7, *P* = 0.0001), and higher histologic grade (aOR, 1.7; 95% CI, 1.0-2.8, *P* = 0.03) were independent factors for patients to receive adjuvant chemotherapy. All enrolled patients received adjuvant hormonal therapy. Adjuvant radiotherapy was administered to all patients after breast conservation surgery \[[@R51], [@R52]\]. After surgery and adjuvant therapy, the patients were regularly followed up in our clinic. If patients were lost to follow-up, information on disease status and survival was obtained from the patients' charts, hospital cancer registry records, and the National Death Registry. Histological subgroup of HR-positive breast cancer {#s4_2} -------------------------------------------------- Histologically, tubular, mucinous, and papillary carcinomas, and ductal carcinomas *in situ* (DCIS) with microinvasions of breast cancers have more favorable prognoses than infiltrating ductal carcinomas (IDCs), infiltrating lobular carcinomas (ILCs), and medullary breast carcinomas. Unlike IDCs, the clinicopathological features of ILCs show greater association with the low-to-intermediate grade positive expression of the ER, and the negative expression or amplification of HER2 \[[@R53], [@R54]\]. However, Lorfida et al. reported that the OS of ILCs might be worse compared with those of stage-matched IDCs \[[@R55]\]. Although the responses to chemotherapy or treatment with aromatase inhibitors may be distinct between cases of ILCs and IDCs \[[@R56], [@R57]\], most clinical trials and practical clinical guidelines suggest that the treatments for ILCs and IDCs should be similar, and these should be considered as a single unified subtype of breast cancer. In addition, our patients with IDCs and ILCs exhibited similar 5-year DFS (81.3% versus 82.3%) and 5-year OS (87.3% versus 90.1%). Park et al. demonstrated that the prognoses of medullary breast carcinomas are not significantly different from those of IDCs, and that the prognoses were also determined by greater tumor sizes and axillary lymph node metastases \[[@R58]\]. As shown in [Supplementary Table 3](#SD1){ref-type="supplementary-material"}, we have demonstrated that the estimated adjusted odds ratios (aOR) for the associations of different histological subtypes with the use of adjuvant chemotherapy were more than 1 for IDCs, ILCs, or medullary carcinomas, unlike the aORs of other histological subtypes, such as mucinous carcinomas \[aOR=0.5\], DCIS with microinvasions, and tubular and papillary carcinomas. Although ILCs may be more endocrine-sensitive than IDCs, based on the similarities in the use of systemic chemotherapy and the prognoses, and the limited sample size of ILCs (n=16), we have included ILCs within the subgroup comprising IDCs and medullary carcinomas. Genotyping {#s4_3} ---------- TaqMan assays were used to genotype specific SNPs, including *CYP19* rs4646, rs1065779, rs1870050, and rs700519; *ESR1*, rs3020314, rs3020396, rs2982684, rs1801132, rs2234693, and rs2046210; *COMT* rs4680; *CYP3A5* rs776746; *CYP2C19* rs4244285 and rs4986893; *UGT1A1* rs4148323; *ABCB1* rs1128503, rs2032582, and rs1045642; *ALDH3A1* rs2231142 and rs2228100; *CYP2C9* rs1057910; *CYP2B6* rs4802101 and rs3211371; *FGFR2* rs2981582; *TNRC9* rs3803662; *MAP3K1* rs889312; *HCN1* rs981782, rs10941679, and rs4415084 in chromosome 5p12. The allelic frequencies of these SNPs are shown in [Supplementary Table 4](#SD1){ref-type="supplementary-material"}. The *CYP2D6*\*10 were determined by rs16947, rs1065852, rs28371725, and rs3892097, whereas the (TTTA)n of *CYP19* were determined by performing a polymerase chain reaction (PCR) that utilized primers and methods described previously \[[@R10]\]. PCR conditions for TaqMan assays {#s4_4} -------------------------------- The thermal cycling conditions were 50°C for 2 minutes, 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds, and 60°C for 60 seconds. The PCR reaction was performed in a total reaction volume of 5 μL containing 10 ng genomic DNA, 2.5 μL of the 2X TaqMan® Universal PCR Master Mix (Applied Biosystems), and 0.125 μL of the 40X primers/probes mixed in the 384-well plate format on ABI7900HT. The primers and probes and genotyping were performed via an Assay-by-Design method or a Made to Order Assay (Applied Biosystems). Statistical analysis {#s4_5} -------------------- Follow-up data available as of December 31, 2011 were analyzed. Distant disease-free survival (DDFS) was measured from the date of the original surgery for breast cancer to distant recurrence or death from any cause, disease-free survival (DFS) was measured from the date of the original surgery for breast cancer to local recurrence, distant recurrence or death from any cause and overall survival (OS) was measured from the date of the original surgery to the date of death from any cause or the last follow-up date \[[@R52]\]. Multiple-adjusted hazard ratio (HR) (aHR) of disease status associated with the individual genotype was assessed after adjustment for age, menopausal status, tumor size, grade, ER, PR, LN status, histopathology, adjuvant chemotherapy, and adjuvant hormonal therapy in the multiple Cox model (data are shown in [Supplementary Table 1](#SD1){ref-type="supplementary-material"}). Furthermore, all SNPs and all the above mentioned variables were further analyzed by the stepwise variable selection procedure with the significance level for entry (SLE) and the significance level for stay (SLS) set to 0.05 (data shown in Table [2](#T2){ref-type="table"}). The stepwise selection Cox model was used to identify the variables that showed significant associations with disease status. In the subgroup analysis, including premenopausal patients, postmenopausal patients, patients receiving adjuvant chemotherapy, and patients receiving adjuvant hormonal therapy alone, stepwise selection was continued as conducted in subgroup analysis. The stepwise selection Cox model has been widely used to predict the hazard rates in patients in various clinical epidemiological studies, such as, those conducted by Yang et al. \[[@R59]\], and Pande et al. \[[@R60]\]. Stepwise regression is a combination of the forward and backward selection techniques. During the iterative process of variable selection, variables are removed from the model if they are deemed non-significant. Furthermore, the whole stepwise procedure repeats between the forward and backward steps until no additional variables are added to the current model. Therefore, in our study, after the stepwise selection procedures were completed, sets of significant variables were selected and listed in Table [2](#T2){ref-type="table"}. All analyses were performed using SAS statistical software for Windows version 9.2 (SAS Institute, Cary, NC, USA). SUPPLEMENTARY MATERIALS FIGURES AND TABLES {#s5} ========================================== This study was supported by research grants MOST 104-2314-B-002-189-MY3, 105-2811-B-002-041, and 104-2314-B-002-152-MY3 from the Ministry of Science and Technology, Taiwan, and MOHW104-TD-B-111-04 and MOHW105-TDU-B-211-134005 from the Ministry of Health and Welfare, Taiwan. **CONFLICTS OF INTEREST** The authors declare no conflict of interest.

THE PRE-TRAVEL CONSULTATION {#subchapter1} =========================== Acosta Rebecca W. The pre-travel consultation is a risk-based assessment process that provides a guide to prioritizing and customizing pre-travel health care to the traveler\'s itinerary, risks, and needs. The goal of the pre-travel consultation is the effective and efficient preparation of travelers with the appropriate counseling, vaccinations, and medications to help reduce their risk of illness and injury during travel. To conduct a risk-based assessment, health-care providers involved in preparing travelers must---•Have a working knowledge of destination-specific disease and health risks and standard recommendations to promote health and prevent illness among travelers. The information and recommendations presented in this publication, the Body of Knowledge in Travel Medicine (as published by the International Society of Travel Medicine \[ISTM\]), and other authoritative sources (see Appendix B) form the basis for this knowledge.•Understand the standard for and expectations of conducting a pre-travel consultation and gain expertise in the process. The well-organized and well-executed pre-travel consultation supports consistent, appropriate, and efficient pre-travel health preparation with the following three essential elements:1.Risk Assessment2.Risk Communication3.Risk Management Risk Assessment {#cesec1} --------------- The risk assessment provides the foundation for the recommendations given during the consultation. A risk assessment involves gathering pertinent information about the itinerary ("where and when") and traveler ("who, why, what, and how") to highlight the traveler\'s risks and alert the provider to any contraindications and precautions to vaccinations or medications that may be indicated. A questionnaire designed to collect and organize the itinerary and traveler data is an essential tool to help support the risk assessment process and facilitate consistent practice (see Box 1-1). The most important information to gather includes the following:•**Itinerary data** ○Countries and regions to be visited; urban versus rural○Dates and length of travel○Purpose of travel (e.g., business, vacation, visiting friends and relatives)○Mode(s) of transportation○Planned and possible activities○Types of accommodations•**Traveler demographic and health/medical history** ○Age, sex○Vaccination history, including prior adverse events○Medical and psychiatric history (past and current)○Medications○Allergies○Pregnancy and breastfeeding status (current status and plans) A basic example of using the itinerary and traveler data includes determining if there will be a risk of yellow fever disease or a requirement for yellow fever vaccination based on the itinerary, and if there is a contraindication (e.g., egg allergy) or a precaution (e.g., \>60 years of age) to the traveler\'s receiving the vaccine. Malaria risk is another important consideration. Will the traveler be going to a region endemic for malaria, and what are the appropriate measures to help prevent malaria based on the details of the itinerary and traveler\'s medical history? During the risk assessment, the provider must remain alert to other factors about "who" will be traveling. Such factors include the traveler\'s previous travel experience, perception of risk, cultural background, peer group(s), and possible barriers to care, such as economic issues, attitudes regarding vaccine safety, and fear of vaccines. These factors may greatly affect the traveler\'s ability and willingness to accept and adhere to the recommendations, and therefore affect the pre-travel consultation. Anticipating the unique needs of high-risk travelers and preparing them for healthy travel will help prevent illness and injury. The following travelers may be considered high risk:•Travelers visiting friends and relatives (VFRs). These individuals have typically migrated from a less-developed area to a developed area and are now returning to the region of their birth. This is especially important when these individuals are traveling with new family members or children. The traveler returning to his or her country of origin may not understand the dynamics of risk and waning immunity (see the VFR section in Chapter 8).•The elderly•Families with young children•Persons traveling to adopt children abroad•Persons with weakened immune systems•Women who are pregnant or breastfeeding Risk Communication {#cesec2} ------------------ The next phase of the consultation process is focused on risk communication and includes the presentation of reliable, evidence-based information in a context appropriate for the individual traveler. Time should be allocated for discussion of the risks with the traveler to promote informed decision making about risk avoidance and prevention measures, such as vaccinations and malaria chemoprophylaxis. Risk communication depends heavily upon the risk assessment for the individual traveler, as well as that traveler\'s perception of risk. For example, three travelers may be going to the same country: one for a week-long, urban-based, business visit; the next on an adventure-seeking, backpack trip to rural areas over several months; and the third is a pregnant VFR traveler. The recommendations and preparation for each of these travelers will vary, even though the destination country is the same. It is important to give both verbal and written information to the traveler to help guide and focus the discussion and reinforce important issues based on his or her risk assessment. Examples include information pamphlets, malaria risk maps, and vaccine information statements (VISs). Through careful risk assessment and thoughtful risk communication, a risk management plan (i.e., vaccinations, medications, and targeted risk-avoidance education) takes shape. Risk Management {#cesec3} --------------- The essential elements of risk management include the following:•Selection, administration, and documentation of vaccinations○Required, recommended, and routine vaccinations should all be considered (see below)○Providers should consider indications, contraindications, precautions, and timing of dosages•Prescribing and advising about preventive medications○Where appropriate according to risk, antimalarial chemoprophylaxis and medications for travelers\' diarrhea, motion sickness, and altitude sickness•Education related to malaria prevention and adherence to chemoprophylaxis (if indicated by the risk assessment)•Information on risk and prevention of other insect-borne diseases•Instruction on methods to reduce foodborne and waterborne illness and the self-management of travelers\' diarrhea•Instruction about animal avoidance and rabies•Information to help reduce the negative effect of○Other itinerary risks (e.g., altitude, pollution)○Activity-specific risks (e.g., diving, rafting, rural road travel)○Personal behavior risks (e.g., sexually transmitted diseases)•General guidance on○Symptoms (e.g., fever, gastrointestinal or dermatologic symptoms) that may require medical attention during or after travel○Preparing a travel health kit (see the [Travel Health Kits section](#subchapter43){ref-type="sec"} later in this chapter)○Accessing medical care abroad and obtaining medical/evacuation insurance When considering vaccinations, common terms used include "required," "recommended," and "routine." Required vaccines are those needed when a destination country requires documentation of vaccine administration or some sort of medical waiver. Recommended vaccines are those vaccines that are considered based on the actual disease risk the traveler may encounter during travel. Routine vaccines refer to those vaccines that are recommended in the United States, regardless of travel. These routine vaccines are an important part of pre-travel care because many of the diseases they protect against are more common in countries outside the United States. Careful documentation of all vaccinations, medications, and specific recommendations given to the traveler helps to complete the care plan record. Providers who are registered to give yellow fever vaccine should be familiar with properly completing the International Certificate of Vaccination or Prophylaxis (ICVP) to ensure that this documentation will be accepted at the borders of destination countries (see the [Yellow Fever](#subchapter6){ref-type="sec"} section later in this chapter). Using an electronic record or standardized form facilitates documentation and helps ensure consistency of practice. Providers should plan to spend an average of 30--45 minutes conducting a complete pre-travel consultation, based on the risk assessment, given the potential complexities in preparing the traveler. Providers with limited knowledge and expertise in travel medicine and the pre-travel consultation should consider referring travelers with complex itineraries or special needs (see Chapters 7 and 8) to a travel medicine clinic or travel medicine specialist through CDC\'s Travelers\' Health website at [www.cdc.gov/travel](http://www.cdc.gov/travel){#interref7}. References 1--4 can assist those providers interested in gaining a more in-depth perspective on the expectations for providing pre-travel health care and the pre-travel consultation process. GENERAL RECOMMENDATIONS FOR VACCINATION AND IMMUNOPROPHYLAXIS {#subchapter2} ============================================================= Atkinson William Kroger Andrew Recommendations for the use of vaccines and other biologic products (e.g., immune globulin products) in the United States are developed by the Advisory Committee on Immunization Practices (ACIP) and other groups, such as the American Academy of Pediatrics. These recommendations are based on scientific evidence of benefits (immunity to the disease) and risks (vaccine adverse reactions) and, where few or no data are available, on expert opinion. The recommendations include information on general immunization issues and the use of specific vaccines. When these recommendations are issued or revised, they are published in CDC\'s Morbidity and Mortality Weekly Report (MMWR) ([www.cdc.gov/mmwr](http://www.cdc.gov/mmwr){#interref8}). This section is based primarily on the ACIP General Recommendations on Immunization. Vaccinations against diphtheria, tetanus, pertussis, measles, mumps, rubella, varicella, poliomyelitis, hepatitis A, hepatitis B, *Haemophilus influenzae* type b, rotavirus, influenza, human papillomavirus, and pneumococcal and meningococcal invasive disease are routinely administered in the United States, usually in childhood or adolescence. If persons do not have a history of adequate protection against these diseases, immunizations appropriate to their age and previous immunization status should be obtained, whether or not international travel is planned. A visit to a provider for immunizations for travel should be seen as an opportunity to bring an incompletely vaccinated person up-to-date on his or her routine vaccinations. Both the child and adolescent vaccination schedule and an adult vaccination schedule are published annually in the MMWR. Vaccine providers should obtain the most current schedules from the CDC Vaccines and Immunization website at [www.cdc.gov/vaccines/](http://www.cdc.gov/vaccines/){#interref9}. The text and [Table 2-1](#cetable1){ref-type="table"}, [Table 2-2](#cetable2){ref-type="table"}, [Table 2-3](#cetable3){ref-type="table"}, [Table 2-4](#cetable4){ref-type="table"}, [Table 2-5](#cetable5){ref-type="table"}, [Table 2-6](#cetable6){ref-type="table"}, [Table 2-7](#cetable7){ref-type="table"} , [2-9](#cetable9){ref-type="table"} --[2-10](#cetable10){ref-type="table"} , [2-18](#cetable18){ref-type="table"} --[2-19](#cetable19){ref-type="table"} , [2-21](#cetable21){ref-type="table"} , 5-2, 7-2,7-3,7-4,7-5, 8-1, 8-7 and 8-8 of this publication present recommendations for the use, number of doses, dose intervals, adverse reactions, precautions, and contraindications for vaccines and toxoids that may be indicated for travelers. For specific vaccines and toxoids, additional details on background, adverse reactions, precautions, and contraindications are found in the respective ACIP statements.Table 2-1Revaccination (booster) schedulesVaccineRecommendationJapanese encephalitisFull duration of protection unknown. Neutralizing antibodies may persist at least 2 years after primary immunization.Hepatitis A (HAV)Booster doses not recommended for adults and children who have completed the primary series (2 doses) according to the routine scheduleHepatitis B (HBV)Booster doses not recommended for adults and children who have completed the primary series (3 doses) according to the routine schedule[1](#cetablefn1){ref-type="table-fn"}Influenza1 annual dose (children 6 months to 9 years of age and certain incompletely vaccinated children should receive 2 doses separated by at least 4 weeks the first time that influenza vaccine is administered). Live attenuated influenza vaccine is approved only for healthy nonpregnant persons 2--49 years of age.Measles--mumps--rubella (MMR)2 doses of MMR vaccine separated by at least 4 weeks or other evidence of immunity (e.g., serologic testing) is recommended for persons born after 1956 who travel outside the United States. Revaccination is not recommended.Meningococcal Quadrivalent A,C,Y, W-135Revaccination after 5 years is recommended for persons who received meningococcal polysaccharide vaccine and who remain at increased risk for meningococcal disease (including some international travelers). Revaccination is not recommended after receipt of meningococcal conjugate vaccine.Pneumococcal (polysaccharide)One-time revaccination 5 years after original dose for persons with certain underlying medical conditions (e.g., asplenia) or persons who were first vaccinated at younger than 65 years of ageRotavirusBooster doses not recommendedPolio (IPV)A single lifetime booster dose is recommended for adults who have written documentation of having completed a primary series.Rabies pre-exposure vaccineNo serologic testing or boosters recommended for travelers. For persons in higher risk groups (e.g., rabies laboratory workers) serologic testing and booster doses are recommended. See [Table 2-17](#cetable17){ref-type="table"}.Tetanus/diphtheria, and acellular pertussis (Tdap)Tetanus and diphtheria booster dose is recommended every 10 years. A single dose of adolescent/adult formulation Td that includes acellular pertussis vaccine (Tdap) is recommended to replace one Td booster dose for persons 11--64 years of age. See ACIP statement for details.Typhoid oralRepeat series every 5 years.Typhoid IMBooster dose every 2 yearsVaricellaRevaccination is not recommended.Yellow feverRepeat vaccination every 10 years.[^1]Table 2-2Recommended intervals between administration of antibody-containing products and measles-containing vaccine or varicella-containing vaccine[1](#cetablefn2){ref-type="table-fn"}IndicationDoseRecommended Interval Before Measles or Varicella VaccinationTetanus (TIG)250 units (10 mg IgG/kg) IM[2](#cetablefn3){ref-type="table-fn"}3 monthsHepatitis A (IG), duration of international travel\<3-month stay\>3-month stay 0.02 mL/kg (3.3 mg IgG/kg) IM0.06 mL/kg (10 mg IgG/kg) IM 3 months3 monthsHepatitis B prophylaxis (HBIG)0.06 mL/kg (10 mg IgG/kg) IM3 monthsRabies prophylaxis (HRIG)20 IU/kg (22 mg IgG/kg) IM4 monthsVaricella prophylaxis (VZIG)125 units/10 kg (20--40 mg IgG/kg) IM (maximum 625 units)5 monthsMeasles prophylaxis (IG)Immunocompetent contactImmunocompromised contact 0.25 mL/kg (40 mg IgG/kg) IM0.50 mL/kg (80 mg IgG/kg) IM 5 months6 monthsBlood transfusion Red blood cells (RBCs), washedRBCs, adenine-saline addedPacked RBCs (Hct 65%)[3](#cetablefn4){ref-type="table-fn"}Plasma/platelet products 10 mL/kg (negligible IgG/kg) IV10 mL/kg (10 mg IgG/kg) IV10 mL/kg (60 mg IgG/kg) IV10 mL/kg (160 mg IgG/kg) IV None3 months6 months7 monthsCytomegalovirus prophylaxis (CMV IGIV)150 mg/kg maximum6 monthsRespiratory syncytial virus (RSV) monoclonal antibody (Synagis)[4](#cetablefn5){ref-type="table-fn"}15 mg/kg IMNoneIntravenous immune globulin (IGIV)Replacement therapyImmune thrombocytopenic purpura (ITP)ITPITP or Kawasaki disease 300--400 mg/kg IV400 mg/kg IV1 gm/kg IV1.6--2 gm/kg IV 8 months8 months10 months11 months[^2][^3][^4][^5]Table 2-3Recommended and minimum ages and intervals between vaccine doses[1](#cetablefn6){ref-type="table-fn"}Vaccine and Dose NumberRecommended Age for this DoseMinimum Age for this DoseRecommended Interval to Next DoseMinimum Interval to Next DoseHepatitis B (HepB)-1[2](#cetablefn7){ref-type="table-fn"}BirthBirth1--4 months4 weeksHep B-21--2 months4 weeks2--17 months8 weeksHep B-3[3](#cetablefn8){ref-type="table-fn"}6--18 months24 weeksNANADiphtheria--tetanus--acellular pertussis (DTaP)-1[2](#cetablefn7){ref-type="table-fn"}2 months6 weeks2 months4 weeksDTaP-24 months10 weeks2 months4 weeksDTaP-36 months14 weeks6--12 months6 months[4](#cetablefn9){ref-type="table-fn"},[5](#cetablefn10){ref-type="table-fn"}DTaP-415--18 months12 months3 years6 months[4](#cetablefn9){ref-type="table-fn"}DTaP-54--6 years4 yearsNANA*Haemophilus influenzae* type b (Hib)-1[2](#cetablefn7){ref-type="table-fn"},[6](#cetablefn11){ref-type="table-fn"}2 months6 weeks2 months4 weeksHib-24 months10 weeks2 months4 weeksHib-3[7](#cetablefn12){ref-type="table-fn"}6 months14 weeks6--9 months8 weeksHib-412--15 months12 monthsNANAInactivated poliovirus (IPV)-1[2](#cetablefn7){ref-type="table-fn"}2 months6 weeks2 months4 weeksIPV-24 months10 weeks2--14 months4 weeksIPV-36--18 months14 weeks3--5 years4 weeksIPV-44--6 years18 weeksNANAPneumococcal conjugate (PCV)-1[6](#cetablefn11){ref-type="table-fn"}2 months6 weeks2 months4 weeksPCV-24 months10 weeks2 months4 weeksPCV-36 months14 weeks6 months8 weeksPCV-412--15 months12 monthsNANAMeasles--mumps--rubella (MMR)-1[8](#cetablefn13){ref-type="table-fn"}12--15 months12 months3--5 years4 weeksMMR-2[8](#cetablefn13){ref-type="table-fn"}4--6 years13 monthsNANAVaricella (Var)-112--15 months12 months3--5 years12 weeks[9](#cetablefn14){ref-type="table-fn"}Var-24--6 years15 monthsNANAHepatitis A (HepA)-112--23 months12 months6--18 months[4](#cetablefn9){ref-type="table-fn"}6 months[4](#cetablefn9){ref-type="table-fn"}HepA-218--41 months18 monthsNANAInfluenza, inactivated[10](#cetablefn15){ref-type="table-fn"}6--18 years6 months4 weeks4 weeksInfluenza, live attenuated[10](#cetablefn15){ref-type="table-fn"}NA2 years4 weeks4weeksMeningococcal conjugate (MCV)11--12 years11 yearsNANAMeningococcal polysaccharide (MPSV)-1NA2 years5 years[11](#cetablefn16){ref-type="table-fn"}5 years[11](#cetablefn16){ref-type="table-fn"}MPSV-2[12](#cetablefn17){ref-type="table-fn"}NA7 yearsNANATd11--12 years7 years10 years5 yearsTdap[13](#cetablefn18){ref-type="table-fn"}≥11 years10 yearsNANAPneumococcal polysaccharide (PPV)-1NA2 years5 years5 yearsPPV-2[14](#cetablefn19){ref-type="table-fn"}NA7 yearsNANAHuman papillomavirus (HPV)-1[15](#cetablefn20){ref-type="table-fn"}11--12 years9 years2 months4 weeksHPV-22 months after dose 19 years, 4 weeks4 months12 weeks[15](#cetablefn20){ref-type="table-fn"}HPV-36 months after dose 19 years, 24 weeksNANARotavirus (RV)-1[16](#cetablefn21){ref-type="table-fn"}2 months6 weeks2 months4 weeksRV-24 months10 weeks2 months4 weeksRV-3[16](#cetablefn21){ref-type="table-fn"}6 months14 weeksNANAHerpes zoster[17](#cetablefn22){ref-type="table-fn"}60 years60 yearsNANATyphoid, inactivated (ViCPS)≥2 years≥2 yearsNANATyphoid, live attenuated (Ty21a)≥6 years≥6 yearsSee footnote [18](#cetablefn23){ref-type="table-fn"}See footnote [18](#cetablefn23){ref-type="table-fn"}Yellow Fever\>9 months[19](#cetablefn24){ref-type="table-fn"}\>9 months[19](#cetablefn24){ref-type="table-fn"}10 years10 yearsJapanese encephalitis (JE)-1≥1 year1 year7 days7 daysJE-27 days after dose 11 year, 7 days30 days14 daysJE-330 days after dose 11 year, 21 daysNANARabies-1 (pre-exposure)See footnote [20](#cetablefn25){ref-type="table-fn"}See footnote [20](#cetablefn25){ref-type="table-fn"}7 days7 daysRabies-27 days after dose 17 days after dose 121 days14 daysRabies-321 days after dose 121 days after dose 1NANA[^6][^7][^8][^9][^10][^11][^12][^13][^14][^15][^16][^17][^18][^19][^20][^21][^22][^23][^24][^25][^26]Adapted from [Table 1](#cetable1){ref-type="table"}, CDC. General recommendations on immunization. Recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Recomm Rep. 2006; 55(RR-15):1--48.Table 2-4Licensed schedule for HAVRIX[1](#cetablefn26){ref-type="table-fn"}Age Group (YRS)Dose (EL.U.)[2](#cetablefn27){ref-type="table-fn"}VolumeNo. of DosesSchedule (Months)1--187200.5 mL20, 6--2≥1914401.0 mL20, 6--12[^27][^28]Table 2-5Licensed schedule for VAQTA[1](#cetablefn28){ref-type="table-fn"}Age Group (YRS)Dose (U.)[2](#cetablefn29){ref-type="table-fn"}VolumeNo. of DosesSchedule (Months)1--18250.5 mL20, 6--18≥19501.0 mL20, 6--18[^29][^30]Table 2-6Licensed schedule for TWINRIX[1](#cetablefn30){ref-type="table-fn"}Age Group (YRS)Dose (EL.U./20 mg)[2](#cetablefn31){ref-type="table-fn"}VolumeNo. of DosesSchedule (Months)≥187201.0 mL30, 1, 6 months≥187201.0 mL40, 7, 21 days +1 year[^31][^32]Table 2-7Recommended doses of immune globulin (IG) for protection against hepatitis ASettingDuration of CoverageDose (Ml/Kg)[1](#cetablefn32){ref-type="table-fn"}Pre-exposureShort-term (1--2 months)0.02Long-term (3--5 months)0.06[2](#cetablefn33){ref-type="table-fn"}PostexposureNA0.02[^33][^34]Table 2-9Recommended doses of currently licensed formulations of hepatitis B vaccineGroupSingle-Antigen VaccineCombination VaccineRECOMBIVAX HBENGERIX-BCOMVAX[1](#cetablefn43){ref-type="table-fn"}PEDIARIX[2](#cetablefn44){ref-type="table-fn"}TWINRIX[3](#cetablefn45){ref-type="table-fn"}Dose (μg)[4](#cetablefn46){ref-type="table-fn"}Volume (ml)Dose (μg)[4](#cetablefn46){ref-type="table-fn"}Volume (ml)Dose (μg)[4](#cetablefn46){ref-type="table-fn"}Volume (ml)Dose (μg)[4](#cetablefn46){ref-type="table-fn"}Volume (ml)Dose (μg)[4](#cetablefn46){ref-type="table-fn"}Volume (ml)**Infants (\<1 year)**5[7](#cetablefn49){ref-type="table-fn"}0.510[7](#cetablefn49){ref-type="table-fn"}0.550.5100.5NA[5](#cetablefn47){ref-type="table-fn"}NA**Children (1--10 years)**50.5100.550.5100.5NANA**Adolescents11--15 years**10[6](#cetablefn48){ref-type="table-fn"}1.0NANANANANANANANA**11--19 years**50.5100.5NANANANA20[3](#cetablefn45){ref-type="table-fn"}1.0[3](#cetablefn45){ref-type="table-fn"}**Adults (\>20 years)**101.0201.0NANANANA201.0**Hemodialysis patients and other immunocompromised persons**[7](#cetablefn49){ref-type="table-fn"}\<20 years[7](#cetablefn49){ref-type="table-fn"}50.5100.5NANANANANANA**≥20 years**40[8](#cetablefn50){ref-type="table-fn"}1040[9](#cetablefn51){ref-type="table-fn"}2.0NANANANANANA[^35][^36][^37][^38][^39][^40][^41][^42][^43]Table 2-10Dosage and schedule for typhoid fever vaccinationVaccinationAge (Years)Dose/Mode of AdministrationNo. of DosesDosing IntervalBoosting Interval**Oral, live, attenuated Ty21a vaccine (Vivotif)**Primary series≥61 capsule,[1](#cetablefn52){ref-type="table-fn"} oral448 hoursNot applicableBooster≥61 capsule,[1](#cetablefn52){ref-type="table-fn"} oral448 hoursEvery 5 years**Vi Capsular polysaccharide vaccine (Typhim Vi)**Primary series≥20.50 mL, intramuscular1Not applicableNot applicableBooster≥20.50 mL, intramuscular1Not applicableEvery 2 years[^44]Table 2-18Pre-exposure immunization for rabies[1](#cetablefn69){ref-type="table-fn"}VaccineDose (Ml)No. of DosesSchedule (Days)RouteHDCV1.030, 7, and 21 or 28IntramuscularPCEC1.030, 7, and 21 or 28Intramuscular[^45]Table 2-19Postexposure immunization for rabies[1](#cetablefn70){ref-type="table-fn"}Immunization StatusVaccine/ProductDoseNo. of DosesSchedule (Days)RouteNot previously immunizedRIG plus20 IU/kg body weight10Infiltrated at bite site (if possible possible); remainder intramuscularHDCV or PCEC1.0 mL50, 3, 7, 14, 28IntramuscularPreviously immunized[2](#cetablefn71){ref-type="table-fn"},[3](#cetablefn72){ref-type="table-fn"}HDCV or PCEC1.0 mL20, 3Intramuscular[^46][^47][^48][^49]Table 2-21Summary guide to tetanus prophylaxis in routine wound management[1](#cetablefn73){ref-type="table-fn"}History of Tetanus Immunization (Doses)Clean, Minor WoundsAll Other WoundsTd[2](#cetablefn74){ref-type="table-fn"}TIGTd[2](#cetablefn74){ref-type="table-fn"}TIGUncertain or \<3 dosesYesNoYesYes3 or more dosesNo[3](#cetablefn75){ref-type="table-fn"}NoNo[4](#cetablefn76){ref-type="table-fn"}No[^50][^51][^52][^53] Spacing of Immunobiologics {#cesec4} -------------------------- ### Simultaneous Administration {#cesec5} All commonly used vaccines can safely and effectively be given simultaneously (i.e., on the same day) at separate sites without impairing antibody responses or increasing rates of adverse reactions. This knowledge is particularly helpful for international travelers, for whom exposure to several infectious diseases might be imminent. Simultaneous administration of all indicated vaccines is encouraged for persons who are the recommended age to receive these vaccines and for whom no contraindications exist. If not administered on the same day, an inactivated vaccine may be given at any time before or after a different inactivated vaccine or a live-virus vaccine. The immune response to an injected or intranasal live-virus vaccine (e.g., measles, mumps and rubella (MMR); varicella; yellow fever; or live attenuated influenza vaccine) might be impaired if administered within 28 days of another live-virus vaccine. Whenever possible, injected live-virus vaccines administered on different days should be given at least 28 days apart. If two injected or intranasal live-virus vaccines are not administered on the same day but less than 28 days apart, the second vaccine should be readministered at least 4 weeks after the first vaccine was administered. Live-virus vaccines can interfere with the response to tuberculin testing. Tuberculin testing, if otherwise indicated, can be done either on the day that live-virus vaccines are administered or 4--6 weeks later. Tuberculin skin testing is not a prerequisite for administration of any vaccine. ### Missed Doses and Boosters {#cesec6} Travelers may forget to return for a follow-up dose of vaccine or booster at the specified time. Occasionally the demand for a vaccine may exceed its supply, and providers may have difficulty obtaining vaccines. (Information on vaccine shortages and recommendations can be found on the CDC Vaccines and Immunization website at [www.cdc.gov/vaccines/vac-gen/shortages/default.htm](http://www.cdc.gov/vaccines/vac-gen/shortages/default.htm){#interref10}.) It is unnecessary in these cases to restart the interrupted series or to add any extra doses except for oral typhoid. The next scheduled dose should be given when the patient presents. (There are no data for interrupted dosing with oral typhoid vaccine; thus, a travel medicine specialist should be consulted.) Some vaccines require periodic booster doses to maintain protection ([Table 2-1](#cetable1){ref-type="table"}). ### Antibody-Containing Blood Products {#cesec7} When MMR and varicella vaccines are given shortly before, simultaneously with, or after an antibody-containing blood product, such as immune globulin (IG) or a blood transfusion, response to the vaccine can be diminished. Antibody-containing blood products from the United States do not interfere with the immune response to yellow fever vaccine and are not believed to interfere with the response to live attenuated influenza vaccine or rotavirus vaccine. The duration of inhibition of MMR and varicella vaccines is related to the dose of IG in the product. MMR or its components and varicella vaccines either should be administered at least 2 weeks before receipt of a blood product or should be delayed 3--11 months after receipt of the blood product, depending on the vaccine ([Table 2-2](#cetable2){ref-type="table"}). Immunoglobulin (IG) administration may become necessary for another indication after MMR or its individual components or varicella vaccines have been given. In such a situation, the IG may interfere with the immune response to the MMR or varicella vaccines. Vaccine virus replication and stimulation of immunity usually occur 2--3 weeks after vaccination. If the interval between administration of one of these vaccines and the subsequent administration of an IG preparation is 14 days or more, the vaccine need not be readministered. If the interval is less than 14 days, the vaccine should be readministered after the interval shown in [Table 2-2](#cetable2){ref-type="table"}, unless serologic testing indicates that antibodies have been produced. If administration of IG becomes necessary, MMR or its components or varicella vaccines can be administered simultaneously with IG, with the recognition that vaccine-induced immunity can be compromised. The vaccine should be administered at a body site different from that chosen for the IG injection. Vaccination should be repeated after the interval noted in [Table 2-2](#cetable2){ref-type="table"}, unless serologic testing indicates antibodies have been produced. When IG is given with the first dose of hepatitis A vaccine, the proportion of recipients who develop a protective level of antibody is not affected, but antibody concentrations are lower. Because the final concentrations of antibody are many times higher than those considered protective, this reduced immunogenicity is not expected to be clinically important. IG preparations interact minimally with other inactivated vaccines and toxoids. Other inactivated vaccines may be given simultaneously or at any time interval after or before an antibody-containing blood product is used. However, such vaccines should be administered at different sites from the IG. Vaccination of Persons with Acute Illnesses {#cesec8} ------------------------------------------- Every opportunity should be taken to provide appropriate vaccinations. The decision to delay vaccination because of a current or recent acute illness depends on the severity of the symptoms and their cause. Although a moderate or severe acute illness is sufficient reason to postpone vaccination, minor illnesses (e.g., diarrhea, mild upper respiratory infection with or without low-grade fever, other low-grade febrile illness) are not contraindications to vaccination. Persons with moderate or severe acute illness, with or without fever, should be vaccinated as soon as the condition has improved. This precaution is to avoid superimposing adverse effects from the vaccine on underlying illness or mistakenly attributing a manifestation of underlying illness to the vaccine. Antimicrobial therapy is not a contraindication to vaccination, with three exceptions. Antibacterial agents may interfere with the response to oral typhoid vaccine. Antiviral agents active against herpesviruses (e.g., acyclovir) may interfere with the response to varicella-containing vaccines (varicella, MMRV, zoster). Antiviral agents active against influenza virus (e.g., zanamivir, oseltamivir) may interfere with the response to live attenuated influenza vaccine. A physical examination or temperature measurement is not a prerequisite for vaccinating a person who appears to be in good health. Asking if a person is ill, postponing a vaccination for someone with moderate or severe acute illness, and vaccinating someone without contraindications are appropriate procedures for clinic immunizations. Altered Immunocompetence {#cesec9} ------------------------ Altered immunocompetence is a general term that is often used interchangeably with the terms immunosuppression and immunodeficiency. It can be caused either by a disease (e.g., leukemia, HIV infection) or by drugs or other therapies (e.g., cancer chemotherapy, prolonged high dose corticosteroids). It can also include conditions such as asplenia and chronic renal disease. Determination of altered immunocompetence is important because the incidence or severity of some vaccine-preventable diseases is higher in persons with altered immunocompetence. Therefore, certain vaccines (e.g., inactivated influenza vaccine, pneumococcal vaccines) are recommended specifically for persons with these diseases. Inactivated vaccine may be safely administered to a person with altered immunocompetence, although response to the vaccine may be suboptimal. The vaccine may need to be repeated after immune function has improved. Persons with altered immunocompetence may be at increased risk for an adverse reaction following administration of live attenuated vaccines because of reduced ability to mount an effective immune response. Live vaccines should generally be deferred until immune function has improved. This is particularly important when planning to give yellow fever vaccine (see the [Yellow Fever](#subchapter6){ref-type="sec"} section later in this chapter). MMR and varicella vaccines are recommended for persons with mild or moderate immunosuppression. For an in-depth discussion, see The Immunocompromised Traveler section in Chapter 8. Vaccination Scheduling for Last-Minute Travelers {#cesec10} ------------------------------------------------ As noted in the Simultaneous Administration section, most vaccine products can be given during one visit for persons anticipating imminent travel. Unless the vaccines given are booster doses of those typically given during childhood, vaccines may require a month or more to induce a sufficient immune response, depending on the vaccine and the number of doses in the series. Some vaccines require more than one dose for best protection. Recommended spacing should be maintained between doses ([Table 2-3](#cetable3){ref-type="table"}). Doses given at less than minimum intervals can lessen the antibody response. Administration of a vaccine earlier than the recommended minimum age or at an interval shorter than the recommended minimum is discouraged. [Table 2-3](#cetable3){ref-type="table"} lists the minimum age and minimum interval between doses for vaccines routinely recommended in the United States. Because some travelers visit their health-care providers without ample time for administration of the vaccine doses recommended for optimal protection against certain diseases, studies have been performed and others are ongoing to determine whether accelerated scheduling is adequate. This concern is primarily the case for hepatitis B vaccine or the combined hepatitis A and B vaccine. An accelerated schedule for combined hepatitis A and hepatitis B vaccine has been approved by the U.S. Food and Drug Administration (FDA). It is unclear what level of protection any given traveler will have if a full series of multidose vaccination is not completed. Allergy to Vaccine Components {#cesec11} ----------------------------- Vaccine components can cause allergic reactions in some recipients. These reactions can be local or systemic and can include anaphylaxis or anaphylactic-like responses. The vaccine components responsible can include the vaccine antigen, animal proteins, antibiotics, preservatives (e.g., thimerosal), or stabilizers (e.g., gelatin). The most common animal protein allergen is egg protein in vaccines prepared by using embryonated chicken eggs (influenza and yellow fever vaccines). Generally, persons who can eat eggs or egg products safely may receive these vaccines, while those with histories of anaphylactic allergy (e.g., hives, swelling of the mouth and throat, difficulty breathing, hypotension, shock) to eggs or egg proteins ordinarily should not. Screening persons by asking whether they can eat eggs without adverse effects is a reasonable way to identify those who might be at risk from receiving yellow fever and influenza vaccines. Recent studies have indicated that other components in vaccines in addition to egg proteins (e.g., gelatin) may cause allergic reactions, including anaphylaxis in rare instances. Protocols have been developed for testing and vaccinating persons with anaphylactic reactions to egg ingestion. Some vaccines contain a preservative or trace amounts of antibiotics to which people might be allergic. Those administering the vaccine(s) should carefully review the information provided in the package insert before deciding if the rare person with such an allergy should receive the vaccine. No currently recommended vaccine contains penicillin or penicillin derivatives. Some vaccines (e.g., MMR and its individual component vaccines, inactivated polio vaccine \[IPV\], varicella, rabies) contain trace amounts of neomycin or other antibiotics; the amount is less than would normally be used for the skin test to determine hypersensitivity. However, persons who have experienced anaphylactic reactions to this antibiotic generally should not receive these vaccines. Most often, neomycin allergy is a contact dermatitis---a manifestation of a delayed-type (cell-mediated) immune response rather than anaphylaxis. A history of delayed-type reactions to neomycin is not a contraindication to receiving these vaccines. Thimerosal, an organic mercurial compound in use since the 1930s, has been added to certain immunobiologic products as a preservative. Thimerosal is present at preservative concentrations (trace quantities) in multidose vials of some brands of inactivated influenza vaccine, pediatric DT, single-antigen tetanus toxoid, meningococcal polysaccharide vaccine, and Japanese encephalitis vaccine. Receiving thimerosal-containing vaccines has been postulated to lead to induction of allergy. However, there is limited scientific evidence for this assertion. Allergy to thimerosal usually consists of local delayed-type hypersensitivity reactions. Thimerosal elicits positive delayed-type hypersensitivity patch tests in 1%--18% of persons tested, but these tests have limited or no clinical relevance. The majority of persons do not experience reactions to thimerosal administered as a component of vaccines, even when patch or intradermal tests for thimerosal indicate hypersensitivity. A localized or delayed-type hypersensitivity reaction to thimerosal is not a contraindication to receipt of a vaccine that contains thimerosal. Since mid-2001, vaccines routinely recommended for infants have been manufactured without thimerosal as a preservative. Additional information about thimerosal and the thimerosal content of vaccines is available on the FDA website at [www.fda.gov/cber/vaccine/thimerosal.htm](http://www.fda.gov/cber/vaccine/thimerosal.htm){#interref11}. Reporting Adverse Events Following Immunization {#cesec12} ----------------------------------------------- Modern vaccines are extremely safe and effective. Benefits and risks are associated with the use of all immunobiologics---no vaccine is completely effective or completely free of side effects. Adverse events following immunization have been reported with all vaccines, ranging from frequent, minor, local reactions to extremely rare, severe, systemic illness, such as that associated with yellow fever vaccine (see [Yellow Fever](#subchapter6){ref-type="sec"} section later in this chapter). Side effects and adverse events following specific vaccines and toxoids are discussed in detail in each ACIP statement. Health-care providers are required by law to report selected adverse events occurring after vaccination with tetanus vaccine in any combination; pertussis in any combination; measles, mumps or rubella alone or in any combination, oral polio vaccine (OPV), IPV, hepatitis B; varicella; *Haemophilus influenzae* type b (conjugate); pneumococcal conjugate; and rotavirus vaccines. In addition, CDC strongly recommends that all vaccine adverse events be reported to the Vaccine Adverse Event Reporting System (VAERS), even if a causal relation to vaccination is not certain. VAERS reporting forms and information are available electronically at [www.vaers.hhs.gov](http://www.vaers.hhs.gov){#interref12} or may be requested by telephone: 800-822-7967. Health-care providers are encouraged to report electronically at <https://secure.vaers.org/VaersDataEntryintro.htm>. Injection Route and Injection Site {#cesec13} ---------------------------------- Injectable vaccines are administered by intramuscular and subcutaneous routes. The method of administration of injectable vaccines depends in part on the presence of an adjuvant in some vaccines. The term adjuvant refers to a vaccine component distinct from the antigen, which enhances the immune response to the antigen. Vaccines containing an adjuvant (i.e., DTaP, DT, human papillomavirus, Td, Tdap, pneumococcal conjugate, Hib, hepatitis A, hepatitis B) should be injected into a muscle mass because administration subcutaneously or intradermally can cause local irritation, induration, skin discoloration, inflammation, and granuloma formation. Routes of administration are recommended by the manufacturer for each immunobiologic. Deviation from the recommended route of administration may reduce vaccine efficacy or increase local adverse reactions. Detailed recommendations on the appropriate route and site for all vaccines have been published in ACIP recommendations; a compiled list of these publications is available on the CDC website at [www.cdc.gov/vaccines/pubs/ACIP-list.htm](http://www.cdc.gov/vaccines/pubs/ACIP-list.htm){#interref14} (also see Appendix C: Travel Vaccine Summary Table).   {#cesec14} = Travel-Related Vaccine-Preventable Diseases HEPATITIS A {#subchapter3} =========== Wiersma Steven T. Infectious Agent {#cesec15} ---------------- Hepatitis A virus (HAV), a 27-nm RNA virus classified as a picornavirus. Mode of Transmission {#cesec16} -------------------- •Transmission can occur through direct person-to-person contact; through exposure to contaminated water, ice, or shellfish harvested from sewage-contaminated water; or from fruits, vegetables, or other foods that are eaten uncooked and that were contaminated during harvesting or subsequent handling.•HAV is shed in the feces of persons with HAV infection. The virus reaches peak levels the week or two before onset of symptoms and diminishes rapidly after liver dysfunction or symptoms appear, which is concurrent with the appearance of circulating antibodies to HAV. Infants and children, however, may shed virus for up to 6 months following infection. Occurrence {#cesec17} ---------- •Worldwide, geographic areas can be characterized by high, intermediate, or low levels of endemicity ([Map 2-1](#f4){ref-type="fig"} ). Levels of endemicity are related to hygienic and sanitary conditions in the geographic areas.Map 2-1Prevalence of antibody to hepatitis A virus, 2006.[1](#fn1){ref-type="fn"}*(See*[note 1](#fn1){ref-type="fn"}*on opposite page.)*•HAV infection is common (high or intermediate endemicity) throughout the developing world, where infections most frequently are acquired during early childhood and usually are asymptomatic or mild.•In areas of high endemicity, adults are usually immune and epidemics of hepatitis A are uncommon.•In developed countries, HAV infection is less common (low endemicity), but community-wide outbreaks may occur.•[Map 2-1](#f4){ref-type="fig"} indicates the seroprevalence of antibody to HAV (total anti-HAV) as measured in selected cross-sectional studies among each country\'s residents. The seroprevalence of anti-HAV provides an estimate of the endemicity of HAV infections, including asymptomatic infections, within a population. Box 2-1Hepatitis A pre-travel case study**Case study:** You are a travel medicine professional preparing a group for travel to an eastern European country that is shown in [Maps 2-1](#f4){ref-type="fig"} to have an intermediate prevalence of antibody to hepatitis A virus (anti-HAV). You have recommended that all members of the group take precautions to prevent hepatitis A. The leader of the group questions your advice on the basis of an editorial from a major newspaper written by the ambassador to the U.S. from this country. The editorial claims that her country is stigmatized in the U.S. and cites the example of the Yellow Book, which indicates that travelers from the U.S. to this country are at risk for hepatitis A. The editorial states that cases of hepatitis A are at very low levels in this country and seem to be declining over time. How would you respond to the members of the group?Points to consider in your response:•In most intermediate and high anti-HAV-endemic countries, many long-term residents are infected as children, at a time when they may not get symptoms. Cases of hepatitis A in the resident population will be very low; however, travelers from low endemic settings such as the United States are at risk for HAV infection and should be protected.•The determination of risk is based on CDC estimates of prevalence of anti-HAV, a marker of previous HAV infection. This country-level estimate is based on limited data and might not reflect the current prevalence.•Prevention of hepatitis A in travelers with vaccination should be used liberally because the vaccine is safe and effective and will give long-term benefits that go beyond the risk posed by any specific trip. Risk for Travelers {#cesec19} ------------------ •Hepatitis A is one of the most common vaccine-preventable infections acquired during travel.•In 2006 in the United States, among cases for which information regarding exposures during the incubation period was collected, the most frequently identified risk factor for hepatitis A was international travel (reported by 15% of case-patients overall).•As in previous years, most travel-related cases (72%) were associated with travel to Mexico and Central/South America. As HAV transmission in the United States has decreased, cases among travelers to countries in which hepatitis is endemic have accounted for an increased proportion of all cases.•The risk of acquiring HAV infection for U.S. residents traveling abroad varies with living conditions, length of stay, and the incidence of HAV infection in the area visited. For travelers to other countries, risk for infection increases with duration of travel and is highest for those who live in or visit rural areas, trek in back-country areas, or frequently eat or drink in settings of poor sanitation.•Nevertheless, many cases of travel-related hepatitis A occur in travelers to developing countries with "standard" tourist itineraries, accommodations, and food consumption behaviors. Clinical Presentation {#cesec20} --------------------- •HAV infection may be asymptomatic, or its clinical manifestations may range in severity from a mild illness lasting 1--2 weeks to a severely disabling disease lasting several months.•Clinical manifestations of hepatitis A often include the abrupt onset of fever, malaise, anorexia, nausea, and abdominal discomfort, followed within a few days by jaundice.•The incubation period for hepatitis A averages 28 days (range: 15--50 days).•The likelihood of having symptoms with HAV infection is related to the infected person\'s age. In children \<6 years of age, most (70%) infections are asymptomatic; if illness does occur, its duration is usually less than 2 months.•No chronic or long-term infection is associated with hepatitis A, but 10% of infected persons will have prolonged or relapsing symptoms over a 6- to 9-month period.•The overall case--fatality rate among cases reported to CDC is 0.3%; however, the rate is 1.8% among adults \>50 years of age. Diagnosis {#cesec21} --------- •Demonstration of IgM antibodies against hepatitis A virus (IgM anti-HAV) in the serum of acutely or recently ill patients establishes the diagnosis.•IgM anti-HAV becomes detectable 5--10 days after exposure. A fourfold or greater rise in specific antibodies in paired sera, detected by commercially available EIA, also establishes the diagnosis.•If laboratory tests are not available, epidemiologic evidence may provide support for the diagnosis in a clinically compatible case.•HAV RNA can be detected in blood and stools of most persons during the acute phase of infection through nucleic acid amplification methods, but these are not generally used for diagnostic purposes. Treatment {#cesec22} --------- No specific treatment is available for persons with hepatitis A. Treatment is supportive. Preventive Measures for Travelers {#cesec23} --------------------------------- Health-care providers should administer hepatitis A vaccination for persons traveling for any purpose, frequency, or duration to countries that have high or intermediate endemicity of HAV infection. Providers may also consider its administration to persons for travel to any destination. ### Vaccine and Immune Globulin {#cesec24} #### Monovalent Vaccines {#cesec25} •Two monovalent hepatitis A vaccines are currently licensed in the United States for persons at least 12 months of age:○HAVRIX, manufactured by GlaxoSmithKline ([Table 2-4](#cetable4){ref-type="table"}), and○VAQTA manufactured by Merck & Co., Inc. ([Table 2-5](#cetable5){ref-type="table"}).•Both vaccines are made of inactivated HAV adsorbed to aluminum hydroxide as an adjuvant. HAVRIX is prepared with 2-phenoxyethanol as a preservative, while VAQTA is formulated without a preservative.•All hepatitis A vaccines should be administered intramuscularly in the deltoid muscle. #### Combination Vaccine {#cesec26} •TWINRIX, manufactured by GlaxoSmithKline, is a combined hepatitis A and hepatitis B vaccine licensed for persons Δ18 years of age, containing 720 EL.U. of hepatitis A antigen (50% of the HAVRIX adult dose) and 20 mg of recombinant hepatitis B surface antigen protein (the same as the ENGERIX-B adult dose) ([Table 2-6](#cetable6){ref-type="table"}).•Primary immunization consists of three doses, given on a 0-, 1-, and 6-month schedule, the same schedule as that commonly used for monovalent hepatitis B vaccine.•TWINRIX contains aluminum phosphate and aluminum hydroxide as adjuvants and 2-phenoxyethanol as a preservative.•An accelerated schedule of TWINRIX (i.e., doses at days 0, 7, and 21) for travelers has been approved by the FDA. A booster dose should be given at 1 year.•The immunogenicity of TWINRIX is equivalent to that of the monovalent hepatitis vaccines when tested after completion of the licensed schedule. #### Vaccination of Travelers {#cesec27} •All susceptible persons traveling to or working in countries that have high or intermediate hepatitis A endemicity should be vaccinated or receive IG before departure. Hepatitis A vaccine at the age-appropriate dose is preferred to IG. The first dose of hepatitis A vaccine should be administered as soon as travel to countries with high or intermediate endemicity is considered.•One dose of monovalent hepatitis A vaccine administered at any time before departure can provide adequate protection for most healthy persons \<40 years of age.•Completion of the vaccine series according to the licensed schedule is necessary for long-term protection.•Many persons will have detectable anti-HAV in response to the monovalent vaccine by 2 weeks after the first vaccine dose. The proportion of persons who develop a detectable antibody response at 2 weeks may be lower when smaller vaccine dosages are used, such as with the use of TWINRIX.•For optimal protection, older adults, immunocompromised persons, and persons with chronic liver disease or other chronic medical conditions planning to depart to an area in \<2 weeks should receive the initial dose of vaccine along with IG (0.02 mL/kg) at a separate anatomic injection site.•Travelers who receive hepatitis A vaccine less than 2 weeks before traveling to an endemic area and who do not receive IG (either by choice or because of lack of availability) will be at lower risk for infection than those who do not receive hepatitis A vaccine or IG.•Although vaccination of an immune traveler is not contraindicated and does not increase the risk for adverse effects, screening for total anti-HAV before travel can be useful in some circumstances to determine susceptibility and eliminate unnecessary vaccination or IG prophylaxis of immune travelers. Such serologic screening for susceptibility might be indicated for adult travelers who are \>40 years of age and those born in areas of the world with intermediate or high endemicity who are likely to have had prior HAV infection, if the cost of screening (laboratory and office visit) is less than the cost of vaccination or IG prophylaxis and if testing will not delay vaccination and interfere with timely receipt of vaccine or IG before travel. Postvaccination testing for serologic response is not indicated.•Travelers who are \<12 months of age, are allergic to a vaccine component, or who otherwise elect not to receive vaccine should receive a single dose of IG (0.02 mL/kg), which provides effective protection against HAV infection for up to 3 months ([Table 2-7](#cetable7){ref-type="table"}).•Those who do not receive vaccination and plan to travel for \>3 months should receive an IG dose of 0.06 mL/kg, which must be repeated if the duration of travel is \>5 months.•In addition, health-care providers should be alert to opportunities to provide vaccination for all travelers whose plans might include travel at some time in the future to an area of high or intermediate endemicity, including those whose current medical evaluation is for travel to an area where hepatitis A vaccination is not currently recommended.•Those who refuse vaccine and IG should be advised to closely adhere to prevention tips listed below. #### Other Vaccine Considerations {#cesec28} •Using the vaccines according to the licensed schedules is preferable. However, an interrupted series does not need to be restarted.•Given their similar immunogenicity, a series that has been started with one brand of monovalent vaccine (i.e., HAVRIX or VAQTA) may be completed with the other brand.•Hepatitis A vaccine may be administered at the same time as IG or other commonly used vaccines for travelers, at different injection sites.•In adults and children who have completed the vaccine series, anti-HAV has been shown to persist for at least 5--12 years after vaccination. Results of mathematical models indicate that, after completion of the vaccination series, anti-HAV will likely persist for 20 years or more. For children and adults who complete the primary series, booster doses of vaccine are not recommended. #### Vaccine Safety and Adverse Reactions {#cesec29} •Among adults, the most frequently reported side effects occurring 3--5 days after a vaccine dose are tenderness or pain at the injection site (53%--56%) or headache (14%--16%).•Among children, the most common side effects reported are pain or tenderness at the injection site (15%--19%), feeding problems (8% in one study), or headache (4% in one study).•No serious adverse events in children or adults that could be definitively attributed to the vaccine or to increases in serious adverse events among vaccinated persons compared with baseline rates have been identified.•IG for intramuscular administration prepared in the United States has few side effects (primarily soreness at the injection site) and has never been shown to transmit infectious agents (hepatitis B virus, hepatitis C virus \[HCV\], or HIV).•Since December 1994, all IG products commercially available in the United States have had to undergo a viral inactivation procedure or be negative for HCV RNA before release. #### Precautions and Contraindications {#cesec30} •These vaccines should not be administered to travelers with a history of hypersensitivity to any vaccine component.•HAVRIX or TWINRIX should not be administered to travelers with a history of hypersensitivity reactions to the preservative 2-phenoxyethanol.•TWINRIX should not be administered to persons with a history of hypersensitivity to yeast.•Because hepatitis A vaccine consists of inactivated virus and hepatitis B vaccine consists of a recombinant protein, no special precautions need to be taken for vaccination of immunocompromised travelers. #### Pregnancy {#cesec31} •The safety of hepatitis A vaccine for pregnant women has not been determined.•However, because hepatitis A vaccine is produced from inactivated HAV, the theoretical risk to either the pregnant woman or the developing fetus is thought to be very low.•The risk of vaccination should be weighed against the risk of hepatitis A in female travelers who might be at high risk for exposure to HAV.•Pregnancy is not a contraindication to using IG. ### Other Prevention Tips {#cesec32} •Boiling or cooking food and beverage items for at least 1 minute to 185° F (85° C) inactivates HAV. Foods and beverages heated to this temperature and for this length of time cannot serve as vehicles for HAV infection unless they become contaminated after heating.•Adequate chlorination of water as recommended in the United States will inactivate HAV.•Travelers should be advised that, to minimize their risk of hepatitis A and other enteric diseases in developing countries, they should avoid potentially contaminated water or food.•Travelers should also be advised to avoid drinking beverages (with or without ice) of unknown purity, eating uncooked shellfish, and eating uncooked fruits or vegetables that are not peeled or prepared by the traveler personally. HEPATITIS B {#subchapter4} =========== Chaves Sandra S. Infectious Agent {#cesec33} ---------------- Hepatitis B is caused by the hepatitis B virus (HBV), a small, circular, partially double-stranded DNA molecule in the *Hepadnaviridae* family. Mode of Transmission {#cesec34} -------------------- HBV is transmitted through activities that involve contact with blood or blood-derived fluids. Such activities include the following:•Unprotected sex with an HBV-infected partner•Shared needles used for injection of illegal drugs•Shared glucose-monitoring equipment•Work in health-care fields (e.g., medical, dental, laboratory) that entails direct exposure to potentially infected human blood•Transfusions with blood or blood products that have not been screened for HBV•Dental, medical, or cosmetic (e.g., tattooing, body piercing) procedures with needles or other equipment that are contaminated with HBV In addition, open skin lesions, such as those due to impetigo, scabies, or scratched insect bites, can play a role in HBV transmission if direct exposure to wound exudates from HBV-infected persons occurs. Occurrence {#cesec35} ---------- •The prevalence of chronic HBV infection is low (\<2%) in the general population in Northern and Western Europe, North America, Australia, New Zealand, Mexico, and southern South America ([Map 2-2](#f5){ref-type="fig"} ).Map 2-2Prevalence of chronic infection with hepatitis B virus, 2006.•The prevalence of chronic HBV infection is intermediate (2%--7%) in South, Central, and Southwest Asia, Israel, Japan, Eastern and Southern Europe, Russia, most areas surrounding the Amazon River basin, Honduras, and Guatemala (see [Maps 2-2](#f5){ref-type="fig"}).•The prevalence of chronic HBV infection is high (Δ8%) in all socioeconomic groups in: all of Africa; Southeast Asia, including China, Korea, Indonesia, and the Philippines; the Middle East, except Israel; South and Western Pacific islands; the interior Amazon River basin; and certain parts of the Caribbean (Haiti and the Dominican Republic) (see [Maps 2-2](#f5){ref-type="fig"}). Risk for Travelers {#cesec36} ------------------ There are no data with which to assess the risk for HBV infection among U.S. travelers. The risk for HBV infection for international travelers is considered generally low, except for travelers to countries where the prevalence of chronic HBV infection is intermediate or high. Some travelers, such as adventure travelers, Peace Corps volunteers, missionaries, and military personnel, may be at increased risk for infection. Situations or activities that may carry increased risk for HBV infection for travelers while overseas include the following:•An injury or illness that requires invasive medical attention (e.g., injection, IV drip, transfusion, stitching)•Dental treatment•Unprotected sexual contact•Sharing illegal drug injection equipment•Skin-perforation practices (e.g., tattooing, ear piercing, acupuncture)•Cosmetic practices with risk for skin perforation (e.g., manicure/pedicure)•Sharing personal grooming items (e.g., earrings, toothbrush, razor) Clinical Presentation {#cesec37} --------------------- •Incubation period of hepatitis B is typically 90 days (range: 60--150 days) from exposure to onset of jaundice.•Constitutional symptoms such as malaise and anorexia may precede jaundice by 1--2 weeks.•Clinical symptoms and signs include nausea, vomiting, abdominal pain, and jaundice.•Skin rashes, joint pain, and arthritis may occur.•Infants, children \<5 years of age, and immunosuppressed adults with newly acquired HBV infection typically are asymptomatic.•Infected persons Δ5 years of age, including immunocompetant adults, 30%--50% have initial clinical signs or symptoms.•The case--fatality rate of acute hepatitis B is approximately 1%.•Acute HBV infection causes chronic (long-term) infection in 30%--90% of persons infected as infants or young children and in \<5% of adolescents and adults.•Chronic infection can lead to chronic liver disease, liver scarring (cirrhosis), and liver cancer. Diagnosis {#cesec38} --------- At least one serologic marker is present during each of the different phases of HBV infection. The serologic markers are typically used to differentiate between acute, resolving, and chronic infection ([Table 2-8](#cetable8){ref-type="table"} ).Table 2-8Typical interpretation of serologic test results for hepatitis B virus infectionSerologic MarkerInterpretationHBsAg[1](#cetablefn34){ref-type="table-fn"}Total anti-HBc[2](#cetablefn35){ref-type="table-fn"}IgM[3](#cetablefn36){ref-type="table-fn"} anti-HBcAnti-HBs[4](#cetablefn37){ref-type="table-fn"}--[5](#cetablefn38){ref-type="table-fn"}------Never infected+[6](#cetablefn39){ref-type="table-fn"},[7](#cetablefn40){ref-type="table-fn"}------Early acute infection; transient (up to 18 days) after vaccination+++--Acute infection--+++ or −Acute resolving infection--+--+Recovered from past infection and immune++----Chronic infection--+----False-positive (i.e., susceptible); past infection; "low-level" chronic infection;[8](#cetablefn41){ref-type="table-fn"} or passive transfer of anti-HBc to infant born to HBsAg-positive mother------+Immune if concentration is ≥10 mIU/mL after vaccine series completion;[9](#cetablefn42){ref-type="table-fn"} passive transfer after hepatitis B immune globulin administration[^54][^55][^56][^57][^58][^59][^60][^61][^62]From CDC. MMWR Recomm Rep. 2006; 55(RR-16):1--25. Treatment {#cesec39} --------- No specific treatment is available for acute illness caused by hepatitis B. Antiviral drugs are approved for the treatment of chronic hepatitis B. Preventive Measures for Travelers {#cesec40} --------------------------------- ### Vaccine {#cesec41} •Hepatitis B vaccination should be administered to all unvaccinated persons traveling to areas with intermediate to high levels of endemic HBV transmission (i.e., with hepatitis B surface antigen \[HBsAg\] prevalence Δ2%).•Hepatitis B vaccination is currently recommended for all U.S. residents who work in health-care fields (e.g., medical, dental, laboratory) that involve potential exposure to human blood.•All unvaccinated U.S. children and adolescents (\<19 years of age) should receive hepatitis B vaccine.•Unvaccinated persons who have indications for hepatitis B vaccination independent of travel should be vaccinated (e.g., men who have sex with men, injection drug users, anyone who has recently had a sexually transmitted disease or has had more than one sex partner in the previous 6 months). #### Vaccine Dose and Administration {#cesec42} •The vaccine is usually administered as a three-dose series on a 0-, 1-, and 6-month schedule (see [Table 2-9](#cetable9){ref-type="table"}). The second dose should be given 1 month after the first dose; the third dose should be given at least 2 months after the second dose and at least 4 months after the first dose.•Alternatively, the vaccine ENGERIX-B, manufactured by GlaxoSmithKline, is also approved for administration on a four-dose schedule at 0, 1, 2, and 12 months.•There is also a two-dose schedule for RECOMBIVAX HB, a vaccine produced by Merck & Co., Inc., which has been licensed for children and adolescents 11--15 years of age. Using the two-dose schedule, the adult dose of RECOMBIVAX HB is administered, with the second dose given 4--6 months after the first dose.•A three-dose series that has been started with one brand of vaccine may be completed with the other brand.•TWINRIX, manufactured by GlaxoSmithKline, is a combined hepatitis A and hepatitis B vaccine licensed for persons 18 years of age or older. Primary immunization consists of three doses, given on a 0-, 1-, and 6-month schedule. #### Special Situations {#cesec43} •Ideally, vaccination should begin at least 6 months before travel so the full vaccine series can be completed before departure. Because some protection is provided by one or two doses, the vaccine series should be initiated, if indicated, even if it cannot be completed before departure. Optimal protection, however, is not conferred until after the final vaccine dose. Travelers should be advised to return for completion of the vaccine series.•An accelerated vaccine schedule could be used for those traveling to endemic areas at short notice and facing imminent exposure because of behavioral risks or to emergency responders to disaster areas. The monovalent hepatitis B vaccines can be used at 0, 7, and 14 days. If an accelerated schedule is used, the patient should receive a booster dose at least 6 months after the start of the series to promote long-term immunity.•An accelerated vaccine schedule with TWINRIX (hepatitis A and hepatitis B vaccine) can also be used (doses at 0, 7, and 21--30 days). In this situation, a booster dose should be given at 12 months to promote long-term immunity.•For children and adults whose immune status is normal, booster doses of vaccine are not recommended. Serologic testing to assess antibody levels is not necessary for most vaccinees (see the Vaccine Recommendations for Infants and Children section in Chapter 7). #### Vaccine Safety and Adverse Reactions {#cesec44} •Hepatitis B vaccines have been shown to be safe for persons of all ages. Pain at the injection site (3%--29%) and elevated temperature higher than 37.7° C (99.9° F) (1%--6%) are the most frequently reported side effects among vaccine recipients.•These vaccines should not be administered to persons with a history of hypersensitivity to any vaccine component, including yeast. The vaccine contains a recombinant protein (HBsAg) that is noninfectious. Limited data indicate that there is no apparent risk of adverse events to the developing fetus when hepatitis B vaccine is administered to pregnant women. HBV infection affecting a pregnant woman can result in serious disease for the mother and chronic infection for the newborn. Neither pregnancy nor lactation should be considered a contraindication for vaccination. ### Other Preventive Measures {#cesec45} •As part of the pre-travel education process, all travelers should be given information about the risks for hepatitis B and other bloodborne pathogens from contaminated medical equipment, injection drug use, or sexual activity and informed of prevention measures (see below), including hepatitis B vaccination, that can be used to prevent transmission of HBV.•Regardless of destination, all persons who may engage in practices that put them at risk for HBV infection during travel should receive hepatitis B vaccination if previously unvaccinated.•Any adult seeking protection from HBV infection should be vaccinated. Acknowledgment of a specific risk factor is not a requirement for vaccination.•Behavioral preventive measures for HBV infection are similar to those for HIV infection and AIDS.•When seeking medical or dental care, travelers should be advised to be alert to the use of medical, surgical, and dental equipment that has not been adequately sterilized or disinfected, reuse of contaminated equipment, and unsafe injecting practices (e.g., reuse of disposable needles and syringes).•HBV and other bloodborne pathogens (e.g., HIV and hepatitis C) can be transmitted if tools are not sterile or if the tattoo artist or piercer does not follow other proper infection-control procedures (e.g., washing hands, using latex gloves, and cleaning and disinfecting surfaces and instruments).•Travelers should be advised to consider the health risks in deciding to get a tattoo or body piercing in areas where adequate sterilization or disinfection procedures might not be available or practiced. TYPHOID AND PARATYPHOID FEVER {#subchapter5} ============================= Mintz Eric Infectious Agent {#cesec46} ---------------- Typhoid fever is an acute, life-threatening febrile illness caused by the bacterium *Salmonella enterica* serotype Typhi. Paratyphoid fever is a similar illness caused by *S*. Paratyphi A, B, or C. Mode of Transmission {#cesec47} -------------------- •Humans are the only source. No animal or environmental reservoirs have been identified.•Typhoid and paratyphoid fever are most often acquired through consumption of water or food that have been contaminated by feces of an acutely infected or convalescent individual or a chronic asymptomatic carrier.•Transmission through sexual contact, especially among men who have sex with men, has rarely been documented. Occurrence {#cesec48} ---------- •An estimated 22 million cases of typhoid fever and 200,000 related deaths occur worldwide each year; an additional 6 million cases of paratyphoid fever are estimated to occur annually.•Approximately 400 cases of typhoid fever and 150 cases of paratyphoid fever are reported to CDC each year among persons with onset of illness in the United States, most of whom are recent travelers. Risk for Travelers {#cesec49} ------------------ •Risk is greatest for travelers to South Asia (6 to 30 times higher than all other destinations). Other areas of risk include East and Southeast Asia, Africa, the Caribbean, and Central and South America.•Travelers to South Asia are at highest risk for infections that are nalidixic acid-resistant or multidrug-resistant (i.e., resistant to ampicillin, chloramphenicol, and trimethoprim--sulfamethoxazole).•Travelers who are visiting friends or relatives are at increased risk (see the VFR section in Chapter 8).•Although the risk of acquiring typhoid or paratyphoid fever increases with the duration of stay, travelers have acquired typhoid fever even during visits of less than 1 week to countries where the disease is endemic. Clinical Presentation {#cesec50} --------------------- •The incubation period of typhoid and paratyphoid infections is 6--30 days. The onset of illness is insidious, with gradually increasing fatigue and a fever that increases daily from low-grade to as high as 102° F--104° F (38.5° C--40° C) by the third to fourth day of illness. Headache, malaise, and anorexia are nearly universal. Hepatosplenomegaly can often be detected. A transient, macular rash of rose-colored spots can occasionally be seen on the trunk.•Fever is commonly lowest in the morning, reaching a peak in late afternoon or evening. Untreated, the disease can last for a month. The serious complications of typhoid fever generally occur only after 2--3 weeks of illness, mainly intestinal hemorrhage or perforation, which can be life threatening. Diagnosis {#cesec51} --------- •Infection with typhoid or paratyphoid fever results in a very low-grade septicemia. Blood culture is usually positive in only half the cases. Stool culture is not usually positive during the acute phase of the disease. Bone-marrow culture increases the diagnostic yield to about 80% of cases.•The Widal test is an old serologic assay for detecting IgM and IgG antibodies to the O and H antigens of *Salmonella*. The test is unreliable, but is widely used in developing countries because of its low cost. Newer serologic assays are somewhat more sensitive and specific than the Widal test, but are infrequently available.•Because there is no definitive test for typhoid or paratyphoid fever, the diagnosis often has to be made clinically. The combination of a history of being at risk for infection and a gradual onset of fever that increases in severity over several days should raise suspicion of typhoid or paratyphoid fever. Treatment {#cesec52} --------- •Specific antimicrobial therapy shortens the clinical course of typhoid fever and reduces the risk for death.•Empiric treatment of typhoid or paratyphoid fever in most parts of the world would utilize a fluoroquinolone, most often ciprofloxacin. However, resistance to fluoroquinolones is highest in the Indian subcontinent and increasing in other areas. Injectable third-generation cephalosporins are often the empiric drug of choice when the possibility of fluoroquinolone resistance is high.•Patients treated with an appropriate antibiotic still require 3--5 days to defervesce completely, although the height of the fever decreases each day. Patients may actually feel worse during the time that the fever is starting to go away. If fever does not subside within 5 days, alternative antimicrobial agents or other foci of infection should be considered. Preventive Measures for Travelers {#cesec53} --------------------------------- ### Vaccine {#cesec54} •CDC recommends typhoid vaccine for travelers to areas where there is a recognized increased risk of exposure to *S.* Typhi.•The typhoid vaccines currently available do not offer protection against *S.* Paratyphi infection.•Travelers should be reminded that typhoid immunization is not 100% effective, and typhoid fever could still occur.•Two typhoid vaccines are currently available in the United States.○Oral live, attenuated vaccine (Vivotif Berna vaccine, manufactured from the Ty21a strain of *S.* Typhi by the Swiss Serum and Vaccine Institute)○Vi capsular polysaccharide vaccine (ViCPS) (Typhim Vi, manufactured by Sanofi Pasteur) for intramuscular use•Both vaccines protect 50%--80% of recipients.•[Table 2-10](#cetable10){ref-type="table"} provides information on vaccine dosage, administration, and revaccination. The time required for primary vaccination differs for the two vaccines, as do the lower age limits.•Primary vaccination with oral Ty21a vaccine consists of four capsules, one taken every other day. The capsules should be kept refrigerated (not frozen), and all four doses must be taken to achieve maximum efficacy. Each capsule should be taken with cool liquid no warmer than 37° C (98.6° F), approximately 1 hour before a meal. This regimen should be completed 1 week before potential exposure. The vaccine manufacturer recommends that Ty21a not be administered to infants or children \<6 years of age.•Primary vaccination with ViCPS consists of one 0.5-mL (25-mg) dose administered intramuscularly. One dose of this vaccine should be given at least 2 weeks before expected exposure. The manufacturer does not recommend the vaccine for infants and children \<2 years of age. #### Vaccine Safety and Adverse Reactions {#cesec55} Information on adverse reactions is presented in [Table 2-11](#cetable11){ref-type="table"} . Information is not available on the safety of these vaccines in pregnancy; it is prudent on theoretical grounds to avoid vaccinating pregnant women. Live, attenuated Ty21a vaccine should not be given to immunocompromised travelers, including those infected with HIV. The intramuscular vaccine presents a theoretically safer alternative for this group. The only contraindication to vaccination with ViCPS vaccine is a history of severe local or systemic reactions after a previous dose. Neither of the available vaccines should be given to persons with an acute febrile illness.Table 2-11Common adverse reactions to typhoid fever vaccinesVaccineReactionsFeverHeadacheLocal ReactionsTy21a[1](#cetablefn53){ref-type="table-fn"}0%--5%0%--5%Not applicableVi Capsular polysaccharide0%--1%16%--20%7% erythema or induration 1 cm[^63] #### Precautions and Contraindications {#cesec56} Theoretical concerns have been raised about the immunogenicity of live, attenuated Ty21a vaccine in persons concurrently receiving antimicrobials (including antimalarial chemoprophylaxis), IG, or viral vaccines. The growth of the live Ty21a strain is inhibited in vitro by various antibacterial agents. Vaccination with Ty21a should be delayed for \>72 hours after the administration of any antibacterial agent. Available data do not suggest that simultaneous administration of oral polio or yellow fever vaccine decreases the immunogenicity of Ty21a. If typhoid vaccination is warranted, it should not be delayed because of administration of viral vaccines. Simultaneous administration of Ty21a and IG does not appear to pose a problem. YELLOW FEVER {#subchapter6} ============ Gershman Mark Schroeder Betsy Staples J. Erin Infectious Agent {#cesec57} ---------------- Yellow fever virus (YFV) is a single-stranded RNA virus that belongs to the genus *Flavivirus*. Mode of Transmission {#cesec58} -------------------- •Vector-borne transmission occurs via the bite of an infected mosquito, primarily *Aedes* or *Haemagogus* spp.•Nonhuman and human primates are the main reservoirs of the virus, with anthroponotic (human-to-vector-to-human) transmission occurring.•There are three transmission cycles for yellow fever: sylvatic (jungle), intermediate (savannah), and urban.○The sylvatic (jungle) transmission cycle involves transmission of the virus between nonhuman primates and mosquito species found in the forest canopy. The virus is transmitted via mosquitoes from monkeys to humans when the humans encroach into the jungle during occupational or recreational activities.○In Africa, an intermediate (savannah) cycle involves transmission of YFV from tree hole-breeding *Aedes* spp. to humans living or working in jungle border areas. In this cycle, the virus may be transmitted from monkeys to humans or from human to human via these mosquitoes.○The urban transmission cycle involves transmission of the virus between humans and urban mosquitoes, primarily *Ae. aegypti*.•Humans infected with YFV experience the highest levels of viremia and can transmit the virus to mosquitoes shortly before onset of fever and for the first 3--5 days of illness.•Given the high level of viremia attained in humans, bloodborne transmission can also occur (via transfusion, needlestick, and intravenous drug abuse). Occurrence {#cesec59} ---------- •Yellow fever occurs in sub-Saharan Africa and tropical South America ([Maps 2-3](#f6){ref-type="fig"} and [2-4](#f7){ref-type="fig"} ), where it is endemic and intermittently epidemic (see [Table 2-12](#cetable12){ref-type="table"} for a list of countries with risk of yellow fever transmission).Map 2-3Yellow fever-endemic zones in Africa, 2009.Map 2-4Yellow fever-endemic zones in the Americas, 2009.Table 2-12Countries with risk of yellow fever transmission[1](#cetablefn54){ref-type="table-fn"}AfricaCentral and South AmericaAngolaEthiopiaNigeriaArgentina[2](#cetablefn55){ref-type="table-fn"}BeninGabonRwandaBolivia[2](#cetablefn55){ref-type="table-fn"}Burkina FasoThe GambiaSierra LeoneBrazil[2](#cetablefn55){ref-type="table-fn"}BurundiGhanaSão Tomé and PríncipeColombiaCameroonGuineaSenegalEcuador[2](#cetablefn55){ref-type="table-fn"}Central African RepublicGuinea-BissauSomaliaFrench GuianaChad[2](#cetablefn55){ref-type="table-fn"}KenyaSudan[2](#cetablefn55){ref-type="table-fn"}GuyanaCongo, Republic of theLiberiaTanzaniaPanama[2](#cetablefn55){ref-type="table-fn"}Côte d\'IvoireMali[2](#cetablefn55){ref-type="table-fn"}TogoParaguayDemocratic Republic of the CongoMauritania[2](#cetablefn55){ref-type="table-fn"}UgandaPeru[2](#cetablefn55){ref-type="table-fn"}Equatorial GuineaNiger[2](#cetablefn55){ref-type="table-fn"}SurinameTrinidad and Tobago[2](#cetablefn55){ref-type="table-fn"}Venezuela[2](#cetablefn55){ref-type="table-fn"}[^64][^65]•In Africa, natural immunity accumulates with age, and thus infants and children are at greatest risk for disease.•In South America, yellow fever occurs most frequently in unimmunized young men who are exposed to mosquito vectors through their work in forested or transitional areas.•Most yellow fever disease in humans is due to sylvatic or intermediate transmission cycles. However, urban yellow fever does occur periodically in Africa and sporadically in the Americas. Risk for Travelers {#cesec60} ------------------ ### General {#cesec61} A traveler\'s risk for acquiring yellow fever is determined by various factors, including immunization status, location of travel, season, duration of exposure, occupational and recreational activities while traveling, and local rate of virus transmission at the time of travel. Although reported cases of human disease are the principal indicator of disease risk, case reports may be absent because of a low level of transmission, a high level of immunity in the population (e.g., due to vaccination), or failure of local surveillance systems to detect cases. This "epidemiologic silence" does not equate to absence of risk and should not lead to travel without the protection provided by vaccination. ### Africa {#cesec62} YFV transmission in rural West Africa is seasonal, with an elevated risk during the end of the rainy season and the beginning of the dry season (usually July--October). However, YFV may be episodically transmitted by *Ae. aegypti* even during the dry season in both rural and densely settled urban areas. ### South America {#cesec63} •The risk for infection for South America is highest during the rainy season (January--May, with a peak incidence in February and March).•Given the high level of viremia in humans and the widespread distribution of *Ae. aegypti* in many towns and cities, South America is at risk for a large-scale urban epidemic. ### Yellow Fever Cases in Travelers {#cesec64} •During 1970--2002, a total of nine cases of yellow fever were reported in unvaccinated travelers from the United States and Europe who traveled to West Africa (five cases) or South America (four cases). Eight of these nine travelers died.•Only one documented case of yellow fever has occurred, which was in a vaccinated traveler from Spain, who visited several West African countries during 1988. ### Risk Estimates for Travelers {#cesec65} •The risk of acquiring yellow fever is difficult to predict because of variations in ecologic determinants of virus transmission. For a 2-week stay, the risks for illness and death due to yellow fever for an unvaccinated traveler traveling to an endemic area of○West Africa are 50 per 100,000 and 10 per 100,000, respectively○South America are 5 per 100,000 and 1 per 100,000, respectively•These estimates are a rough guideline based on the risk to indigenous populations, often during peak transmission season. Thus, these risk estimates may not accurately reflect the true risk to travelers, who may have a different immunity profile, take precautions against getting bitten by mosquitoes, and have less outdoor exposure.•The risk of acquiring yellow fever in South America is lower than that in Africa because the mosquitoes that transmit the virus between monkeys in the forest canopy do not often come in contact with humans, and there is a relatively high level of immunity in local residents secondary to vaccine use. Clinical Presentation {#cesec66} --------------------- •Asymptomatic or clinically inapparent infection is believed to occur in the majority of persons infected with YFV.•The incubation period is typically 3--6 days.•The initial illness presents as a nonspecific influenza-like syndrome with sudden onset of fever, chills, headache, backache, myalgias, prostration, nausea, and vomiting. Most patients improve after the initial presentation.•After a brief remission of hours to a day, approximately 15% of cases progress to develop a more serious or toxic form of the disease characterized by jaundice, hemorrhagic symptoms, and eventually shock and multisystem organ failure.•The overall case--fatality ratio for cases with jaundice is 20%--50%. Diagnosis {#cesec67} --------- •The preliminary diagnosis is based on the patient\'s clinical features, places and dates of travel, and activities.•Laboratory diagnosis is generally accomplished by testing serum to detect virus-specific IgM and IgG antibodies by serologic assays. Due to cross-reactivity between antibodies raised against other flaviviruses, more specific antibody testing, such as a neutralization test, should be done to confirm the infection.•Early in the illness, YFV or yellow fever viral RNA can often be detected in serum samples by virus isolation or nucleic acid amplification tests (NAAT). However, by the time more overt symptoms are recognized, the virus or viral RNA is usually undetectable. Therefore, virus isolation and NAAT should not be used for ruling out a diagnosis of yellow fever.•Health-care providers should contact their state or local health department or call 800-CDC-INFO (800-232-4636) for assistance with diagnostic testing for yellow fever infections and for questions about antibody response to vaccination. Treatment {#cesec68} --------- •No specific treatments have been found to benefit patients with yellow fever.•Treatment is symptomatic. Rest, fluids, and use of analgesics and antipyretics may relieve symptoms of fever and aching. Care should be taken to avoid certain medications, such as aspirin or other nonsteroidal anti-inflammatory drugs, which may increase the risk for bleeding.•Infected persons should be protected from further mosquito exposure (staying indoors and/or under a mosquito net) during the first few days of illness, so they do not contribute to the transmission cycle. Preventive Measures for Travelers {#cesec69} --------------------------------- ### Personal Protection Measures {#cesec70} •No drugs for preventing infection are available.•The best way to prevent mosquito-borne diseases, including yellow fever, is to avoid mosquito bites (see the [Protection Against Mosquitoes, Ticks, and Other Insects and Arthropods](#subchapter29){ref-type="sec"} section later in this chapter):○Use insect repellent containing DEET, Picaridin, oil of lemon eucalyptus, or IR3535 on exposed skin. Always follow the directions on the package.○Wear long sleeves, pants, and socks. If possible, treat clothes with permethrin.○Stay in screened or air-conditioned accommodations to keep mosquitoes out.○Get rid of mosquito sources by emptying standing water from flowerpots, buckets, car tires and barrels. ### Yellow Fever Vaccine {#cesec71} •Yellow fever is preventable by a relatively safe, effective vaccine.•All yellow fever vaccines currently manufactured are live attenuated viral vaccines.•YF-VAX, the only yellow fever vaccine approved for use in the United States, is manufactured by sanofi pasteur.•Studies comparing the reactogenicity and immunogenicity of various yellow fever vaccines, including those manufactured outside of the United States, suggest that there is no significant difference in the reactogenicity or immune response generated by the various vaccines. Thus, individuals who receive yellow fever vaccines in other countries should be considered protected against yellow fever. #### Recommendations for the Use of Yellow Fever Vaccine for Travelers {#cesec72} •Persons aged Δ9 months of age who are traveling to or living in areas with risk of yellow fever transmission in South America and Africa should be vaccinated. In addition, some countries require proof of yellow fever vaccination for entry. See the following section in this chapter ([Yellow Fever Vaccine Requirements and Recommendations, by Country](#subchapter7){ref-type="sec"}) for more detailed information on the requirements and recommendations for yellow fever vaccination for specific countries.•However, because severe adverse events (see below) can follow yellow fever vaccination, physicians should be careful to administer the vaccine only to persons truly at risk of exposure to YFV.•Refer to Yellow Fever Vaccine Recommendations of the Advisory Committee on Immunization Practices (ACIP) for additional information at [www.cdc.gov/vaccines/pubs/ACIP-list.htm](http://www.cdc.gov/vaccines/pubs/ACIP-list.htm){#interref16}. #### Vaccine Dose and Administration {#cesec73} •For all eligible persons, a single injection of 0.5 mL of reconstituted vaccine should be administered subcutaneously.•The International Health Regulations (IHR) published by WHO require revaccination at 10-year intervals. #### Vaccine Safety and Adverse Reactions {#cesec74} ##### Common Adverse Events {#cesec75} •Reactions to yellow fever vaccine are generally mild, with 10%--30% of vaccinees reporting mild systemic adverse events.•Reported events typically include low-grade fever, headache, and myalgias that begin within days after vaccination and last 5--10 days.•Approximately 1% of vaccinees temporarily curtail their regular activities because of these reactions. ##### Severe Adverse Events {#cesec76} ###### Hypersensitivity {#cesec77} Immediate hypersensitivity reactions, characterized by rash, urticaria, or asthma or a combination of these, are uncommon. Anaphylaxis following yellow fever vaccine is reported to occur at a rate of 1.8 cases per 100,000 doses administered. ###### Yellow Fever Vaccine-Associated Neurologic Disease (YEL-AND) {#cesec78} •YEL-AND represents a conglomerate of different clinical syndromes, including meningoencephalitis, Guillain--Barré syndrome (GBS), acute disseminated encephalomyelitis (ADEM), bulbar palsy, and Bell\'s palsy.•Historically, YEL-AND was seen primarily among infants as encephalitis, but more recent reports have been among persons of all ages.•The onset of illness for documented cases ranges 3--28 days after vaccination, and almost all cases were in first-time vaccine recipients.•YEL-AND is rarely fatal.•The incidence of YEL-AND in the United States is 0.8 per 100,000 doses administered. The rate is higher in persons Δ60 years of age, with a rate of 1.6 per 100,000 doses in persons 60--69 years of age and 2.3 per 100,000 doses in persons Δ70 years of age. ###### Yellow Fever Vaccine-Associated Viscerotropic Disease (YEL-AVD) {#cesec79} •YEL-AVD is a severe illness similar to wild-type disease, with vaccine virus proliferating in multiple organs and often leading to multisystem organ failure and death.•Since the initial cases of YEL-AVD were published in 2001, more than 40 confirmed and suspected cases have been reported throughout the world.•The onset of illness for YEL-AVD cases averaged 3.5 days (range: 1--8 days) after vaccination. YEL-AVD appears to occur after the first dose of yellow fever vaccine rather than with booster doses.•The case--fatality ratio for reported YEL-AVD cases is 53%.•The incidence of YEL-AVD in the United States is 0.4 cases per 100,000 doses of vaccine administered. The rate is higher for persons Δ60 years of age, with a rate of 1 per 100,000 doses in persons 60--69 years of age and 2.3 per 100,000 doses in persons aged Δ70 years of age. #### Contraindications {#cesec80} ##### Infants \<9 Months of Age {#cesec81} •The vaccine is contraindicated for routine use in infants \<9 months of age by the manufacturer and the FDA because of the increased risk of postvaccine encephalitis. However, ACIP and WHO recognize that situations occur in which vaccination of an infant 6--8 months of age might be considered, such as residence in or unavoidable travel to a yellow fever endemic or epidemic zone. The decision to immunize infants who are 6--8 months of age must balance the infant\'s risk for exposure with the risk for vaccine-associated encephalitis. **YF vaccine should never be administered to infants \<6 months of age**.•Physicians considering vaccinating infants aged \<9 months of age should contact their state health department or call 800-CDC-INFO (800-232-4636) for further advice. ##### Hypersensitivity {#cesec82} •Yellow fever vaccine is contraindicated in anyone with a history of acute hypersensitivity reaction to any of the vaccine components, including gelatin. Because the yellow fever vaccine is produced in chicken embryos, vaccine should not be administered to anyone with a history of acute hypersensitivity to egg or chicken proteins.•If vaccination of a person with a questionable history of hypersensitivity to one of the vaccine components is considered essential because of a high risk for acquiring yellow fever, desensitizing and vaccinating procedures are described in the vaccine package insert and should be performed under close medical supervision. ##### Immunosuppression {#cesec83} •The vaccine is contraindicated in persons with immunocompromising conditions, including symptomatic HIV infection or AIDS, malignancy, or diseases of the thymus (e.g., thymectomy) or those receiving immunosuppressant therapy (e.g., corticosteroids, alkylating agents, antimetabolites) or radiation therapy.•Immunosuppressed persons should not be immunized, and travel to yellow fever-endemic areas should be postponed or avoided.•If travel to yellow fever endemic areas is unavoidable, persons who cannot be immunized because of their immunosuppressive condition should be advised of the risk for acquiring yellow fever disease, instructed in methods for avoiding vector mosquitoes, and, if warranted, issued a medical waiver to fulfill international health regulations (see information in [Exemption from Vaccination and Waiver Letters](#cesec97){ref-type="sec"} in this section).•Physicians considering vaccinating an immunosuppressed individual can contact their state health department or call for more information.•Family members of immunosuppressed or HIV-infected persons who themselves have no contraindications can receive yellow fever vaccine. ###### AIDS or Symptomatic HIV {#cesec84} No large-scale trials have been done to evaluate the safety of the yellow fever vaccine in individuals with HIV or AIDS. However, because yellow fever vaccine is a live, viral vaccine, it is contraindicated in persons with symptomatic HIV infection or AIDS. (For persons with asymptomatic HIV infection, see [Precautions](#cesec87){ref-type="sec"} below.) ###### History of Thymus Disease {#cesec85} •A history of thymus disease is a contraindication to yellow fever vaccine.•Four persons with a history of thymectomy for a thymoma were noted among the first 23 cases of YEL-AVD, suggesting that compromised thymus function is an independent risk factor for YEL-AVD.•Health-care providers should be careful to ask about a history of thymus disorder, including myasthenia gravis, thymoma, or prior thymectomy, when screening a patient before administering yellow fever vaccine. ###### Immunosuppressive Medication {#cesec86} •Although no studies have been done to evaluate the safety of yellow fever vaccine in persons receiving immunosuppressive or immunomodulating medicines, the vaccine is contraindicated in those receiving medications that alter the ability to resist viral infections. The vaccine should not be given to individuals who are taking medications with a warning in the package insert against the use of live viral vaccines.•Low-dose (i.e., 20 mg or less of prednisone or equivalent/day); short-term (i.e., \<2 weeks) systemic corticosteroid therapy or intra-articular, bursal, or tendon injections with corticosteroids; and intranasal corticosteroids are not thought to be sufficiently immunosuppressive to constitute an increased hazard to recipients of yellow fever vaccine (see The Immunocompromised Traveler section in Chapter 8). #### Precautions {#cesec87} ##### Adults 60 Years of Age or Older {#cesec88} •Analysis of adverse events passively reported to the Vaccine Adverse Event Reporting System (VAERS) indicate that persons 60 years of age or older may be at increased risk for systemic adverse events following vaccination compared with younger persons.•The rate of any serious adverse event following vaccination is 1.5 times higher than the average rate for persons 60--69 years of age and 3 times higher for persons 70 years or older.•To determine if vaccination should be administered to travelers 60 years of age or older, the risks and benefits of vaccination should be weighed against their destination-specific risk for exposure to YFV. ##### Asymptomatic HIV {#cesec89} •Persons who are HIV-infected but who do not have AIDS or other symptomatic manifestations of HIV infection, who have established laboratory verification of adequate immune system function (e.g., CD4+ T cell counts \>200/mm^3^), and who cannot avoid potential exposure to YFV should be offered the choice of vaccination.•If international travel requirements are the only reason to vaccinate an asymptomatic HIV-infected person, rather than an increased risk for acquiring yellow fever, the person should be excused from immunization and issued a medical waiver to fulfill health regulations (see information in [Exemption from Vaccination and Waiver Letters](#cesec97){ref-type="sec"} in this section).•Data are limited regarding seroconversion rates after yellow fever vaccination among asymptomatic HIV-infected persons, but indicate that the seroconversion rate among such persons may be reduced. Because vaccination of asymptomatic HIV-infected persons might be less effective than that of persons not infected with HIV, measurement of the neutralizing antibody response to vaccination should be considered before travel. ##### Pregnancy {#cesec90} •The safety of yellow fever vaccination during pregnancy has not been studied in a large prospective trial. However, a recent study of women who were vaccinated with yellow fever vaccine early in their pregnancies found no major malformations in their infants. There was slight increased risk noted for minor, mostly skin, malformations.•In a similar study, a higher rate of spontaneous abortions in pregnant women receiving the vaccine was reported but not substantiated.•The proportion of women vaccinated during pregnancy who develop YF IgG-specific antibodies is variable depending on the study (38.6% or 98.2%) and may be correlated with the trimester in which they received the vaccine. Because pregnancy may affect immunologic function, serologic testing can be considered to document a protective immune response to the vaccine.•For pregnant women, if travel is unavoidable and the vaccination risks are felt to outweigh the risks of YF exposure, these women should be excused from immunization and, if applicable, issued a medical waiver to fulfill international health regulations (see information in [Exemption from Vaccination and Waiver Letters](#cesec97){ref-type="sec"} in this section). Pregnant women who must travel to areas where the risk of yellow fever infection is high should be vaccinated, and their infants should be monitored after birth for evidence of congenital infection and other possible adverse effects resulting from yellow fever vaccination.•Although there are no specific data, it is recommended that a woman wait 4 weeks after receiving the live virus yellow fever vaccine before conceiving. ##### Breastfeeding {#cesec91} •Whether the yellow fever vaccine is excreted in breast milk is not known.•One suspect case of YEL-AND has been reported in a 1-month old infant whose mother was vaccinated with yellow fever vaccine and the infant was exclusively breastfed. Testing was unable to determine if the breast milk was the mode of transmission.•It is recommended that vaccination of nursing mothers should be avoided. However, when travel of nursing mothers to high-risk yellow fever-endemic areas cannot be avoided or postponed, these women should be vaccinated. #### Simultaneous Administration of Other Vaccines and Drugs {#cesec92} •One study suggested that the immune response to yellow fever vaccine is not inhibited by administration of measles vaccine (also a live, attenuated vaccine) given concurrently or at various intervals of a few days to 1 month prior. However, to minimize the potential risk for interference, injectable or nasally administered live vaccines not administered on the same day should be given at least 4 weeks apart.•A prospective study of persons given yellow fever vaccine along with 5 mL of commercially available IG showed no alteration of the immunologic response to yellow fever vaccine when compared with controls. International Certificate of Vaccination or Prophylaxis (ICVP) {#cesec93} -------------------------------------------------------------- ### Background {#cesec94} •The International Health Regulations (IHR) allow countries to require proof of yellow fever vaccination for entry and from travelers arriving from certain countries, even if only in transit, to prevent importation and indigenous transmission of YFV.•Some countries require evidence of vaccination from all entering travelers, which includes direct travel from the United States ([Table 2-13](#cetable13){ref-type="table"} ).Table 2-13Countries that require proof of yellow fever vaccination for all arriving travelers[1](#cetablefn56){ref-type="table-fn"}AngolaFrench GuianaBeninGabonBolivia (or signed affidavit at point of entry)GhanaBurkina FasoLiberiaBurundiMaliCameroonNigerCentral African RepublicRwandaCongo, Republic of theSão Tomé and PríncipeCôte d\'IvoireSierra LeoneDemocratic Republic of the CongoTogo[^66]•Travelers who arrive in a country with a yellow fever vaccination entry requirement without proof of yellow fever vaccination may be quarantined up to 6 days.•Travelers with a specific contraindication to yellow fever vaccine should request a waiver from a physician before traveling to countries requiring vaccination (see below). ### Authorization to Provide Vaccinations and to Validate the ICVP {#cesec95} •Under the revised IHR (2005), effective December 15, 2007, all state parties (countries) are required to issue a new ICVP. This is intended to replace the former International Certificate of Vaccination against Yellow Fever (ICV).○Persons who received a yellow fever vaccination after December 15, 2007, must provide proof of vaccination on an ICVP.○If the person received the vaccine before December 15, 2007, the original ICV is still valid, provided that the vaccination was given less than 10 years previously.•Vaccinees should receive a completed ICVP ([Figure 2-1](#f1){ref-type="fig"} ), validated (stamped and signed) with the center\'s stamp where the vaccine was given (see below).○An ICVP must be complete in every detail; if incomplete or inaccurate, it is not valid.○Failure to secure validations can cause a traveler to be quarantined, denied entry, or possibly revaccinated at the point of entry to a country. This is not a recommended option for the traveler. Figure 2-1Example International Certificate of Vaccination or Prophylaxis (ICVP).•A copy of the ICVP, CDC 731 (formerly PHS 731) may be purchased from the U.S. Government Printing Office, Washington, D.C., <http://bookstore.gpo.gov/,> telephone 866-512-1800. The stock number is 017-001-00567-3 for 25 copies and 017-001-00566-5 for 100 copies.•This certificate of vaccination is valid for a period of 10 years, beginning 10 days after vaccination. With booster doses of the vaccine, the certificate is considered valid from the day of vaccination. ### Persons Authorized to Sign the Certificate and Designated Yellow Fever Vaccination Centers {#cesec96} •The ICVP must be signed by a licensed physician or by a health-care worker designated by the physician supervising the administration of the vaccine ([Figure 2-1](#f1){ref-type="fig"}). A signature stamp is not acceptable.•Yellow fever vaccination must be given at a certified center in possession of an official "Uniform Stamp," which can be used to validate the ICVP.•State health departments are responsible for designating nonfederal yellow fever vaccination centers and issuing Uniform Stamps to physicians.•Information about the location and hours of yellow fever vaccination centers may be obtained by visiting CDC\'s Travelers\' Health website at [wwwn.cdc.gov/travel/yellowfever.aspx](http://wwwn.cdc.gov/travel/yellowfever.aspx){#interref18}. ### Exemption from Vaccination and Waiver Letters {#cesec97} •Some countries do not require an ICVP for infants younger than a certain age (e.g., \<6 months, \<9 months, or \<1 year of age, depending on the country). Age requirements for vaccination for individual countries can be found in the Yellow Fever Vaccine Requirements and Recommendations section in this chapter.•For medical contraindications, a physician who has decided to issue a waiver should fill out and sign the Medical Contraindications to Vaccination section of the ICVP ([Figure 2-2](#f2){ref-type="fig"} ). The physician should also---○Give the traveler a signed and dated exemption letter on the physician\'s letterhead stationery, clearly stating the contraindications to vaccination and bearing the stamp used by the yellow fever vaccination centers to validate the ICVP.○Inform the traveler of any increased risk of yellow fever infection associated with nonvaccination and how to minimize this risk by using mosquito protection measures. Figure 2-2Example International Certificate of Vaccination or Prophylaxis (ICVP) medical contraindication to vaccination.•Reasons other than medical contraindications are not acceptable for exemption from vaccination.•The traveler should be advised that issuance of a waiver does not guarantee its acceptance by the destination country. On arrival at the destination, the traveler may be faced with quarantine, refusal of entry, or vaccination on site.•To potentially improve the likelihood of acceptance of a waiver upon arrival at the destination country, the provider can suggest that the traveler take the following additional measures before initiating travel:○Obtain specific and authoritative advice from the embassy or consulate of the country or countries he or she plans to visit.○Request documentation of requirements for waivers from embassies or consulates and retain these along with the completed Medical Contraindication to Vaccination section of the ICVP. ### Requirements Versus Recommendations {#cesec98} •Country entry **requirements** for proof of yellow fever vaccination under the IHRs are different from CDC\'s **recommendations**.•Yellow fever vaccine entry **requirements** are established by countries in order to prevent the importation and transmission of YFV, and are allowed under the IHRs. Travelers must comply with these to enter the country, unless they have been issued a medical waiver. Certain countries require vaccination from travelers arriving from all countries, while some countries require vaccination only for travelers coming from "a country with risk of yellow fever transmission" ([Table 2-14](#cetable14){ref-type="table"} ). WHO defines those areas "at risk of yellow fever transmission" as countries or areas where yellow fever has been reported currently or in the past, plus where vectors and animal reservoirs currently exist. Country requirements are subject to change at any time; therefore, CDC encourages travelers to check with the appropriate embassy or consulate before departure.Table 2-14Yellow fever vaccine requirements and recommendations, by countryCountryYellow Fever Vaccine Requirements[1](#cetablefn57){ref-type="table-fn"},[2](#cetablefn58){ref-type="table-fn"}CDC Yellow Fever Vaccine Recommendations[2](#cetablefn58){ref-type="table-fn"},[3](#cetablefn59){ref-type="table-fn"},[4](#cetablefn60){ref-type="table-fn"}**Afghanistan**If traveling from a country with risk of yellow fever transmissionNone**Albania**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Algeria**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Andorra**Not requiredNone**Angola**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Anguilla** (U.K.)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Antarctica**Not requiredNone**Antigua and Barbuda**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Argentina**Not requiredYellow fever vaccination is recommended for all travelers ≥9 months of age who are going to the northern and northeastern forested areas of Argentina, including Iguassu Falls and all areas bordering Paraguay and Brazil. These areas include all departments of Misiones and Formosa Provinces; and the Department of Bermejo in Chaco Province; Departments of Berón de Astrada, Capital, General Alvear, General Paz, Ituzaingó, Itatí, Paso de los Libres, San Cosme, San Miguel, San Martín, and Santo Tomé in Corrientes Province; Departments of Valle Grande, Ledesma, Santa Bárbara, and San Pedro in Jujuy Province; and Departments of General José de San Martín, Oran, Rivadavia, and Anta in Salta Province.**Armenia**Not requiredNone**Aruba**If traveling from a country with risk of yellow fever transmission and ≥6 months of age[5](#cetablefn61){ref-type="table-fn"}None**Australia**All persons ≥1 year of age who, within 6 days of arrival in Australia, have been in or have passed through a country with risk of yellow fever transmissionNone**Austria**Not requiredNone**Azerbaijan**Not requiredNone**Azores** (Portugal)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Bahamas, The**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Bahrain**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Bangladesh**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Barbados**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Belarus**Not requiredNone**Belgium**Not requiredNone**Belize**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Benin**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Bermuda** (U.K.)Not requiredNone**Bhutan**If traveling from a country with risk of yellow fever transmissionNone**Bolivia**Required for all travelers ≥1 year of age\ \ For U.S. citizens: Medical waivers must be translated into Spanish and accompany the International Certificate of Vaccination or Prophylaxis (ICVP). Travelers who do not have a valid ICVP will still be allowed to enter Bolivia if they agree to sign an affidavit exempting the Bolivian state from any liability in the event the traveler gets sick with yellow fever within the Bolivian territory. This last option may cause delays at the point of entry.For all travelers ≥9 months of age traveling to areas east of the Andes Mountains (see [Map 2-4](#f7){ref-type="fig"}). Vaccination is NOT recommended for travel only to the cities of La Paz or Sucre.**Bosnia and Herzegovina**Not requiredNone**Botswana**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Brazil**Not requiredFor all travelers ≥9 months of age going to the following areas at risk of yellow fever transmission, including the ENTIRE states of Acre, Amapá, Amazonas, Distrito Federal (including the capital city of Brasilia), Goiás, Maranhão, Mato Grosso, Mato Grosso do Sul, Minas Gerais, Pará, Rondônia, Roraima, and Tocatins; and the designated areas of the following states: northwest and west Bahia, central and west Paraná, southwest Piauí, northwest and west central Rio Grande do Sul, far west Santa Catarina, and north and west São Paulo.\ \ Vaccination is recommended for travelers visiting Iguassu Falls. Vaccination is NOT recommended for travel to the following coastal cities: Rio de Janeiro, São Paulo, Salvador, Recife, and Fortaleza.**British Indian Ocean Territory** includes **Diego Garcia** (U.K.)Not requiredNone**Brunei**Required from travelers ≥1 year of age arriving within 6 days from countries with risk of yellow fever transmission or having passed through areas partly or wholly at risk of yellow fever transmission within the preceding 6 days.None**Bulgaria**Not requiredNone**Burkina Faso**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Burma (Myanmar)**If traveling from a country with risk of yellow fever transmission. Required also for nationals and residents of Burma (Myanmar) departing for a country with risk of yellow fever transmissionNone**Burundi**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Cambodia**If traveling from a country with risk of yellow fever transmissionNone**Cameroon**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Canada**Not requiredNone**Canary Islands** (Spain)Not requiredNone**Cape Verde**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Cayman Islands** (U.K.)Not requiredNone**Central African Republic**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Chad**If traveling from a country with risk of yellow fever transmissionFor all travelers ≥9 months of age traveling to areas south of the Sahara Desert**Chile**Not requiredNone**China**If traveling from a country with risk of yellow fever transmissionNone**Christmas Island** (Australia)All travelers ≥1 year of age, who within the past 6 days have traveled or passed through an endemic area, as listed by WHONone**Cocos (Keeling) Islands**All persons ≥1 year of age who, within 6 days of arrival, have been in or have passed through a country with risk of yellow fever transmissionNone**Colombia**Not requiredFor all travelers ≥9 months of age. Travelers whose itinerary is limited to the cities of Bogotá, Cali, or Medellín are at lower risk and may consider foregoing vaccination.**Comoros**Not requiredNone**Congo, Republic of the (Congo-Brazzaville)**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Cook Islands** (New Zealand)Not requiredNone**Costa Rica**Required from travelers coming from countries with risk of yellow fever transmission. No certificate is required for travelers \<9 months of age and ≥60 years of age, pregnant or lactating women, persons with allergy to eggs or gelatin, immunosuppression, thymus disease, a personal or family history of adverse reactions associated with the yellow fever vaccine, or asymptomatic HIV infection with laboratory evidence of satisfactory immune functions.\ \ The following countries are considered at risk of yellow fever transmission:\ \ *South America*: Boliva, Brazil, Colombia, Ecuador, French Guyana, Peru, Venezuela\ \ *Africa*: Angola, Benin, Burkina Faso, Cameroon, Gabon, The Gambia, Ghana, Guinea, Liberia, Nigeria, Democratic Republic of the Congo, Sierra Leone, SudanNone**Côte d\'Ivoire (Ivory Coast)**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Croatia**Not requiredNone**Cuba**Not requiredNone**Cyprus**Not requiredNone**Czech Republic**Not requiredNone**Democratic Republic of the Congo (Congo-Kinshasa)**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Denmark**Not requiredNone**Djibouti**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Dominica**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Dominican Republic**Not requiredNone**Easter Island** (Chile)Required for travelers coming from a country with risk of yellow fever transmissionNone**Ecuador** (including the **Galápagos Islands**)Required from travelers ≥1 year of age coming from countries with risk of yellow fever transmission. Nationals and residents of Ecuador are required to possess certificates of vaccination on their departure to an area with risk of yellow fever transmission.For all travelers ≥9 months of age who are traveling to the following provinces in the Amazon Basin: Morona-Santiago, Napo, Orellana, Pastaza, Sucumbíos, and Zamorarisk Chinchipe, and all other areas in the eastern part of the Andes Mountains, NOT including the cities of Quito and Guayaquil or the Galápagos Islands**Egypt**If traveling from countries with risk of yellow fever transmission and ≥1 year of age. Air passengers without a certificate in transit, but coming from these countries or areas, will be detained in the precincts of the airport until they resume their journey. All travelers arriving from Sudan are required to have a vaccination certificate or a location certificate issued by a Sudanese official center, stating that they have not been in Sudan south of 15° N within the previous 6 days.\ \ Required also for travelers arriving or transiting from:\ \ *Africa*: Angola, Benin, Burkina Faso, Burundi, Cameroon, Central African Republic, Chad, Congo, Côte d\'Ivoire, Democratic Republic of the Congo, Equatorial Guinea, Ethiopia, Gabon, The Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Liberia, Mali, Niger, Nigeria, Rwanda, São Tomé and Príncipe, Senegal, Sierra Leone, Somalia, Sudan (south of 15° N), Tanzania, Togo, Uganda, and Zambia\ \ *Americas*: Belize, Bolivia, Brazil, Colombia, Costa Rica, Ecuador, French Guiana, Guyana, Panama, Peru, Suriname, Trinidad and Tobago, and VenezuelaNone**El Salvador**If traveling from a country with risk of yellow fever transmission and ≥6 months of age[5](#cetablefn61){ref-type="table-fn"}None**Equatorial Guinea**If traveling from a country with risk of yellow fever transmissionFor all travelers ≥9 months of age**Eritrea**If traveling from a country with risk of yellow fever transmissionNone**Estonia**Not requiredNone**Ethiopia**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageFor all travelers ≥9 months of age**Falkland Islands** (U.K.)Not requiredNone**Faroe Islands** (Denmark)Not requiredNone**Fiji**If traveling from a country with risk of yellow fever transmission within 10 days of having stayed overnight or longer and ≥1 year of ageNone**Finland**Not requiredNone**France**Not requiredNone**French Guiana**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**French Polynesia**, includes the island groups of **Society Islands (Tahiti, Moorea**, and **Bora-Bora), Marquesas Islands (Hiva Oa** and **Ua Huka**), and **Austral Islands (Tubuai** and **Rurutu**)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Gabon**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Gambia, The**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageFor all travelers ≥9 months of age**Georgia**Not requiredNone**Germany**Not requiredNone**Ghana**Required upon arrival from all countriesFor all travelers ≥9 months of age**Gibraltar** (U.K.)Not requiredNone**Greece**Not requiredNone**Greenland** (Denmark)Not requiredNone**Grenada**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Guadeloupe**, including **St. Barthelemy** and **Saint Martin** (France)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Guam** (U.S.)Not requiredNone**Guatemala**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Guinea**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageFor all travelers ≥9 months of age**Guinea-Bissau**If traveling from a country with risk of yellow fever transmission and ≥1 year of age\ \ Required also for travelers arriving from:\ \ *Africa:* Angola, Benin, Burkina Faso, Burundi, Cape Verde, Central African Republic, Chad, Congo, Côte d\'Ivoire, Democratic Republic of the Congo, Djibouti, Equatorial Guinea, Ethiopia, Gabon, The Gambia, Ghana, Guinea, Kenya, Liberia, Madagascar, Mali, Mauritania, Mozambique, Niger, Nigeria, Rwanda, São Tomé and Príncipe, Senegal, Sierra Leone, Somalia, Tanzania, Togo, Uganda, and Zambia\ \ *Americas:* Bolivia, Brazil, Colombia, Ecuador, French Guiana, Guyana, Panama, Peru, Suriname, and VenezuelaFor all travelers ≥9 months of age**Guyana**If traveling from a country with risk of yellow fever transmission\ \ Required also for travelers arriving from:\ \ *Africa*: Angola, Benin, Burkina Faso, Burundi, Cameroon, Central African Republic, Chad, Congo, Côte d\'Ivoire, Democratic Republic of the Congo, Equatorial Guinea, Ethiopia, Gabon, The Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Liberia, Mali, Mauritania, Niger, Nigeria, Rwanda, São Tomé and Príncipe, Senegal, Sierra Leone, Somalia, Sudan, Tanzania, Togo, and Uganda\ \ *Americas*: Belize, Bolivia, Brazil, Colombia, Ecuador, French Guiana, Guyana, Panama, Peru, Suriname, and VenezuelaFor all travelers ≥9 months of age**Haiti**If traveling from a country with risk of yellow fever transmissionNone**Holy See**Not requiredNone**Honduras**Required from travelers ≥1 year of age coming from countries with risk of yellow fever transmission. The Government of Honduras is also recommending vaccine for travelers coming from Panama.None**Hong Kong SAR** (China)Not requiredNone**Hungary**Not requiredNone**Iceland**Not requiredNone**India**Required from travelers ≥6 months of age coming from a country with risk of yellow fever transmission\ \ Specifically, per the Government of India, anyone (except infants ≥6 months old) arriving by air or sea without a certificate is detained in isolation for up to 6 days if that person (i) arrives within 6 days of departure from an area with risk of yellow fever transmission, or (ii) has been in such an area in transit (except those passengers and members of crew who, while in transit through an aiport situated in an area with risk of yellow fever transmission, remained within the airport premises during the period of their entire stay and the Health Officer agrees to such an exemption), or (iii) has come on a ship that started from or touched at any port in a yellow fever area with risk of yellow fever transmission up to 30 days before its arrival into India, unless such a ship has been disinsected in accordance with the procedure laid down by WHO, or (iv) has come by an aircraft which has been in an area with risk of yellow fever transmission and has not been disinsected in accordance with the provisions laid down in the Indian Aircraft Public Health Rules, 1954, or those recommended by WHO.\ \ The following countries and areas are regarded as at risk of yellow fever transmission:\ \ *Africa*: Angola, Benin, Burkina Faso, Burundi, Cameroon, Central African Republic, Chad, Congo, Côlte d\'Ivoire, Democratic Republic of the Congo, Equatorial Guinea, Ethiopia, Gabon, The Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Liberia, Mali, Niger, Nigeria, Rwanda, São Tomé and Principe, Senegal, Sierra Leone, Somalia, Sudan, Togo, Uganda, Tanzania, and Zambia\ \ *Americas*: Bolivia, Brazil, Colombia, Ecuador, French Guiana, Guyana, Panama, Peru, Suriname, Trinidad and Tobago, and Venezuela\ \ *Note*: When a case of yellow fever is reported from any country, that country is regarded by the Government of India as a country with risk of yellow fever and is added to the above list.None**Indonesia**If traveling from a country with risk of yellow fever transmission and ≥9 months of ageNone**Iran**If traveling from a country with risk of yellow fever transmissionNone**Iraq**If traveling from a country with risk of yellow fever transmissionNone**Ireland**Not requiredNone**Israel**Not requiredNone**Italy**Not requiredNone**Jamaica**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Japan**Not requiredNone**Jordan**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Kazakhstan**If traveling from a country with risk of yellow fever transmissionNone**Kenya**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageFor all travelers ≥9 months of age. The cities of Nairobi and Mombasa have lower risk of transmission than rural areas.**Kiribati** (formerly **Gilbert Islands**), includes **Tarawa, Tabuaeran (Fanning Island**), and **Banaba (Ocean Island)**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Korea, North**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Korea, South**Not requiredNone**Kosovo**Not requiredNone**Kuwait**Not requiredNone**Kyrgyzstan**Not requiredNone**Laos**If traveling from a country with risk of yellow fever transmissionNone**Latvia**Not requiredNone**Lebanon**If traveling from a country with risk of yellow fever transmission and ≥6 months of age[5](#cetablefn61){ref-type="table-fn"}None**Lesotho**If traveling from a country with risk of yellow fever transmissionNone**Liberia**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Libya**If traveling from a country with risk of yellow fever transmissionNone**Liechtenstein**Not requiredNone**Lithuania**Not requiredNone**Luxembourg**Not requiredNone**Macau SAR** (China)Not requiredNone**Macedonia**Not requiredNone**Madagascar**If traveling from a country with risk of yellow fever transmissionNone**Madeira Islands** (Portugal)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Malawi**If traveling from a country with risk of yellow fever transmissionNone**Malaysia**Required from travelers ≥1 year of age arriving within 6 days from countries with risk of yellow fever transmissionNone**Maldives**If traveling from a country with risk of yellow fever transmissionNone**Mali**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age going to areas south of the Sahara Desert**Malta**If traveling from a country with risk of yellow fever transmission and ≥9 months of age\ \ If indicated on epidemiological grounds, infants \<9 months of age are subject to isolation or surveillance if coming from an area with risk of yellow fever transmission.None**Marshall Islands**Not requiredNone**Martinique** (France)Not requiredNone**Mauritania**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageFor all travelers ≥9 months of age traveling to areas south of the Sahara Desert**Mauritius**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Mayotte** (French territorial collectivity)Not requiredNone**Mexico**Not requiredNone**Micronesia Federated States of**; includes **Yap Islands, Pohnpei, Chuuk**, **and Kosrae**,Not requiredNone**Moldova**Not requiredNone**Monaco**Not requiredNone**Mongolia**Not requiredNone**Montenegro**Not requiredNone**Montserrat** (U.K.)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Morocco**Not requiredNone**Mozambique**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Namibia**If traveling from a country with risk of yellow fever transmission\ \ The countries, or parts of countries, included in the endemic zones in Africa and South America are regarded as areas with risk of yellow fever transmission. Travelers on scheduled flights that originated outside the countries with risk of yellow fever transmission, but who have been in transit through these areas, are not required to possess a certificate provided that they remained at the scheduled airport or in the adjacent town during transit. All passengers whose flights originated in countries with risk of yellow fever transmission or who have been in transit through these countries on unscheduled flights are required to possess a certificate.\ \ The certificate is not insisted upon in the case of children \<1 year of age, but such infants may be subject to surveillance.None**Nauru**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Nepal**If traveling from a country with risk of yellow fever transmissionNone**Netherlands**Not requiredNone**Netherlands Antilles (Bonaire, Curaçao, Saba, St. Eustasius**, and **St. Maarten)**If traveling from a country with risk of yellow fever transmission and ≥6 months of age[5](#cetablefn61){ref-type="table-fn"}None**New Caledonia** (France)If traveling from a country with risk of yellow fever transmission and ≥1 year of age\ \ *Note*: In the event of an epidemic threat to the territory, a specific vaccination certificate may be required.None**New Zealand**Not requiredNone**Nicaragua**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Niger**Required upon arrival from all countries if traveler is ≥1 year of age. The Government of Niger recommends vaccine for travelers leaving Niger.For all travelers ≥9 months of age traveling to areas south of the Sahara Desert**Nigeria**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageFor all travelers ≥9 months of age**Niue** (New Zealand)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Norfolk Island** (Australia)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Northern Mariana Islands** (U.S.), includes **Saipan, Tinian**, and **Rota Island**Not requiredNone**Norway**Not requiredNone**Oman**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Pakistan**Required from travelers coming from any part of a country in which there is a risk of yellow fever transmission; infants \<6 months of age are exempt if the mother\'s vaccination certificate shows that she was vaccinated before the birth of the child.[5](#cetablefn61){ref-type="table-fn"}None**Palau**Required from all travelers ≥1 year of age coming from countries with risk of yellow fever transmission or from countries in any part of which there is a risk of yellow fever transmissionNone**Panama**If traveling from a country with risk of yellow fever transmissionFor all travelers ≥9 months of age traveling to the provinces of Darien, Kuna Yala (old San Blas), Comarca Emberá, and Panama east of the Canal Zone, EXCLUDING the Canal Zone, Panama City, and San Blas Islands**Papua New Guinea**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Paraguay**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageFor all travelers ≥9 months of age**Peru**Not requiredFor all travelers ≥9 months of age traveling to the areas east of the Andes Mountains (see [Map 2-4](#f7){ref-type="fig"}) and for those who intend to visit any jungle areas of the country \<2,300 m (\<7,546 ft). Travelers who are limiting travel to the cities of Cuzco and Machu Picchu do NOT need vaccination.**Philippines**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Pitcairn Islands** (U.K.)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Poland**Not requiredNone**Portugal**Required only for travelers ≥1 year of age arriving from a country with risk of yellow fever transmission and destined for the Azores and Madeira. However, no certificate is required for travelers in transit at Funchal, Santa Maria, and Porto Santo.None**Puerto Rico** (U.S.)Not requiredNone**Qatar**Not requiredNone**Réunion** (France)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Romania**Not requiredNone**Russia**Not requiredNone**Rwanda**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Saint Helena** (U.K.)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Saint Kitts (Saint Christopher)** and **Nevis** (U.K.)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Saint Lucia**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Saint Pierre and Miquelon** (France)Not requiredNone**Saint Vincent and the Grenadines**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Samoa** (formerly **Western Somoa**)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Samoa, American** (U.S.)Not requiredNone**San Marino**Not requiredNone**São Tomé and Príncipe**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Saudi Arabia**If traveling from a country with risk of yellow fever transmissionNone**Senegal**If traveling from a country with risk of yellow fever transmissionFor all travelers ≥9 months of age**Serbia**Not requiredNone**Seychelles**If traveling from a country with risk of yellow fever transmission within the preceding 6 days and ≥1 year of ageNone**Sierra Leone**Required upon arrival from all countriesFor all travelers ≥9 months of age**Singapore**If traveling from a country with risk of yellow fever transmission within the preceding 6 days and ≥1 year of ageNone**Slovakia**Not requiredNone**Slovenia**Not requiredNone**Solomon Islands**If traveling from a country with risk of yellow fever transmissionNone**Somalia**If traveling from a country with risk of yellow fever transmissionFor all travelers ≥9 months of age**South Africa**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**South Georgia**Not requiredNone**South Sandwich Islands**Not requiredNone**Spain**Not requiredNone**Sri Lanka**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Sudan**If traveling from a country with risk of yellow fever transmission and ≥9 months of age. A certificate may be required for travelers leaving Sudan.For all travelers ≥9 months of age traveling to areas south of the Sahara Desert, EXCLUDING the city of Khartoum**Suriname**If traveling from a country with risk of yellow fever transmission fever transmission and ≥1 year of ageFor all travelers ≥9 months of age**Swaziland**If traveling from a country with risk of yellow fever transmissionNone**Sweden**Not requiredNone**Switzerland**Not requiredNone**Syria**If traveling from a country with risk of yellow fever transmissionNone**Taiwan**If traveling from a country with risk of yellow fever transmissionNone**Tajikistan**Not requiredNone**Tanzania**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageFor all travelers ≥9 months of age. The city of Dar es Salaam has a lower risk of transmission than rural areas.**Thailand**If traveling from a country with risk of yellow fever transmission and ≥9 months of ageNone**Timor-Leste (East Timor)**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Togo**Required upon arrival from all countries if traveler is ≥1 year of ageFor all travelers ≥9 months of age**Tokelau** (New Zealand)Not requiredNone**Tonga**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Trinidad and Tobago**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageFor all travelers ≥9 months of age whose itinerary includes Trinidad. Port of Spain has lower risk of transmission than rural or forested areas. Cruise ship passengers who do not disembark from the ship or travelers visiting only the urban area of Port of Spain (including passengers in-transit only) may consider foregoing vaccination. Vaccination is NOT recommended for those visiting only Tobago.**Tunisia**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Turkey**Not requiredNone**Turkmenistan**Not requiredNone**Turks and Caicos Islands** (U.K.)If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Tuvalu**Not requiredNone**Uganda**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageFor all travelers ≥9 months of age**Ukraine**Not requiredNone**United Arab Emirates**Not requiredNone**United Kingdom** (with **Channel Islands** and **Isle of Man**)Not requiredNone**United States**Not requiredNone**Uruguay**If traveling from a country with risk of yellow fever transmissionNone**Uzbekistan**Not requiredNone**Vanuatu**Not requiredNone**Venezuela**Not requiredFor all travelers ≥9 months of age traveling to Venezuela, EXCEPT the northern coastal area (see [Map 2-4](#f7){ref-type="fig"}). The cities of Caracas and Valencia are NOT in the endemic zone.**Vietnam**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Virgin Islands British**,Not requiredNone**Virgin Islands, U.S.**Not requiredNone**Wake Island, U.S.**Not requiredNone**Western Sahara**Not requiredNone**Yemen**If traveling from a country with risk of yellow fever transmission and ≥1 year of ageNone**Zambia**Not requiredNone**Zimbabwe**If traveling from a country with risk of yellow fever transmissionNone[^67][^68][^69][^70][^71]**Please note:** Country requirements for yellow fever vaccine are subject to change at any time, and CDC vaccine recommendations are subject to change at any time if disease conditions change; therefore, CDC encourages health-care providers and travelers to check for updates on the CDC website [www.cdc.gov/travel,](http://www.cdc.gov/travel,){#interref4} and with the destination country\'s embassy or consulate in sufficient time to receive yellow fever vaccination or to obtain a waiver if recommendations or requirements have changed.•The information in the section on yellow fever vaccine **recommendations** is advice given by CDC to prevent yellow fever infections among travelers. **Recommendations** are subject to change at any time if disease conditions change; therefore, CDC encourages travelers to check for relevant travel notices on the CDC website [www.cdc.gov/travel](http://www.cdc.gov/travel){#interref19} before departure. ### Vaccination for Travel on Military Orders {#cesec99} Because military requirements may exceed those indicated in this publication, any person who plans to travel on military orders (civilians and military personnel) should be advised to contact the nearest military medical facility to determine the requirements for the trip. YELLOW FEVER VACCINE REQUIREMENTS AND RECOMMENDATIONS, BY COUNTRY {#subchapter7} ================================================================= Gershman Mark Schroeder Betsy Jentes Emily S. Marano Nina JAPANESE ENCEPHALITIS (JE) {#cesec100} -------------------------- **Marc Fischer, Anne Griggs, J. Erin Staples** ### Infectious Agent {#cesec101} Japanese encephalitis virus (JEV) is a single-stranded RNA virus that belongs to the genus *Flavivirus* and is closely related to West Nile and St. Louis encephalitis viruses. ### Mode of Transmission {#cesec102} •JEV is transmitted to humans through the bite of an infected mosquito, primarily *Culex* species. Wading birds are the main animal reservoir for the virus, but the presence of pigs greatly amplifies the transmission of JEV.•Humans are a dead-end host in the JEV transmission cycle. ### Occurrence {#cesec103} •JEV is the most common cause of encephalitis in Asia, occurring throughout most of Asia and parts of the western Pacific ([Map 2-5](#f8){ref-type="fig"} ). JEV has not been locally transmitted in Africa, Europe, or the Americas.Map 2-5Geographic distribution of Japanese encephalitis.•JEV transmission principally occurs in rural agricultural areas, often associated with rice production and flooding irrigation. In some areas of Asia, these ecologic conditions may occur near or occasionally within urban centers.•In temperate areas of Asia, transmission is seasonal, and human disease usually peaks in summer and fall. In the subtropics and tropics, seasonal transmission varies with monsoon rains and irrigation practices and may be extended or even occur year-round.•In endemic countries, JE is primarily a disease of children. However, travel-associated JE can occur among persons of any age. ### Risk for Travelers {#cesec104} •The risk for JE for most travelers to Asia is extremely low but varies according to season, destination, duration, and activities. Fewer than 40 cases of confirmed JE have been reported in travelers in the last 40 years.•The overall incidence of JE reported among people from nonendemic countries traveling to Asia is \<1 case per 1 million travelers. However, expatriates and travelers staying for prolonged periods in rural areas with active JEV transmission are likely at similar risk as the susceptible resident population (0.1--2 cases per 100,000 persons per week).•Travelers on even brief trips are probably at increased risk if they have extensive outdoor or nighttime exposure in rural areas, including persons staying in resort areas or with family.•Short-term travelers whose visits are restricted to major urban areas are at very minimal risk for JE.•In endemic areas where there are few human cases among residents because of vaccination or natural immunity, JEV is often maintained in an enzootic cycle between animals and mosquitoes. Therefore, susceptible visitors still may be at risk for infection. ### Clinical Presentation {#cesec105} •Most human infections with JEV are asymptomatic; \<1% of people infected with JEV develop clinical disease.•Acute encephalitis is the most commonly recognized clinical manifestation of JEV infection. Milder forms of disease such as aseptic meningitis or undifferentiated febrile illness can also occur.•The incubation period is 5--15 days. Illness usually begins with sudden onset of fever, headache, and vomiting. Mental status changes, focal neurologic deficits, generalized weakness, and movement disorders may develop over the next few days.○A parkinsonian syndrome resulting from extrapyramidal involvement is a very distinctive clinical presentation of JE.○Acute flaccid paralysis, with clinical and pathological features similar to poliomyelitis, has also been associated with JEV infection.○Seizures are very common, especially among children.•Clinical laboratory findings include moderate leukocytosis, mild anemia, hyponatremia, and cerebrospinal fluid (CSF) pleocytosis with a lymphocytic predominance.•Case--fatality ratio is approximately 20%--30%. Among survivors, 30%--50% may still have significant neurologic or psychiatric sequelae, even years after their acute illness. ### Diagnosis {#cesec106} •JE should be suspected in a patient with evidence of a neurologic infection (e.g., encephalitis, meningitis, or acute flaccid paralysis) who has recently traveled or resided in an endemic country in Asia or the western Pacific.•Laboratory diagnosis of JEV infection should be performed by using JE-specific IgM-capture enzyme-linked immunosorbent assay (ELISA) on CSF or serum. JE-specific IgM antibodies will be present in the CSF or blood of almost all patients by 7 days following onset of symptoms. A fourfold or greater rise in JEV-specific neutralizing antibodies between acute- and convalescent-phase serum specimens may be used to confirm the diagnosis.•Vaccination history, date of onset of symptoms, and information regarding other flaviviruses known to circulate in the geographic area that may cross-react in serologic assays need to be considered when interpreting results.•Humans have low levels of transient viremia and usually have neutralizing antibodies by the time distinctive clinical symptoms are recognized. Virus isolation and nucleic-acid amplification tests (NAATs) are insensitive for the detection of JEV or JE viral RNA in blood or CSF and should not be used for ruling out a diagnosis of JE.•Health-care providers should contact their state or local health department or CDC\'s Division of Vector Borne Infectious Diseases at 970-221-6400 for assistance with diagnostic testing. ### Treatment {#cesec107} There is no specific antiviral treatment for JE; therapy consists of supportive care and management of complications. ### Preventive Measures for Travelers {#cesec108} #### Personal Protection Measures {#cesec109} •No drugs for preventing JEV infection are available.•The best way to prevent mosquito-borne diseases, including JE, is to avoid mosquito bites (see the [Protection Against Mosquitoes, Ticks, and Other Insects and Arthropods](#subchapter29){ref-type="sec"} section later in this chapter). #### JE Vaccines {#cesec110} •An inactivated mouse brain-derived JE vaccine (JE-VAX, manufactured by sanofi pasteur) has been licensed for use in adult and pediatric travelers (Δ1 year of age) in the United States since 1992. Although production of JE-VAX was discontinued in 2006, stockpiles of the vaccine will be used for U.S. travelers until they are depleted.•An inactivated cell culture-derived JE vaccine (IXIARO, manufactured by Intercell) was approved for adult travelers (Δ18 years of age) in the United States on March 30, 2009. Recommendations for its use will be available at [www.cdc.gov/travel](http://www.cdc.gov/travel){#interref22}. Other inactivated and live attenuated JE vaccines are manufactured and used in Asia but not licensed for use in the United States. ##### Inactivated Mouse Brain-Derived JE Vaccine (JE-VAX) {#cesec111} •A randomized controlled trial among 65,000 children in Thailand showed an efficacy of 91% (95% CI 70%--97%) after two doses.•From 88% to 100% of adults from nonendemic settings developed neutralizing antibodies after receiving three doses of vaccine.•The duration of protection after primary immunization is unknown, but circulating neutralizing antibodies appear to last for at least 2--3 years.•Booster doses produce an anamnestic response in neutralizing antibody titers. ##### Recommendations for the Use of JE Vaccine for Travelers {#cesec112} •Decisions regarding the use of JE vaccine for travelers must balance the low risk for disease and the small chance of an adverse event following immunization.•The U.S. Advisory Committee on Immunization Practices (ACIP) currently recommends JE vaccine for travelers who plan to spend a month or longer in endemic areas or areas with ongoing transmission. Vaccine should also be considered for shorter-term travelers whose itineraries may put them at increased risk for JEV exposure, such as rural stays during the rainy season.•Evaluation of an individual traveler\'s risk should take into account itinerary, activities, and best-available information on the current level of JE activity in the travel area ([Table 2-15](#cetable15){ref-type="table"} ). Complicating the concept of risk assessment, however, is the fact that sporadic cases of JE have very rarely occurred in travelers whose itineraries would not ordinarily have indicated a risk for JE (e.g., a resort hotel in Bali, a standard tour of China).Table 2-15**Risk for Japanese encephalitis, by country**[1](#cetablefn62){ref-type="table-fn"}CountryAffected AreasTransmission SeasonComments**Australia**Outer islands of Torres StraitDecember to May; all human cases reported from February to AprilOne human case reported from north Queensland mainland**Bangladesh**Limited data; probably widespreadUnknown; most human cases reported from May to OctoberOne outbreak of human disease reported from Tangail District in 1977. Sentinel surveillance has identified human cases in Chittagong, Khulna, and Rajshahi divisions, and Mymensingh district.**Bhutan**No dataNo data**Brunei**No data; presumed to be endemic countrywideUnknown; presumed year-round transmission**Burma (Myanmar)**Limited data; presumed be endemic countrywideUnknown; most human cases reported from May to OctoberOutbreaks of human disease documented in Shan State. JEV antibodies documented in animals and humans in other areas**Cambodia**Presumed to be endemic countrywideProbably year round with peaks reported from May to OctoberSentinel surveillance has identified human cases in at least 14 provinces including Phnom Penh, Takeo, Kampong, Cham, Battambang, Svay Rieng, and Siem Reap.**China**Human cases reported from all provinces except Xizang (Tibet), Xinjiang, and Qinghai*Hong Kong and Macau:* Not considered endemic. Rare cases reported from the New TerritoriesMost human cases reported from April to OctoberHighest rates reported from the southwest and south central provincesVaccine not routinely recommended for travel limited to Beijing or other major cities**India**Human cases reported from all states except Dadra, Daman, Diu, Gujarat, Himachal, Jammu, Kashmir, Lakshadweep, Meghalaya, Nagar Haveli, Punjab, Rajasthan, and SikkimMost human cases reported from May to October especially in northern India. The season may be extended or year round in some areas especially in southern India.Highest rates of human disease reported from the states of Andhra Pradesh, Assam, Bihar, Goa, Haryana, Karnataka, Kerala, Tamil Nadu, Uttar Pradesh, and West Bengal**Indonesia**Presumed to be endemic countrywideHuman cases reported year round; peak season varies by islandSentinel surveillance has identified human cases in Bali, Kalimantan, Java, Nusa Tenggara, Papua, and Sumatra.**Japan**[2](#cetablefn63){ref-type="table-fn"}Rare sporadic human cases on all islands except Hokkaido. Enzootic activity ongoingMost human cases reported from May to OctoberLarge number of human cases reported until routine JE vaccination introduced in 1968. Most recent small outbreak reported from Chugoku district in 2002. Sporadic cases reported among U.S. military personnel on Okinawa. Enzootic transmission without human cases observed on HokkaidoVaccine not routinely recommended for travel limited to Tokyo or other major cities**Korea, North**No dataNo data**Korea, South**[2](#cetablefn63){ref-type="table-fn"}Rare sporadic human cases countrywide. Enzootic activity ongoingMost human cases reported from May to OctoberLarge number of human casesreported until routine JE vaccinationintroduced in 1968. Highest rates ofdisease were reported from thesouthern provinces. Last majoroutbreak reported in 1982Vaccine not routinely recommendedfor travel limited to Seoul or othermajor cities**Laos**No data; presumed to be endemic countrywidePresumed to be May to October**Malaysia**Endemic in Sarawak; sporadic cases or outbreaks reported from all states of Peninsula, and probably SabahYear-round transmissionMost human cases reported from Penang and SarawakVaccine not routinely recommended for travel limited to Kuala Lumpur or other major cities**Mongolia**Not considered endemic**Nepal**Endemic in southern lowlands (Terai). Sporadic cases or outbreaks reported from the Kathmandu valleyMost human cases reported from May to NovemberHighest rates of human disease reported from western Terai districts, including Bankey, Bardia, Dang, and KailaliVaccine not routinely recommended for travel limited to high-altitude areas**Pakistan**Limited data; human cases reported from around KarachiMost human cases reported from May to October**Papua New Guinea**Limited data; sporadic human cases reported from Western, Gulf, and South Highland ProvincesUnknownA case of JE was reported from near Port Moresby in 2004. Human cases documented in Papua Indonesia**Philippines**Limited data; presumed to be endemic on all islandsUnknown; probably year-roundOutbreaks reported in Nueva Ecija, Luzon, and Manila**Russia**Rare human cases reported from the Far Eastern maritime areas south of KhabarouskMost human cases reported from July to September**Singapore**Rare sporadic human cases reportedYear-round transmissionVaccine not routinely recommended**Sri Lanka**Endemic countrywide except in mountainous areasYear-round with variable peaks based on monsoon rainsHighest rates of human disease reported from Anuradhapura, Gampaha, Kurunegala, Polonnaruwa, and Puttalam districts**Taiwan**[2](#cetablefn63){ref-type="table-fn"}Rare sporadic human cases island-wideMost human cases reported from May to OctoberLarge number of human cases reported until routine JE vaccination introduced in 1968Vaccine not routinely recommended for travel limited to Taipei or other major cities**Thailand**Endemic countrywide; seasonal epidemics in the northern provincesYear-round with seasonal peaks from May to October, especially in the northHighest rates of human disease reported from the Chiang Mai Valley. Sporadic human cases reported from Bangkok suburbs**Timor-Leste**Limited data; anecdotal reports of sporadic human casesNo data**Vietnam**Endemic countrywide; seasonal epidemics in the northern provincesYear-round with seasonal peaks from May to October, especially in the northHighest rates of disease in the northern provinces around Hanoi and northwestern provinces bordering China**Western Pacific Islands**Outbreaks of human disease reported in Guam in 1947--1948 and Saipan in 1990Unknown; most human cases reported from October to MarchEnzootic cycle might not be sustainable; outbreaks may follow introductions of JE virus.[^72][^73] ##### Vaccine Safety and Adverse Events {#cesec113} •Inactivated mouse brain-derived JE vaccine has been associated with localized erythema, tenderness, and swelling at the injection site in about 20% of recipients.•Mild systemic side effects (e.g., fever, chills, headache, rash, myalgia, gastrointestinal symptoms) have been reported in approximately 10% of vaccinees.•Serious allergic hypersensitivity reactions, including generalized urticaria and angioedema of the extremities, face, and oropharynx, have been reported. Accompanying bronchospasm, respiratory distress, and hypotension have been observed in some of these patients, although there have been no fatalities among vaccine recipients with these symptoms.•Estimates of the frequency of these reactions range from 20 to 600 cases per 100,000 vaccinees and vary by country, year, case definition, surveillance method, and vaccine lot.•Most hypersensitivity reactions occur within 24--48 hours after the first dose, when they occur following a subsequent dose, the onset of symptoms is often delayed (median: 3 days; range: up to 2 weeks).•Most reactions can be treated with antihistamines or corticosteroids on an outpatient basis; however, up to 10% of vaccinees with rare severe reactions are hospitalized.•At least four deaths due to anaphylactic shock temporally associated with receipt of this vaccine have been reported in the world literature, which includes endemic country national vaccine programs. None of these patients had evidence of urticaria or angioedema, and two had received other vaccines simultaneously.•Moderate to severe neurologic symptoms, including encephalitis, seizures, gait disturbances, and parkinsonian syndrome have been reported, with an incidence of 0.1 to 2 cases per 100,000 vaccinees.•In addition, there have been case reports of children in Japan and Korea with severe or fatal acute disseminated encephalomyelitis (ADEM) temporally associated with JE vaccination. ##### Vaccine Dose and Administration {#cesec114} •For travelers Δ3 years of age, the recommended primary immunization series for JE-VAX is three doses of 1.0 mL each, administered subcutaneously on days 0, 7, and 30.•An abbreviated schedule (days 0, 7, and 14) provides similar rates of seroconversion but significantly lower neutralizing antibody titers.•Immunization routes and schedules for children 1 and 2 years of age are identical except that 0.5-mL doses should be administered.•Vaccine recipients should be observed for a minimum of 30 minutes after immunization and warned about the possibility of delayed allergic reactions.•The last dose should be administered at least 10 days before beginning travel to ensure an adequate immune response and access to medical care in the event of any delayed adverse reactions.•Booster doses may be administered 2--3 years after the primary series. The timing and immune response of subsequent boosters have not been studied in travelers. ##### Precautions and Contraindications {#cesec115} •A history of allergy or hypersensitivity reaction to a previous dose of mouse brain-derived JE vaccine is a contraindication to receiving additional doses.•Proven or suspected hypersensitivity to thimerosal or proteins of rodent or neural origin is a contraindication to vaccination.•Persons with a previous history of urticaria are more likely to develop a hypersensitivity reaction following receipt of JE vaccine. This history should be considered when weighing the risks and benefits of the vaccine for an individual patient.•No specific information is available on the safety of JE vaccine in pregnancy. Therefore, the vaccine should not be routinely administered during pregnancy. Pregnant women who must travel to an area where risk for JE is high should be vaccinated when the theoretical risk for immunization is outweighed by the risk for infection.•No data are available on vaccine safety and efficacy in infants \<1 year of age.•Two small studies of inactivated JE vaccine in children with underlying medical conditions did not show a change in the adverse reactions or immune response after vaccination. MENINGOCOCCAL DISEASE {#subchapter8} ===================== Cohn Amanda Jackson Michael L. Infectious Agent {#cesec116} ---------------- •The infectious agent is a gram-negative diplococci, *Neisseria meningitidis*. Meningococci are classified into serogroups on the basis of the composition of the capsular polysaccharide.•The five major meningococcal serogroups associated with disease are A, B, C, Y, and W-135. Mode of Transmission {#cesec117} -------------------- Person-to-person transmission occurs by close contact with respiratory secretions or saliva. Occurrence {#cesec118} ---------- •*Neisseria meningitidis* is found worldwide. At any time, 5%--10% of the population may be carriers of *N. meningitidis*.•Invasive disease is much rarer, occurring at a rate of 0.5--10 cases per 100,000 population in nonepidemic areas and up to 1,000 cases per 100,000 population in epidemic regions.•The incidence of meningococcal disease is highest in the "meningitis belt" of sub-Saharan Africa ([Map 2-6](#f9){ref-type="fig"} ). The incidence of meningococcal disease is several times higher in the meningitis belt than in the United States, with periodic epidemics during the dry season (December--June). During nonepidemic periods the rate of meningococcal disease is roughly 5--10 cases per 100,000 population per year. During epidemics the rate can be as high as 1,000 cases per 100,000 population.Map 2-6Areas with frequent epidemics of meningococcal meningitis.•Serogroup A predominates in the meningitis belt, although serogroups C, X, and W-135 are also found.•Young children have the highest risk for meningococcal disease. Risk for Travelers {#cesec119} ------------------ •Travelers to the meningitis belt may be at risk for meningococcal disease, particularly during the dry season.•Risk is likely highest in travelers who will have prolonged contact with local populations in the meningitis belt during an epidemic.•The incidence of meningococcal disease in international travelers who acquire sporadic disease is very low, estimated at 0.4 per 100,000 in one retrospective study.•The Hajj pilgrimage to Saudi Arabia has been associated with outbreaks of meningococcal disease in returning pilgrims and their contacts. Clinical Presentation {#cesec120} --------------------- •Meningococcal disease generally occurs 1--14 days after exposure.•Meningococcal disease presents as meningitis in 50% or more of cases. Meningococcal meningitis is characterized by sudden onset of headache, fever, and stiffness of the neck, sometimes accompanied by nausea, vomiting, photophobia, and/or altered mental status.•Up to 20% of persons with meningococcal disease present with meningococcal sepsis. Meningococcal sepsis is characterized by an abrupt onset of fever and a petechial or purpuric rash. The rash may progress to purpura fulminans. Meningococcal sepsis may often involve hypotension, acute adrenal hemorrhage, and multiorgan failure.•Among infants and children \<2 years of age, meningococcal disease may have nonspecific symptoms. Neck stiffness may be absent. Diagnosis {#cesec121} --------- •Early diagnosis and treatment are critical.•If possible, a lumbar puncture should be done before starting antibiotic therapy to ensure that bacteria, if any, can be cultured from cerebrospinal fluid (CSF).•Diagnosis is generally made isolating *N. meningitidis* from blood or CSF, by detecting meningococcal antigen in CSF by latex agglutination, or by evidence of *N. meningitidis* DNA by polymerase chain reaction.•The signs and symptoms of meningococcal meningitis are similar to those of other causes of bacterial meningitis, such as *Haemophilus influenzae* and *Streptococcus pneumoniae*. Identification of the causative organism is important for selecting the correct antibiotics for treatment and prophylaxis. Treatment {#cesec122} --------- •Invasive meningococcal disease is potentially fatal and should always be viewed as a medical emergency.•Antibiotic treatment must be started early in the course of the disease. Several antibiotic choices are available, including ceftriaxone, chloramphenicol, cefotaxime, and benzylpenicillin. Preventive Measures for Travelers {#cesec123} --------------------------------- ### Vaccine {#cesec124} •ACIP recommends vaccination against meningococcal disease to persons who travel to or reside in countries where *N. meningitidis* is hyperendemic or epidemic, particularly if contact with the local population will be prolonged.•Vaccination is advised for persons traveling to the meningitis belt of Africa during the dry season (December through June).•Advisories for travelers to other countries will be issued when epidemics of meningococcal disease caused by vaccine-preventable serogroups are recognized (see the CDC Travelers\' Health website at [www.cdc.gov/travel](http://www.cdc.gov/travel){#interref23}).•Quadrivalent meningococcal polysaccharide--protein conjugate vaccine (MCV4) is licensed for use among persons 2--55 years of age.•Quadrivalent meningococcal polysaccharide vaccine (MPVS4) is licensed for use among persons 2 years of age or older.•Both vaccines protect against meningococcal disease caused by serogroups A, C, Y, and W-135. Approximately 7--10 days are required following vaccination for development of protective antibody levels.•MCV4 is the preferred vaccine for persons 2--55 years of age; MPSV4 should be used for persons \>55 years of age. There is no licensed vaccine for persons \<2 years old, but MPSV4 is safe to give to children \<2 years of age who require vaccination before traveling to Mecca for the Hajj pilgrimage. #### Requirement for Travel {#cesec125} Proof of quadrivalent vaccination against meningococcal disease is required for persons traveling to Mecca during the annual Hajj and Umrah pilgrimage. ### Antibiotic Chemoprophylaxis {#cesec126} •Antibiotic chemoprophylaxis among close contacts of a patient with invasive meningococcal disease is recommended for prevention of secondary cases in the United States and most industrialized countries.•Antibiotic regimens for prophylaxis include rifampin, ciprofloxacin, and ceftriaxone. Ceftriaxone is recommended for pregnant women. RABIES {#subchapter9} ====== Rupprecht Charles E. Shlim David R. Infectious Agent {#cesec127} ---------------- Rabies is an acute, progressive, fatal encephalomyelitis caused by neurotropic viruses in the family *Rhabdoviridae*, genus *Lyssavirus*. Regardless of the viral variant found throughout the world, all lyssaviruses cause rabies. Mode of Transmission {#cesec128} -------------------- •The disease is almost always transmitted by an animal bite that inoculates virus into wounds.•Very rarely, rabies virus has been transmitted by exposures other than bites that introduce the virus into open wounds or mucous membranes.•All mammals are believed to be susceptible, but reservoirs are carnivores and bats.•Although dogs are the main reservoir in developing countries, the epidemiology of the disease differs sufficiently enough from one region or country to another to warrant the medical evaluation of all mammal bites.•Bat bites anywhere in the world are a cause of concern and an indication for prophylaxis. Pathophysiology of Rabies {#cesec129} ------------------------- •Rabies virus is present in the saliva of the biting mammal.•The virus that is inoculated into the wound does not enter the bloodstream. The virus must be taken up at a nerve synapse to travel to the brain, where it causes a fatal encephalitis. The virus may enter a nerve rapidly, or it may remain at the bite site for an extended period before gaining access to the nervous system. The density of nerve endings in the region of the bite increases the risk of developing rabies encephalitis more rapidly. The hands and face, because of the density of nerve endings, are considered higher-risk exposures.•Prevention of rabies encephalitis is dependent upon preventing the virus from entering a peripheral nerve. This can be accomplished by wound cleansing and disinfection, instillation of rabies virus neutralizing antibodies into the wound, and stimulating an active immune response with a series of rabies vaccine injections. Occurrence {#cesec130} ---------- •Rabies is found on all continents except Antarctica.•In certain areas of the world, canine rabies remains highly endemic, including but not limited to parts of Africa, Asia, and Central and South America. [Table 2-16](#cetable16){ref-type="table"} lists countries that have reported no cases of rabies during the most recent period for which information is available (formerly referred to as "rabies-free" countries).Table 2-16Countries and political units reporting no indigenous cases of rabies during 2005[1](#cetablefn64){ref-type="table-fn"}RegionCountriesAfricaCape Verde, Libya, Mauritius, Réunion, São Tomé and Príncipe, and SeychellesAmericasNorth: Bermuda, St. Pierre and Miquelon\ \ **Caribbean:** Antigua and Barbuda, Aruba, Bahamas, Barbados, Cayman Islands, Dominica, Guadeloupe, Jamaica, Martinique, Montserrat, Netherlands Antilles, Saint Kitts (Saint Christopher) and Nevis, Saint Lucia, Saint Martin, Saint Vincent and Grenadines, Turks and Caicos, and Virgin Islands (UK and US)\ \ **South:** Uruguay**Asia**Hong Kong, Japan, Kuwait, Lebanon, Malaysia (Sabah), Qatar, Singapore, United Arab Emirates**Europe**Austria, Belgium, Cyprus, Czech Republic,[2](#cetablefn65){ref-type="table-fn"} Denmark,[2](#cetablefn65){ref-type="table-fn"} Finland, France,[2](#cetablefn65){ref-type="table-fn"} Gibraltar, Greece, Iceland, Ireland, Isle of Man, Italy, Luxemburg, Netherlands,[2](#cetablefn65){ref-type="table-fn"} Norway, Portugal, Spain[2](#cetablefn65){ref-type="table-fn"} (except Ceuta/Melilla), Sweden, Switzerland, and United Kingdom[2](#cetablefn65){ref-type="table-fn"}**Oceania**[3](#cetablefn66){ref-type="table-fn"}Australia,[2](#cetablefn65){ref-type="table-fn"} Northern Mariana Islands, Cook Islands, Fiji, French Polynesia, Guam, Hawaii, Kiribati, Micronesia, New Caledonia, New Zealand, Palau, Papua New Guinea, Samoa, and Vanuatu[^74][^75][^76]•Additional information about the global occurrence of rabies can be obtained from:○The World Health Organization ([www.who.int/rabies/rabnet/en/](http://www.who.int/rabies/rabnet/en/){#interref24})○The Pan American Health Organization ([www.paho.org/english/ad/dpc/vp/rabia.htm](http://www.paho.org/english/ad/dpc/vp/rabia.htm){#interref25})○The Rabies Bulletin---Europe ([www.rbe.fli.bund.de](http://www.rbe.fli.bund.de){#interref26})○The World Organization for Animal Health ([www.oie.int/eng/en_index.htm](http://www.oie.int/eng/en_index.htm){#interref27})○Other sources are local health authorities of the country, the embassy, or the local consulate\'s office in the United States. Lists are provided only as a guide, because up-to-date information may not be available, surveillance standards vary, and reporting status can change suddenly as a result of disease re-introduction or emergence. Risk for Travelers {#cesec131} ------------------ •The actual rate of possible rabies exposure in tourists has not been calculated with accuracy. However, studies have found a range of roughly 16 to 200 per 100,000 based on differing criteria.•Travelers to rabies-endemic countries should be warned about the risk of acquiring rabies and educated in animal bite prevention strategies.•Street dogs represent the most frequent risk for bite exposure to travelers, followed by monkeys, especially those that live near temples in parts of Asia. Travelers should be instructed not to approach these animals and to be aware of their surroundings so that they do not surprise a dog in a confined space. If a dog is charging at a person, stooping to pick up a rock (or pretending to pick up a rock) can often make the dog turn and run away.•Monkeys are attracted to food and may jump on travelers\' backs if there is food in their backpacks.•Children are considered to be at high risk from rabies virus exposures because their small stature makes extensive bites more likely, they are attracted to animals, and there is the remote possibility that they may not report a possible exposure.•Casual exposure to cave air is not a concern, but cavers should be warned not to handle bats. Noncavers can occasionally encounter a bat. Many bats have tiny teeth, and not all wounds may be appreciated compared with the lesions caused by carnivores. Any suspected or documented bite or scratch from a bat should be grounds for seeking postexposure rabies immunoprophylaxis. Clinical Presentation {#cesec132} --------------------- •After infection, the incubation period is highly variable, but lasts approximately 1--3 months.•The disease progresses from a nonspecific prodromal phase to paresis or paralysis; spasms of swallowing muscles can be stimulated by the sight, sound, or perception of water (hydrophobia); and delirium and convulsions can develop, followed rapidly by coma and death. Diagnosis {#cesec133} --------- •Definitive diagnosis can be made by demonstrating the presence of the rabies virus in corneal impressions or nuchal biopsy, either through staining or polymerase chain reaction.•A serologic response to rabies virus can also prove the diagnosis. Treatment {#cesec134} --------- No treatment is effective after the development of clinical signs, but the extremely rare case of recovery after extensive medical intervention offers hope that future experimental therapeutics may be developed. Preventive Measures for Travelers {#cesec135} --------------------------------- •Prevention of possible exposures to rabies virus is best accomplished by avoiding bites from mammals (mainly dogs, monkeys, bats, and cats in some countries).•Although licks from animals to fresh wounds or mucus membranes of humans are a theoretical risk of acquiring rabies and postexposure prophylaxis should be considered, there are no examples of rabies in travelers who were exposed in this way.•Travelers should be counseled to avoid approaching stray animals, to be aware of their surroundings so that they do not accidentally surprise a stray dog, to avoid contact with bats, and not to carry or eat food while walking among monkeys.•In addition to avoiding bite exposures, two strategies are available to travelers for the prevention of rabies: vaccination and management of a possible rabies exposure (see sections below for more specific details).•For certain international travelers, pre-exposure rabies vaccine may be recommended, based on the local incidence of rabies in the country to be visited, the availability of appropriate anti-rabies biologicals, and the intended activity and duration of stay of the traveler.○A decision to receive pre-exposure rabies immunization may also be based on the likelihood of repeat travel to at-risk destinations over time or taking up residence in a high-risk destination.○Pre-exposure vaccination may be recommended for veterinarians, animal handlers, field biologists, cavers, missionaries, and certain laboratory workers. [Table 2-17](#cetable17){ref-type="table"} provides criteria for pre-exposure vaccination.Table 2-17Criteria for pre-exposure immunization for rabiesRisk CategoryNature of RiskTypical PopulationsPre-exposure RegimenContinuousVirus present continuously, often in high concentrations\ \ Specific exposures likely to go unrecognized\ \ Bite, nonbite, or aerosol exposureRabies research laboratory workers,[1](#cetablefn67){ref-type="table-fn"} rabies biologics production workersPrimary course: serologic testing every 6 months; booster vaccination if antibody titer is below acceptable level[2](#cetablefn68){ref-type="table-fn"}FrequentExposure usually episodic with source recognized, but exposure might also be unrecognized\ \ Bite, nonbite, or aerosol exposure possibleRabies diagnostic laboratory workers[1](#cetablefn67){ref-type="table-fn"}, cavers, veterinarians and staff, and animal control and wildlife workers in rabies-epizootic areasPrimary course: serologic testing every 2 years; booster vaccination if antibody titer is below acceptable level[2](#cetablefn68){ref-type="table-fn"}Infrequent (greater than general population)Exposure nearly always episodic with source recognized\ \ Bite or nonbite exposureVeterinarians, animal control and wildlife workers in areas with low rabies rates; veterinary students; and travelers visiting areas where rabies is enzootic and immediate access to appropriate medical care, including biologics, is limitedPrimary course: no serologic testing or booster vaccinationRare (general population)Exposure always episodic, with source recognizedU.S. population at large, including individuals in rabies-epizootic areasNo pre-exposure immunization necessary[^77][^78]•Immediate and adequate medical care after an animal bite is critical to preventing rabies (see [Management of a Possible Rabies Exposure](#cesec139){ref-type="sec"} below). Regardless of whether or not pre-exposure vaccine is administered, travelers going to areas with a high risk for rabies should be especially encouraged to purchase medical evacuation insurance (see the [Travel Insurance and Evacuation Insurance](#subchapter45){ref-type="sec"} section later in this chapter). ### Vaccine {#cesec136} #### Pre-Exposure Vaccination {#cesec137} •In the United States, pre-exposure vaccination consists of a series of three injections with human diploid cell rabies vaccine (HDCV) or purified chick embryo cell (PCEC) vaccine.•The schedule for this series in given in [Table 2-18](#cetable18){ref-type="table"}.•In the event of a possible rabies exposure in someone who has received pre-exposure rabies immunization, rabies immune globulin (RIG) is not given, and two boosters of an acceptable rabies vaccine are given on days 0 and 3. The booster doses need to be modern cell culture vaccines, but they do not need to be the same brand as the vaccine given in the pre-exposure immunization series.•Pre-exposure immunization does not eliminate the need for additional medical attention after a rabies exposure, but it greatly simplifies postexposure prophylaxis.•Pre-exposure vaccination may also provide some degree of protection when there is an unapparent or unrecognized exposure to rabies virus and when postexposure prophylaxis might be delayed.•Travelers should receive all three pre-exposure immunizations before travel. If three doses of rabies vaccine cannot be completed prior to travel, the traveler should not start the series, as it would be problematic to plan postexposure prophylaxis after a partial immunization series. Beginning in 2007 and anticipated to continue through 2009, the United States has experienced a limitation in supply of rabies vaccine. As a result, during this time, pre-exposure rabies vaccination has not been available to international travelers, except under special circumstances. The existing supplies of rabies vaccine are being reserved for postexposure prophylaxis. #### Postexposure Vaccination {#cesec138} •If pre-exposure rabies immunization is not given, the traveler will need to obtain full postexposure rabies prophylaxis in the event of a possible rabies virus exposure. This consists of injections of RIG (20 IU/kg) and a series of five injections of rabies vaccine over a 1-month period.•Because RIG or rabies vaccine may not be available in the destination country, travelers should have a strategy in place before travel as to how to respond to a possible exposure. This strategy may require the traveler to fly to a different country to obtain the appropriate prophylaxis.•Different postexposure vaccine schedules, alternative routes of administration and other rabies vaccines besides HDCV and PCEC may be found abroad. Although not approved for sale in the United States, purified vero cell rabies vaccine and purified chick embryo cell vaccine (manufactured abroad) are acceptable alternatives if available in a destination country.•Historically, rabies vaccine was once manufactured from viruses grown in animal brains, and some of these vaccines are still in use in developing countries. The brain-derived vaccines can be identified if the traveler is offered a large injection (5 mL) daily for 14--21 days. The traveler should not accept these vaccines, but rather travel to where acceptable vaccines and immune globulin are available. ### Management of a Possible Rabies Exposure {#cesec139} •Travelers should be advised that any animal bite or scratch should receive prompt local first aid by thorough cleansing of the wound with copious amounts of soap and water and povidone iodine, if available. This local care will substantially reduce the risk for rabies.•Wounds that might require suturing should have the suturing delayed for a few days. If suturing is necessary for control of bleeding or for functional or cosmetic reasons, RIG should be administered into the wound before closing the wound. The use of local anesthetic is not contraindicated in wound management.•Human rabies immune globulin (HRIG) is manufactured by plasmaphoresis of hyperimmunized volunteers. The manufactured quantity of HRIG falls short of world-wide requirements, and the substance is not available in many developing countries. Equine rabies immune globulin (ERIG) or purified fractions of ERIG have been used effectively in some developing countries where HRIG might not be available. If necessary, such heterologous products are preferable to no RIG administration in human rabies postexposure prophylaxis.•The incidence of adverse reactions after the use of these products has been low (0.8%--6.0%), and most of those reactions were minor. However, such products are neither evaluated by U.S. standards nor regulated by the FDA, and their use cannot be unequivocally recommended at this time. In addition, unpurified antirabies serum of equine origin might still be used in some countries where neither HRIG nor ERIG is available. The use of this antirabies serum is associated with higher rates of serious adverse reactions, including anaphylaxis.•After wound cleansing, as much of the calculated amount of RIG (see [Table 2-19](#cetable19){ref-type="table"}) as is anatomically feasible is infiltrated around the wound. The dose injected around the wound may be as small as 0.5 mL if the wound is small or on a finger. If the wounds are extensive, the calculated dose of RIG must not be exceeded. If the calculated dose is inadequate to inject all the wounds, the RIG should be diluted with normal saline to extend the number of wounds that can be injected. This is a particular issue in children, whose body weight may be small in relation to the size and number of wounds.•The remainder of the RIG dose, if any, should be injected intramuscularly. Care should be taken to guarantee that this remaining amount of RIG is deposited in a muscle and not injected subcutaneously, which may decrease its effectiveness. The remaining RIG can be given in the deltoid muscle, on the opposite side of the initial vaccine dose. The anterior thigh is also an alternative site.•RIG should not be given more than 7 days after the start of the postexposure vaccine series. This 7-day period does not relate to the time of the bite exposure itself.•Postexposure prophylaxis, including RIG, should be initiated after a possible bite exposure even if there has been a considerable delay between the exposure and the traveler presenting for evaluation.•Travelers who have completed a three-dose pre-exposure rabies immunization series or have received the full postexposure prophylaxis are considered pre-immunized and do not require routine boosters, except after a possible rabies exposure.•Periodic serum testing for rabies antibody is not necessary in routine international travelers. ### Vaccine Safety and Adverse Reactions {#cesec140} •Travelers should be advised that they may experience local reactions after vaccination, such as pain, erythema, swelling, or itching at the injection site, or mild systemic reactions, such as headache, nausea, abdominal pain, muscle aches, and dizziness.•Approximately 6% of persons receiving booster vaccinations with HDCV may experience an immune complex-like reaction characterized by urticaria, pruritus, and malaise. The likelihood of these reactions is less with PCECV.•Once initiated, rabies postexposure prophylaxis should not be interrupted or discontinued because of local or mild systemic reactions to rabies vaccine. ### Precautions and Contraindications {#cesec141} •Pregnancy is not a contraindication to postexposure prophylaxis.•In infants and children, the dose of HDCV or PCEC for pre-exposure or postexposure prophylaxis is the same as that recommended for adults. The dose of RIG for postexposure prophylaxis is based on body weight ([Table 2-19](#cetable19){ref-type="table"}).   {#cesec142} = Routine Vaccine-Preventable Diseases DIPHTHERIA {#subchapter10} ========== Tiwari Tejpratap S.P. Infectious Agent {#cesec143} ---------------- •Diphtheria is caused by toxigenic strains of *Corynebacterium diphtheriae* biotype mitis, gravis, intermedius, or belfanti.•The bacteria produce an exotoxin which, if absorbed in the bloodstream, may damage organs such as the heart, kidneys, and nerves. Mode of Transmission {#cesec144} -------------------- •Humans are the only known reservoir of *C. diphtheriae*.•Person-to-person transmission occurs through oral or respiratory droplets, close physical contact, and rarely by fomites.•Cutaneous diphtheria is common in tropical countries, and contact with discharge from skin lesions may play an important role in transmission of infection in these environments. Occurrence {#cesec145} ---------- •Diphtheria is found worldwide. Countries with endemic diphtheria are shown in [Table 2-20](#cetable20){ref-type="table"} .Table 2-20Countries with endemic diphtheriaRegionsCountries**Africa**Algeria, Angola, Egypt, Niger, Nigeria, Sudan, and sub-Saharan countries**Americas**Bolivia, Brazil, Colombia, Dominican Republic, Ecuador, Haiti, and Paraguay**Asia/South Pacific**Afghanistan, Bangladesh, Bhutan, Burma (Myanmar), Cambodia, China, India, Indonesia, Laos, Malaysia, Mongolia, Nepal, Pakistan, Papua New Guinea, Philippines, Thailand, and Vietnam**Middle East**Iran, Iraq, Saudi Arabia, Syria, Turkey, and Yemen**Europe**Albania, Russia, and countries of the former Soviet Union•Diphtheria causes significant morbidity and mortality in developing countries where vaccination coverage is low.•During the 1990s, large epidemics occurred in the newly independent states of the former Soviet Union. More recently in the Americas, diphtheria outbreaks occurred in Paraguay, the Dominican Republic, and Haiti.•Diphtheria is uncommon in industrialized countries because of long-standing routine use of DTP (diphtheria and tetanus toxoids and pertussis vaccine). Diphtheria is uncommon in the United States; and the last case occurred in an elderly traveler returning from Haiti in 2003. Risk for Travelers {#cesec146} ------------------ •Symptomatic infection is extremely rare in adequately immunized persons, although active immunization with diphtheria toxoid does not prevent colonization or transient carriage of *C. diphtheriae*.•Exposure and higher risk of acquiring disease and potentially life-threatening complications are possible in inadequately immunized or unimmunized travelers to countries with endemic diphtheria. Clinical Presentation {#cesec147} --------------------- •The incubation period is 2--5 days (range 1--10 days).•Nasal diphtheria can be asymptomatic or mild, with a blood-tinged discharge.•Affected anatomic sites include the mucous membrane of the upper respiratory tract (nose, pharynx, tonsils, larynx, and trachea \[respiratory diphtheria\]), skin (cutaneous diphtheria), or rarely, mucous membranes at other sites (eye, ear, vulva).•Respiratory diphtheria has a gradual onset and is characterized by a mild fever (rarely \>101° F or \>38.3° C), sore throat, difficulty in swallowing, malaise, loss of appetite, and if the larynx is involved, hoarseness may occur.•The hallmark of respiratory diphtheria is the presence of a membrane that appears within 2--3 days of illness over the mucous membrane of the tonsils, pharynx, larynx, or nares, and which can extend into the trachea. The membrane is firm, fleshy, grey, and adherent, and bleeds following attempts to remove or dislodge it.•Local complications such as life-threatening or fatal airway obstruction can result from extension of the membrane or dislodgement of a piece of the membrane into the larynx or trachea.•In severe respiratory diphtheria, cervical lymphadenopathy and soft-tissue swelling in the neck give rise to a "bull-neck" appearance.•The case--fatality rate of respiratory diphtheria is 5%--10%.•Systemic complications, including myocarditis, and polyneuropathies, can result from absorption of diphtheria toxin from the infection site. However, cutaneous and nasal diphtheria are localized and rarely associated with systemic toxicity. Diagnosis {#cesec148} --------- •A presumptive diagnosis is usually based on clinical features.•A confirmed diagnosis is made by isolation of *C. diphtheriae* from culture of nasal or throat swabs, or membrane tissue.•Toxin production is confirmed by performing a modified Elek test.•Polymerase chain reaction assays can also be performed on isolates, swabs, or membrane specimens to rapidly confirm the presence of tox gene responsible for production of diphtheria toxin, but the test is available only in research or reference laboratories. Treatment {#cesec149} --------- •Patients with respiratory diphtheria require hospitalization to monitor response to treatment and manage complications.•Equine diphtheria antitoxin (DAT) is the mainstay of treatment and is administered after sensitivity testing, without waiting for laboratory confirmation. In the United States, DAT is available to physicians under an FDA-approved Investigational New Drug protocol by contacting CDC at 770-488-7100.•An appropriate antibiotic (erythromycin or penicillin) to eliminate the causative organisms, stop exotoxin production, and reduce communicability.•Supportive care (airway, cardiac monitoring) is required.•Antimicrobial prophylaxis (erythromycin or penicillin) is recommended for close contacts of patients. Preventive Measures for Travelers {#cesec150} --------------------------------- ### Vaccine {#cesec151} •For protection against diphtheria, all travelers should be up-to-date with diphtheria toxoid vaccine before departure. Diphtheria toxoid is not manufactured as a monovalent vaccine but is available in pediatric (D) and adult formulations (d) that are combined with other vaccines such as tetanus toxoid (DT, Td), or tetanus toxoid and acellular or whole-cell pertussis antigens (DTaP, DTwP, Tdap), or as a DTwP/DTaP combination with other antigens (e.g., hepatitis B, inactivated poliovirus vaccines, or Hib vaccine).•In the United States, infants and children \<7 years of age are vaccinated with diphtheria toxoid in combination with tetanus toxoid and acellular pertussis vaccine (DTaP) according to a routine childhood immunization schedule as recommended by the ACIP (see the Vaccine Recommendations for Infants and Children section in Chapter 7).•Immunization for infants and children \<7 years of age consists of five doses of DTaP vaccine. The first three doses are usually given at ages 2, 4, and 6 months, followed by booster doses at ages 12--18 months and 4--6 years (see Table 7-2).•Adolescents 11--18 years of age should receive a dose of Tdap instead of Td for booster immunization against tetanus, diphtheria, and pertussis if they have completed the recommended childhood DTwP/DTaP vaccination series (see Table 7-3).•Adults 19--64 years of age should receive a single dose of age-appropriate Tdap to replace a single dose of Td for active booster immunization against tetanus, diphtheria and pertussis. Thereafter, routine booster doses with Td should be given every 10 years to maintain seroprotection against diphtheria as well as tetanus. This booster is particularly important for travelers who will live or work with local populations in countries where diphtheria is endemic.•Adults \>65 years of age should receive Td; Tdap is not licensed for this age group.•Persons \>7 years of age and with uncertain vaccination history or who have never been vaccinated against tetanus, diphtheria, or pertussis should receive three doses of an age-appropriate tetanus and diphtheria toxoid-containing vaccine. If a person is 10--65 years old, a single dose of Tdap may be substituted for one Td dose for added protection against pertussis. HUMAN PAPILLOMAVIRUS (HPV) {#subchapter11} ========================== Dunne Eileen F. Infectious Agent {#cesec152} ---------------- •Human papillomavirus (HPV) is in the family Papillomaviridae, a family of DNA viruses that has a double-stranded, closed, circular genome of ∼8 kb and a nonenveloped icosahedral capsid.•Infection with human papillomavirus is specific to humans. Mode of Transmission {#cesec153} -------------------- •There are more than 100 HPV types. Some types cause infection on the skin and others cause infection on the mucosa. More than 40 mucosal HPV types are commonly found on the genitals and are transmitted primarily by sexual contact, most commonly sexual intercourse.•Sometimes transmission of genital HPV types occurs by other routes (e.g., mother-to-child transmission). Occurrence {#cesec154} ---------- •HPV is common worldwide.•Studies in multiple countries demonstrate prevalence of HPV from 3% to 70%. Risk for Travelers {#cesec155} ------------------ There are no inherent risks for travelers. HPV is ubiquitous and common worldwide. Risk depends on the behavior of the traveler. Clinical Presentation {#cesec156} --------------------- •HPV infection is usually subclinical and asymptomatic.•HPV infection with cutaneous types can cause the common skin wart.•HPV infection with mucosal types is presumed when anogenital warts or cervical cell changes are detected in screening.•HPV rarely causes recurrent respiratory papillomatosis (RRP); anogenital cancers such as vaginal, vulvar, anal, penile cancers; and some oral cancers. Diagnosis {#cesec157} --------- •Infection is most commonly asymptomatic and transient.•When clinical disease occurs, diagnosis is usually made by detection of the lesion by visual inspection, and in women by a Papanicolaou test (Pap test), HPV test (done in similar fashion to a Pap test), or by colposcopy. Definitive diagnosis is made by biopsy.•Laboratory diagnosis of HPV in the clinical setting can be made by using the Digene hybrid capture 2 test (hc2); this test is recommended only in the setting of cervical cancer screening. Other DNA and serologic tests are available in research settings. Treatment {#cesec158} --------- There is no treatment for HPV, but there are treatments for HPV-associated conditions such as genital warts and cervical cell changes. Preventive Measures for Travelers {#cesec159} --------------------------------- •There are no recommendations for preventive measures for travelers beyond abstinence or engaging in safe sexual behaviors.•The quadrivalent HPV vaccine prevents four HPV types (types 6, 11, 16, 18) commonly associated with genital warts and cervical cancers. Cervical cancer screening with the Pap test should be continued because the vaccine does not prevent all types associated with cancers.•ACIP recommendations for HPV vaccine are that 11- and 12-year-old girls should be routinely vaccinated, and the vaccine series can be started as young as 9 years of age. Girls and young women 13--26 years of age should be vaccinated if they have not received the vaccine or have not completed the series.•Vaccination consists of three intramuscular doses, on day 0, at 2 months, and 6 months; see product insert for some variability in administration timing. The vaccine is generally well tolerated, although pain at the injection site is common. INFLUENZA (SEASONAL, AVIAN, AND PANDEMIC) {#subchapter12} ========================================= McCarron Margaret Shay David K. Infectious Agent {#cesec160} ---------------- •Influenza is caused by infection with influenza viruses.•Human influenza viruses can be divided into three types: A, B, and C. Only types A and B cause widespread illness in humans.•Influenza A viruses are further classified into subtypes on the basis of two surface proteins: hemagglutinin (H) and neuraminidase (N). There are 16 different hemagglutinin subtypes and 9 different neuraminidase subtypes. Mode of Transmission {#cesec161} -------------------- •Person-to-person transmission results from respiratory droplets of coughs and sneezes.○Most healthy adults can infect others beginning 1 day before symptoms develop and up to 5 days after becoming sick.○Children can pass the virus for longer than 7 days.○Fomite transmission is also possible. Occurrence {#cesec162} ---------- ### Seasonal Influenza {#cesec163} •Infection with seasonal influenza viruses is common.•In temperate climates in the Northern Hemisphere, annual seasonal epidemics of influenza generally occur during the winter months, while in the temperate regions of the Southern Hemisphere most activity occurs from April through September.•In tropical and subtropical areas, influenza can occur throughout the year.•CDC has estimated that U.S. epidemics during the 1990s were associated with an annual average of 36,000 influenza-related deaths.•Influenza virus infections cause disease in all age groups. Rates of infection are highest among infants and children, but rates of serious morbidity and mortality are highest among persons Δ65 years of age and persons of any age who have medical conditions (e.g., chronic cardiopulmonary disease) that place them at increased risk for complications from influenza. Children \<2 years of age have rates of influenza-related hospitalization that are as high as those in the elderly. ### Avian Influenza {#cesec164} •Avian influenza refers to influenza A viruses usually found in birds. Influenza A viruses infect a broad range of avian species and several mammalian species, including humans, swine, and horses.○Most cases of avian influenza infection in humans have resulted from contact with infected poultry (e.g., domesticated chickens, ducks, and turkeys) or surfaces contaminated with secretions or excretions from infected birds.○The spread of avian influenza viruses from one ill person to another has been reported very rarely and has thus far been limited, inefficient, and unsustained.•H5N1 is an influenza A virus, which is a type characterized by the ability to constantly undergo change. H5N1 virus has caused serious disease among wild birds and poultry on multiple continents. Human cases of H5N1 are very rare but have occurred in countries in Asia, Africa, Eastern Europe, and the Middle East since 2003. As of June 2008, only 385 human cases of H5N1 infection have been reported worldwide. These cases, however, are a concern because the mortality rate is high. There is a concern that H5N1 may gain the ability to spread easily between people. Vigilant monitoring for human infection and person-to-person transmission has become an important component of pandemic preparedness. For a current list of countries reporting outbreaks of H5N1 among birds, see the World Organization for Animal Health (OIE) website at [www.oie.int/](http://www.oie.int/){#interref28}.•The global influenza research community conducts surveillance for novel influenza viruses in travelers. In a 2007 review of returned U.S. travelers suspected of having H5N1 infection, no evidence of infection with novel viruses was reported. Surveillance continues for human infection with H5N1 in travelers.•For current, continuously updated information, see CDC\'s Avian Influenza website ([www.cdc.gov/flu/avian/index.htm](http://www.cdc.gov/flu/avian/index.htm){#interref29}). ### Pandemic Influenza {#cesec165} •The emergence of a novel human influenza A virus could lead to a global pandemic, during which rates of morbidity and mortality from influenza-related complications could increase dramatically. The public health threat of a pandemic arising from novel influenza A viruses, including influenza A (H5N1), becomes imminent only if the virus gains the ability to spread efficiently from one human to another.•Such transmission has not yet been observed with the currently circulating A (H5N1) viruses. Although a few cases of limited person-to-person spread of H5N1 viruses have been reported as of June 2008, no instances of transmission continuing beyond one person are thought to have occurred.•Because the situation has and may continue to evolve, for current information see the official U.S. government website for pandemic influenza ([www.pandemicflu.gov/](http://www.pandemicflu.gov/){#interref30}). Risk for Travelers {#cesec166} ------------------ ### Seasonal Influenza {#cesec167} •The risk for exposure to seasonal influenza during international travel depends on the time of year and destination. In the tropics, influenza can occur throughout the year, while in the temperate regions of the Southern Hemisphere most activity occurs from April through September.•In temperate climates, travelers can also be exposed to influenza during the summer, especially when traveling as part of large tourist groups with travelers from areas of the world where influenza viruses are circulating. ### Avian Influenza {#cesec168} •In those countries where H5N1 has occurred most people become infected through direct contact with birds (e.g., domesticated chickens, ducks, and turkeys) that were carrying the H5N1 virus or from surfaces contaminated with secretions or excretions from these birds. Direct contact could happen during activities such as---○Visiting poultry farms○Visiting live bird or poultry markets○Preparing or consuming uncooked or undercooked bird products (such as meat, eggs, or blood).•Because the situation has and may continue to evolve, travelers can stay abreast of new developments by checking the following websites that are updated regularly:○U.S. government website for pandemic influenza ([www.pandemicflu.gov/](http://www.pandemicflu.gov/){#interref31})○CDC\'s Travelers\' Health website ([wwwn.cdc.gov/travel/contentAvianFluInformation.aspx](http://wwwn.cdc.gov/travel/contentAvianFluInformation.aspx){#interref32})○CDC\'s Avian Influenza website ([www.cdc.gov/flu/avian/](http://www.cdc.gov/flu/avian/){#interref33})○WHO website ([www.who.int/csr/disease/avian_influenza/en/index.html](http://www.who.int/csr/disease/avian_influenza/en/index.html){#interref34}). Clinical Presentation {#cesec169} --------------------- •Onset of symptoms typically occurs 1--4 days after infection.•Uncomplicated influenza illness is characterized by the abrupt onset of constitutional and respiratory signs and symptoms (e.g., fever, myalgia, headache, malaise, nonproductive cough, sore throat, and rhinitis). Among children, otitis media, nausea, and vomiting are also commonly reported with influenza illness.•Influenza illness typically resolves within 1 week for most persons, although cough and malaise can persist for \>2 weeks. However, influenza virus infections can cause primary influenza viral pneumonia; exacerbate underlying medical conditions (e.g., pulmonary or cardiac disease); lead to secondary bacterial pneumonia, sinusitis, or otitis; or contribute to co-infections with other viral or bacterial pathogens. Influenza-related deaths can result from primary illnesses, secondary bacterial pneumonia, or exacerbations of chronic cardiac or pulmonary conditions.•Young children with influenza virus infection may have initial symptoms mimicking bacterial sepsis with high fevers, and febrile seizures have been reported in 6%--20% of children hospitalized with influenza virus infection. Population-based studies among hospitalized children with laboratory-confirmed influenza have demonstrated that, although the majority of hospitalizations are brief (2 days or less), 4%--11% of children hospitalized with laboratory-confirmed influenza required treatment in the intensive-care unit and 3% required mechanical ventilation. Among 1,308 hospitalized children in one study, 80% were \<5 years of age and 27% were \<6 months of age. Influenza virus infection also has been uncommonly associated with encephalopathy, transverse myelitis, myositis, myocarditis, pericarditis, and Reye syndrome. Diagnosis {#cesec170} --------- •Respiratory illnesses caused by influenza virus infection are difficult to distinguish from illnesses caused by other respiratory pathogens on the basis of signs and symptoms alone. Sensitivity and predictive value of clinical definitions can vary, depending on the degree of circulation of other respiratory pathogens and the level of influenza activity. Among studies conducted with children and adults, the positive predictive value of clinical signs and symptoms for laboratory-confirmed influenza virus infection has ranged from 30% to 88%.•Laboratory testing can aid in diagnosis. Diagnostic tests available for influenza include viral culture, serology, rapid antigen testing, polymerase chain reaction, and immunofluorescence assays.•Respiratory specimens obtained via nasopharyngeal swabs typically yield better detection of influenza than specimens obtained via oropharyngeal swabs.•Commercial rapid diagnostic tests are available that can detect influenza viruses within 30 minutes. Some tests are approved for use in any outpatient setting, whereas others must be used in a moderately complex clinical laboratory. These rapid tests differ in the types of influenza viruses they can detect and whether they can distinguish between influenza types. Some tests can detect only influenza A viruses, some detect both influenza A and B viruses, but cannot distinguish between the two types, and some detect both influenza A and B and can distinguish between the two.•None of the commercially available rapid tests provides any information about influenza A subtypes. The types of specimens acceptable for use (i.e., throat, nasopharyngeal, or nasal aspirates, swabs, or washes) also vary by test. The specificity and, in particular, the sensitivity of rapid tests are lower than for viral culture and vary by test. Because of the lower sensitivity of the rapid tests, physicians should consider confirming negative tests with viral culture or other means because of the possibility of false-negative rapid test results. Treatment {#cesec171} --------- •Influenza-specific antiviral drugs for chemoprophylaxis of influenza are important adjuncts to the influenza vaccine.•The four currently licensed U.S. antiviral agents are amantadine, rimantadine, zanamivir, and oseltamivir. Amantadine and rimantadine have a mechanism of action effective only against influenza A viruses, while the neuraminidase inhibitors oseltamivir and zanamivir are effective against both influenza A and B viruses.•Antiviral drug testing results conducted at CDC during the 2005--2006 influenza season, indicated 79% resistance among influenza A H3N2 viruses and 10% among H1N1. CDC recommends that neither amantadine nor rimantadine be used for the treatment or chemoprophylaxis of influenza A in the United States until susceptibility to these antiviral medications has been re-established among circulating influenza A viruses.•Oseltamivir or zanamivir can be prescribed if antiviral treatment of influenza is indicated. Oseltamivir is approved for treatment of persons aged Δ1 year, and zanamivir is approved for treatment of persons Δ7 years of age. Oseltamivir and zanamivir can be used for chemoprophylaxis of influenza; oseltamivir is licensed for use in persons Δ1 year of age, and zanamivir is licensed for use in persons Δ5 years of age. These two drugs differ in dosing, approved age groups for use, side effects, and cost. The package inserts should be consulted for more information. ### Antiviral Resistance {#cesec172} •In the winter of 2007--2008, oseltamivir-resistant influenza A (H1N1) viruses were found to be circulating in several countries, including the United States, but the highest rates of resistance were found in Northern Europe. Oseltamivir resistance appears to be geographically variable, both within Europe and globally. In the United States, approximately 12% of H1N1 viruses from the 2007--2008 season were resistant to oseltamivir, resistance was highest in Europe at 26%, and overall global resistance was approximately 16%. The oseltamivir-resistant H1N1 viruses remained sensitive to amantadine and rimantadine. This strain of oseltamivir-resistant influenza A (H1N1) may continue to circulate in future seasons or may spread geographically. Oseltamivir and zanamivir remain the preferred antiviral drugs to be used in the treatment of influenza virus infection, given their relatively low levels of resistance compared with the high resistance found to amantadine and rimantadine among currently circulating influenza A viruses.•Some H5N1 viruses currently infecting birds and humans are resistant to amantadine and rimantadine. Most of the H5N1 viruses tested have been susceptible to the antiviral medications oseltamivir and zanamivir, but resistance has been reported. The effectiveness of antivirals for treating H5N1 virus infections is unknown. For more information about influenza antiviral drugs, see [www.cdc.gov/flu/avian/gen-info/avian-flu-humans.htm\#antiviral](http://www.cdc.gov/flu/avian/gen-info/avian-flu-humans.htm#antiviral){#interref35}.•For more detailed information on influenza vaccines, treatment, and general prevention and control, please refer to "Prevention and Control of Influenza: Recommendations of the Advisory Committee on Immunization Practices (ACIP)" (the most recent version is available at: [www.cdc.gov/vaccines/pubs/ACIP-list.htm](http://www.cdc.gov/vaccines/pubs/ACIP-list.htm){#interref36}). Preventive Measures for Travelers {#cesec173} --------------------------------- •Handwashing and cough hygiene can play important roles in limiting person-to-person transmission of influenza. Where handwashing is not available, use of hand sanitizing gels containing greater than 60% alcohol can be used.•Annual vaccination of persons at high risk for complications and vaccination of health-care workers and close contacts of high risk persons before the influenza season are the most effective measure for preventing seasonal influenza and associated complications.•Vaccination of travelers is recommended when the vaccine is available and if there are no contraindications. ### Vaccine {#cesec174} •Two types of influenza vaccines are currently available for use in the United States: trivalent inactivated vaccine (TIV), administered by intramuscular injection; and live, attenuated influenza vaccine (LAIV), administered by nasal spray. LAIV is approved currently for use only in healthy persons 2--49 years of age who are not pregnant.•In the United States, annual influenza vaccination is recommended by CDC and ACIP for certain groups of people, primarily those who are at high risk of having serious flu complications or those who live with or care for those at high risk for serious complications. Annual recommendations are published by CDC and ACIP, including information about the season\'s vaccine composition, dosage and administration, and recommendations for specific populations. The current version of these routine recommendations is available at [www.cdc.gov/vaccines/pubs/ACIP-list.htm](http://www.cdc.gov/vaccines/pubs/ACIP-list.htm){#interref37}.•The influenza vaccine must be administered annually to optimize protection because vaccine-derived immunity declines over time and because the vaccine strains must be updated regularly to reflect ongoing antigenic changes among circulating influenza viruses.•Dosages differ according to age group and type of vaccine used. For inactivated vaccines, two doses administered at least 1 month apart are required for previously unvaccinated infants and children through 8 years of age. In adults, studies have indicated little or no improvement in antibody response when a second dose of inactivated vaccine is administered during the same season; therefore, a booster is not recommended. Inactivated vaccine should be administered in infants and young children in the anterolateral aspect of the thigh; all other recipients should be vaccinated in the deltoid muscle.•The age groups for which influenza and pneumococcal vaccination are recommended overlap considerably. For travelers at high risk who have not previously been vaccinated with pneumococcal vaccine, health-care providers should strongly consider administering pneumococcal and influenza vaccines concurrently. Both vaccines can be administered at the same time at different sites without increasing side effects. Infants and children can receive influenza vaccine at the same time they receive other routine vaccinations.•Both influenza vaccines contain three strains of influenza viruses. Viruses in inactivated vaccines are killed, while those in LAIV are live. These live viruses are attenuated and do not cause influenza illnesses. The viruses used in both vaccines are representative of viruses likely to circulate in the upcoming season, and usually one or more vaccine strains are updated annually. Because the vaccine is grown in hen eggs, the vaccine may contain small amounts of egg protein. The package insert should be consulted regarding the use of other compounds to inactivate the viruses or to limit bacterial contamination. ### Avian Influenza {#cesec175} •H5N1 infections in humans, though rare, can cause serious disease and death.•CDC advises travelers to countries with known outbreaks of H5N1 to avoid---○Poultry farms○All poultry, whether or not symptomatic, and especially contact with sick or dead poultry○Contact with surfaces that may have been contaminated by poultry feces or secretions○Contact with animals in live food markets•Since transmission of H5N1 viruses to two persons through consumption of uncooked duck blood may have occurred in Vietnam in 2005, uncooked poultry or poultry products, including blood, should not be consumed. Care should be taken when preparing these foods.•For more information, see Human Infection with Avian Influenza A (H5N1) Virus Advice for Travelers ([wwwn.cdc.gov/travel/contentAvianFluAsia.aspx](http://wwwn.cdc.gov/travel/contentAvianFluAsia.aspx){#interref38}) and the WHO Avian Influenza Fact Sheet ([www.who.int/mediacentre/factsheets/avian_influenza/en/index.html\#humans](http://www.who.int/mediacentre/factsheets/avian_influenza/en/index.html#humans){#interref39}).•A vaccine to protect humans against influenza A (H5N1) is not yet available commercially, but candidate vaccines are undergoing human clinical trials in the United States, with one vaccine currently licensed in the United States. This vaccine is approved by the U.S. FDA for stockpiling purposes only. MEASLES (RUBEOLA) {#subchapter13} ================= Parker Amy A. Uzicanin Amra Infectious Agent {#cesec176} ---------------- •Measles virus is a member of the genus *Morbillivirus* of the family Paramyxoviridae.•Humans are the only known natural host for the measles virus.•Measles, also known as rubeola, is one of the most highly communicable infectious diseases. Mode of Transmission {#cesec177} -------------------- •Measles spreads by airborne droplets.•Direct contact with nasal or throat secretions of infected persons.•Less commonly it is spread by articles freshly soiled with nose and throat secretions.•Infected persons are usually contagious from 4 days before onset of signs or symptoms, and until 4 days after the onset of signs or symptoms. Occurrence {#cesec178} ---------- •An estimated 20 million measles cases still occur globally every year, and travelers could be exposed in almost any country they visit. However, the risks are greater in countries where measles remains endemic or where large outbreaks are occurring.•In the Americas, indigenous measles circulation was interrupted in 2002, but risk of measles due to virus importations from other parts of the world still remains.•The number of reported measles cases in the United States has declined from 894,134 in 1941 to fewer than 150 cases each year since 1997. However, from January 1 through April 25, 2008, a total of 64 confirmed measles cases were reported to CDC, which is the largest number of cases reported in the United States for the corresponding period for any year since 2001. Ten of these cases were acquired abroad by unvaccinated travelers (five in visitors to the United States and five in U.S. residents) and the remaining cases were considered to be associated with these importations of measles. Risk for Travelers {#cesec179} ------------------ •All persons who do not have evidence of measles immunity are at risk for contracting measles during international travel.•Acceptable presumptive evidence of immunity to measles for international travelers includes---○For infants 6--11 months of age, documented administration of one dose of live measles-containing vaccine[1](#fn4){ref-type="fn"} and, for persons Δ12 months of age, two doses of MMR[2](#fn5){ref-type="fn"} vaccine at least 28 days apart, on or after the first birthday○Laboratory evidence of immunity○Birth before 1957○Documented physician-diagnosed measles Clinical Presentation {#cesec180} --------------------- •Incubation period is ∼10 days (range 7--18 days) from exposure to onset of fever, usually 14 days before appearance of rash.•Symptoms include prodromal fever, conjunctivitis, coryza, cough, and small spots with white or bluish white centers on an erythematous base on the buccal mucosa (Koplik spots).•Characteristic red, blotchy (maculopapular) rash appears on third to seventh day that begins on the face, becomes generalized, and lasts 4--7 days.•Complications include diarrhea (8%), middle ear infection (7%--9%), and pneumonia (1%--6%). Encephalitis, frequently resulting in permanent brain damage, occurs in approximately 1 per 1,000--2,000 cases of measles. Subacute sclerosing panencephalitis (SSPE), a rare but serious degenerative central nervous system disease, is thought to occur in 1 per 100,000 cases, although a risk of 22 cases of SSPE per 100,000 measles cases was found during the 1989--1991 measles resurgence in the United States. SSPE, which is caused by a persistent infection with a defective measles virus, is manifested by mental and motor deterioration that starts an average of 7 years after measles virus infection (most frequently in children \<2 years of age), progressing to coma and death.•The risk of serious complications and death is highest for children £5 years of age and adults Δ20 years of age. It is also higher in populations with poor nutritional status. Diagnosis {#cesec181} --------- •A clinical case of measles illness is characterized by all of the following:○Generalized maculopapular rash lasting Δ3 days○Temperature Δ101° F (Δ38.3° C)○Cough, coryza, or conjunctivitis•Laboratory criteria for diagnosis is a positive serologic test for measles immunoglobulin M (IgM) antibody, seroconversion or significant rise in measles IgG antibody level by any standard serologic assay, or isolation of measles virus or identification by PCR of measles virus RNA from a clinical specimen.•A confirmed case is either laboratory confirmed or meets the clinical case definition and is epidemiologically linked to a confirmed case. A laboratory-confirmed case does not need to meet the clinical case definition. Treatment {#cesec182} --------- •There is no specific antiviral therapy or treatment for measles.•Supportive therapy includes hydration, antipyretics, and treating complications such as pneumonia.•The WHO currently recommends vitamin A for all children with acute measles, regardless of their country of residence, to reduce morbidity and mortality. Vitamin A is administered once a day for 2 days, at the following doses:○50,000 IU for infants \<6 months of age○100,000 IU for infants 6--11months of age○200,000 IU for children ages 12 months or older•A third age-specific dose of vitamin A is to be given 2--4 weeks later to case-patients with clinical signs and symptoms of vitamin A deficiency. Parenteral and oral formulations of vitamin A are available in the United States. Preventive Measures for Travelers {#cesec183} --------------------------------- ### Vaccine {#cesec184} •Measles vaccine contains live, attenuated measles virus. It is available as a monovalent formulation and in combination formulations, such as measles--rubella (MR), measles--mumps--rubella (MMR), and measles--mumps--rubella--varicella (MMRV).•Ensure that all travelers who do not have evidence of measles immunity (see [Risk for Travelers](#cesec179){ref-type="sec"} earlier in this section) are up to date on measles vaccination prior to departure.○Infants 6--11 months of age should have at least one dose of measles-containing vaccine.[1](#fn4){ref-type="fn"}○Preschool children Δ12 months of age should have two doses of MMR[2](#fn5){ref-type="fn"} vaccine separated by at least 28 days.○School-age children should have two doses of MMR.[2](#fn5){ref-type="fn"}○Adults born in or after 1957 should have two doses of measles-containing vaccine.○If administered at Δ12 months of age, one dose of measles-containing vaccine or MMR is 95% effective in preventing measles disease and two doses are 99% effective. One dose of measles-containing vaccine or MMR is approximately 85% effective if administered at 9 months of age.•For persons Δ12 months of age, combined MMR vaccine is recommended whenever one or more of the individual components is indicated to provide optimal protection against mumps and rubella. For infants \<12 months of age, measles vaccine alone is recommended if it is available; otherwise MMR should be used.•MMR vaccine, if administered within 72 hours of initial measles exposure, may provide some protection. If the exposure does not result in infection, the vaccine should induce protection against subsequent measles virus infection.•Immune globulin (IG) can be used to prevent or mitigate measles in a susceptible person when administered within 6 days of exposure. However, any immunity conferred is temporary unless modified or typical measles occurs, and the person should receive measles-containing vaccine 5--6 months after IG administration. ### Adverse Reactions, Precautions, and Contraindications to Measles Vaccine {#cesec185} #### Allergy {#cesec186} Persons with severe allergy (i.e., hives, swelling of the mouth or throat, difficulty breathing, hypotension, and shock) to gelatin or neomycin or who have had a severe allergic reaction to a prior dose of MMR or MMRV should not be revaccinated except with extreme caution. MMR or MMRV vaccines may be administered to egg-allergic persons without prior routine skin testing or the use of special protocols. #### Immunosuppression {#cesec187} Replication of vaccine viruses can be potentiated in persons who have immune deficiency disorders. Death related to vaccine-associated measles infection has been reported among severely immunocompromised persons. Therefore, severely immunosuppressed individuals should not be vaccinated with MMR or MMRV vaccines•MMR or MMRV should be avoided for at least 1 month after cessation of high-dose corticosteroid therapy. Some experts, however, recommend waiting only 2 weeks after completion of therapy among individuals receiving high doses of systemic corticosteroids daily or on alternate days even if they were receiving therapy for less than 14 days.•Other immunosuppressive therapy: MMR or MMRV vaccines in general should be withheld for at least 3 months. This interval is based on the assumption that the immunologic responsiveness will have been restored in 3 months and the underlying disease for which the therapy was given is in remission. #### Thrombocytopenia {#cesec188} The benefits of primary immunization are usually greater than the potential risks. However, avoiding a subsequent dose of MMR or MMRV vaccine may be prudent if an episode of thrombocytopenia occurred within approximately 6 weeks after a previous dose of vaccine. MUMPS {#subchapter14} ===== Kutty Preeta K. Barskey Albert E. IV Gallagher Kathleen M. Infectious Agent {#cesec190} ---------------- •Mumps virus is an enveloped, negative-strand RNA virus, a member of the genus *Rubulavirus*.•Humans are the only known natural host for mumps virus. Mode of Transmission {#cesec191} -------------------- •Transmission is by respiratory droplets, saliva, or contact with contaminated fomites.•Patients are usually contagious 1--2 days (occasionally as long as 7 days) before symptom onset until 5 days afterward. Occurrence {#cesec192} ---------- •With the exception of the multistate outbreak in 2006, mumps is an uncommon disease in the United States because of a successful vaccination program.•Mumps virus remains endemic in many countries throughout the world because mumps vaccine is used in only 57% of the World Health Organization member countries. Risk for Travelers {#cesec193} ------------------ •The risk of exposure to mumps among travelers can be high in most countries of the world, especially for travelers \>12 months of age who do not have evidence of mumps immunity (see [Preventive Measures for Travelers](#cesec197){ref-type="sec"} later in this section). Although some countries have had variable successes with a national vaccination program---including Finland, which has declared elimination---the risk of contacting imported mumps in these countries is still a concern.•Acceptable presumptive evidence of immunity to mumps for international travelers includes---○Documented administration of two doses of live mumps virus vaccine at least 28 days apart, on or after the first birthday○Laboratory evidence of immunity○Birth before 1957○Documentation of physician-diagnosed mumps Clinical Presentation {#cesec194} --------------------- •Incubation period from exposure to onset of symptoms is generally 16--18 days (range 12--25 days).•Onset of illness is usually nonspecific, with symptoms of fever, headache, malaise, myalgia, and anorexia.•Mumps is characterized by parotitis, either unilateral or bilateral.•Although mumps is generally a mild and self-limited disease, complications of mumps infection can include deafness; orchitis, oophoritis, or mastitis (inflammation of the testicles, ovaries or breasts, respectively); pancreatitis; and meningitis or encephalitis. With the exception of deafness, these complications are more frequent in adults than in children. Diagnosis {#cesec195} --------- •Mumps may occur in epidemics; mumps virus is the only cause of epidemic parotitis.•Diagnosis is usually clinical, based on the presence of parotitis and associated signs, symptoms, or complications.•Clinical case definition: An illness with acute onset of unilateral or bilateral tender, self-limited swelling of the parotid glands, other salivary gland(s), or both, lasting at least 2 days, and without other apparent cause.•Laboratory criteria include---○Isolation of mumps virus from clinical specimen○Detection of mumps nucleic acid (e.g., standard or real-time RT-PCR assays)○Detection of mumps IgM antibody○Demonstration of specific mumps antibody response in the absence of recent vaccination, either a fourfold increase in IgG titer as measured by quantitative assays, or seroconversion from negative to positive by using a standard serologic assay of paired acute- and convalescent-phase serum specimens•Laboratory specimens that can be collected are serum for serology (IgM, IgG) and a buccal swab (or a throat swab) for viral specimens. For more information see [www.cdc.gov/vaccines/vpd-vac/mumps/outbreak/faqs-lab-spec-collect.htm](http://www.cdc.gov/vaccines/vpd-vac/mumps/outbreak/faqs-lab-spec-collect.htm){#interref49}.•Laboratory confirmation is more challenging in highly vaccinated populations. Serologic tests should be interpreted with caution. A negative laboratory test should not rule out a clinically compatible case, especially in a two-dose vaccine recipient. Treatment {#cesec196} --------- There is no specific antiviral therapy for mumps, and the basic treatment consists of supportive care. Preventive Measures for Travelers {#cesec197} --------------------------------- ### Vaccine {#cesec198} •Although vaccination against mumps is not a requirement for entry into any country (including the United States), travelers leaving the United States or living abroad should ensure they are immune to mumps.•Mumps vaccine contains live, attenuated mumps virus. It is available as a monovalent formulation and in combination formulations, such as MMR. Combined MMR vaccine is recommended whenever one or more of the individual components is indicated to provide optimal protection against measles and rubella. Mumps vaccine is highly, but not 100%, effective in preventing mumps. One dose of mumps vaccine is approximately 80%--85% effective in preventing clinical mumps with parotitis, and two doses are approximately 90% effective.•Mumps vaccine has not been demonstrated to be effective in preventing infection after exposure; however, it can be administered postexposure to provide protection against subsequent exposures. Immune globulin is not effective in preventing mumps infection following an exposure and is not recommended. ### Adverse Reactions, Precautions, and Contraindications to Mumps Vaccine {#cesec199} •Refer to the Measles (Rubeola) section earlier in this chapter for information on reactions following MMR vaccine and additional precautions and contraindications. #### General Vaccine Recommendations, Pediatric and Catch-Up Schedules, and Recommendations for Special Populations {#cesec200} •Refer to Chapters 7 and 8. PERTUSSIS {#subchapter15} ========= Skoff Tami H. Thomas Cynthia G. Infectious Agent {#cesec201} ---------------- Pertussis is caused by fastidious gram-negative coccobacillus, *Bordetella pertussis.* Mode of Transmission {#cesec202} -------------------- It is spread by person-to-person transmission via aerosolized respiratory droplets or by direct contact with respiratory secretions. Occurrence {#cesec203} ---------- •*B. pertussis* circulates worldwide, but disease rates are highest among young children in countries where vaccination coverage is low, which is primarily in the developing world.•In developed countries, the incidence of pertussis is highest among unvaccinated infants and increases again among adolescents.•Immunity from childhood vaccination and natural disease wanes with time; therefore, adolescents and adults who have not received a Tdap booster vaccination can become infected or re-infected. Risk for Travelers {#cesec204} ------------------ •Pertussis remains endemic worldwide, even in areas with high vaccination rates.•Travelers who come in close contact with infected persons are at risk for disease. Infants too young to be protected by a complete vaccination series are at greatest risk for severe pertussis requiring hospitalization. Clinical Presentation {#cesec205} --------------------- •In classic disease, mild upper respiratory tract symptoms begin 7--10 days (range 6--21 days) after exposure, followed by a cough that becomes paroxysmal. Coughing paroxysms may be frequent or relatively infrequent and are often followed by vomiting. Fever is absent or minimal. The CDC/Council of State and Territorial Epidemiologists\' clinical case definition for pertussis includes cough for Δ2 weeks with paroxysms, whoop, and/or post-tussive vomiting.•Disease in infants \<6 months of age can be atypical with a short catarrhal stage, gagging, gasping, or apnea as early manifestations; among infants \<2 months of age, the case--fatality rate is approximately 1%.•Recently immunized children may have mild cough illness; older children and adults may have prolonged cough with or without paroxysms. The cough gradually wanes over several weeks to months. Diagnosis {#cesec206} --------- •Factors such as prior vaccination status, stage of disease, antibiotic use, specimen collection and transport conditions, and nonstandardized tests may affect the sensitivity, specificity, and interpretation of available diagnostic tests for *B. pertussis*.•Current CDC guidelines for the laboratory confirmation of pertussis cases include culture and PCR (when the above clinical case definition is met); serology and direct fluorescent antibody (DFA) tests are not confirmatory tests included in the case definition. Treatment {#cesec207} --------- •Macrolide antibiotics (azithromycin, clarithromycin, and erythromycin) are recommended for the treatment of pertussis in persons Δ1 month of age; for infants \<1 month of age, azithromycin is the preferred antibiotic.•Antimicrobial therapy with a macrolide antibiotic administered \<3 weeks after cough onset can limit transmission to others.•Postexposure prophylaxis is recommended for close contacts of cases and for individuals at high risk of developing severe disease. The recommended agents and dosing regimens for prophylaxis are the same as those indicated for the treatment of pertussis. Preventive Measures for Travelers {#cesec208} --------------------------------- ### Vaccine {#cesec209} •Travelers should be up to date with pertussis vaccinations prior to departure.•Complete vaccination of children \<7 years of age with five doses of acellular pertussis vaccine in combination with diphtheria and tetanus toxoids (DTaP) is recommended; an accelerated schedule of doses may be used to complete the DTaP series.•There is no pertussis-containing vaccine licensed for children 7--9 years of age. If a child turns 10 years old during the vaccination series with Td (tetanus and diphtheria toxoids vaccine), a single dose of Tdap may be substituted for one of the Td doses.•Adolescents aged 11--18 years should receive a single dose of Tdap instead of Td for booster immunization against tetanus, diphtheria, and pertussis if they have completed the recommended childhood DTwP/DTaP vaccination series. Adolescents who received their last Td (tetanus and diphtheria toxoids vaccine) 5 years or more previously should also receive a single dose of Tdap.•Adults 19--64 years of age should receive a single dose of Tdap to replace a single dose of Td for booster immunization against tetanus, diphtheria, and pertussis if their last tetanus toxoid-containing vaccine (e.g., Td) was administered 10 years or more prior. Tdap is not licensed for adults 65 years of age or older.•Tdap can be given in intervals \<10 years from the last Td to provide pertussis protection prior to travel, except in those individuals with a contraindication to vaccination.•Adolescents and adults who have never been immunized against pertussis, tetanus, or diphtheria, have incomplete immunization, or whose immunity is uncertain should follow the catch-up schedule established for Td/Tdap. Tdap can be substituted for any one of the Td doses in the series. PNEUMOCOCCAL DISEASE (*STREPTOCOCCUS PNEUMONIAE*) {#subchapter16} ================================================= Nuorti J. Pekka Infectious Agent {#cesec210} ---------------- •*Streptococcus pneumoniae* (pneumococcus) is a bacterium that frequently colonizes the nasopharynx of healthy persons, particularly young children, without causing illness.•There are 91 known pneumococcal serotypes.•The major clinical syndromes include life-threatening infections such as meningitis, bacteremia, and pneumonia.•Pneumococcus is the most commonly identified cause of community-acquired pneumonia. It is also a major cause of milder but more common illnesses, such as sinusitis and otitis media. Mode of Transmission {#cesec211} -------------------- •Direct person-to-person transmission is through close contact via respiratory droplets.•Transmission is thought to be common, but clinical illness occurs infrequently among casual contacts. Occurrence {#cesec212} ---------- •Pneumococcal disease occurs worldwide, and the reported incidence varies by geographic region.•Rates are higher in developing countries than in industrialized countries.•Pneumococcal disease is more common during winter and early spring, when respiratory viruses such as influenza are circulating. Most illnesses are sporadic.•Outbreaks of pneumococcal disease are uncommon but may occur in closed populations such as nursing homes, childcare centers or other institutions.•In the United States, most deaths from pneumococcal disease occur in older adults, although in developing countries, many children die of pneumococcal pneumonia.•Routine use of the 7-valent pneumococcal conjugate vaccine (PCV7) in the United States since 2000 has dramatically reduced the incidence of pneumococcal disease in both children and adults. Because the vaccine interrupts transmission of vaccine-type pneumococci, rates of pneumococcal disease in unvaccinated older children and adults have also decreased.•As of 2008, 18 industrialized countries are routinely using pneumococcal conjugate vaccines, including Canada, Australia, the United Kingdom, and other Western European and Middle Eastern countries. Risk for Travelers {#cesec213} ------------------ •The risk for pneumococcal disease is generally highest among young children, the elderly, and persons of any age who have chronic medical conditions, such as heart disease, lung disease, diabetes or asplenia, or conditions that suppress the immune system, such as HIV.•Cigarette smokers are also at increased risk.•Most travelers, however, are not in these categories. It is important to recognize that healthy travelers in their twenties or thirties have developed pneumococcal pneumonia while traveling in developing countries. Clinical Presentation {#cesec214} --------------------- •Sudden onset with fever and malaise are typical symptoms for all forms of pneumococcal infections and may be the only symptoms in young children with bacteremia.•In pneumococcal pneumonia, fever may precede the usual symptoms of cough, pleuritic chest pain, and the production of purulent or blood-tinged sputum.•In elderly persons, the onset of pneumococcal pneumonia may be less abrupt, with fever, shortness of breath, or altered mental status as the initial symptoms; sputum production may be absent.•Pneumococcal meningitis may present with a stiff neck, headache, lethargy, or seizures; otitis media or sinusitis typically cause pain in the ears or sinuses. Diagnosis {#cesec215} --------- •A definitive diagnosis of pneumococcal infection can be made by isolation of the bacterium from blood or other normally sterile body sites, such as cerebrospinal fluid. Most patients with pneumococcal pneumonia, however, do not have detectable bacteremia.•The diagnosis of pneumococcal pneumonia can be suspected if on microscopy a sputum specimen contains many gram-stain positive diplococci and polymorphonuclear leukocytes and very few epithelial cells.•Typical chest radiography may show lobar, segmental, or multilobar consolidation.•Pneumococcal pneumonia is usually, but not always, associated with a high white blood cell count. High white blood cell counts should raise suspicion for this diagnosis, since other serious travel-related diseases causing fever, such as hepatitis, typhoid fever, malaria, dengue fever, or typhus, all have normal or low white blood cell counts. Treatment {#cesec216} --------- •All types of pneumococcal infections are usually treated with antibiotics.•Worldwide, many strains are increasingly resistant to penicillin, cephalosporin, and macrolides, and some are resistant to multiple classes of drugs, complicating treatment choices. Antimicrobial susceptibility of strains isolated from blood and cerebrospinal fluid should be determined, and treatment should be targeted based on the susceptibility results.•In 2008, the Clinical and Laboratory Standards Institute adopted new susceptibility breakpoints for penicillin treatment of nonmeningitis cases of pneumococcal disease. However, empiric antibiotic therapy should not be delayed and should begin before microbiological confirmation of etiology.•In the United States and other countries where beta-lactam resistance among pneumococcal isolates is common, the initial regimen for suspected pneumococcal meningitis should include vancomycin until the antimicrobial susceptibility pattern of the organism is available. Preventive Measures for Travelers {#cesec217} --------------------------------- ### Vaccine {#cesec218} •No specific recommendations for the use of pneumococcal vaccines in travelers have been formulated.•Currently, two vaccines are available for prevention of pneumococcal disease in the United States.○***Pneumococcal conjugate vaccine***---The 7-valent pneumococcal conjugate vaccine (PCV7) (Prevnar, Wyeth Vaccines) is mainly used in children. It is part of the routine infant immunization schedule in the United States and is now recommended for all children \<5 years of age (see the Vaccine Recommendations for Infants and Children section in Chapter 7). The infant schedule consists of a three-dose primary series at ages 2, 4, and 6 months and a booster dose at 12--15 months of age. Fewer doses are required for children who begin the series after 7 months of age.○***Pneumococcal polysaccharaide vaccine***---A 23-valent pneumococcal polysaccharide vaccine (PPV23) (Pneumovax, Merck) is mainly used in older adults and persons with underlying medical conditions. PPV23 is recommended for all adults Δ65 years of age and for persons 2--64 years of age with underlying medical conditions at the time the condition is recognized. In 2006, only about 57% of adults aged Δ65 years of age had received the vaccine. Children 2--4 years of age who have underlying medical conditions that are indications for PPV23 should also receive polysaccharide vaccine after receiving the conjugate vaccine series.•Both vaccines induce antibodies to the specific types of pneumococcal capsule and have been shown to be effective against invasive disease.•Additional pneumococcal conjugate vaccine formulations are expected to be licensed soon. The WHO recommends that inclusion of pneumococcal conjugate vaccines in all national immunization programs should be a priority.•Routine revaccination is not recommended for most people. A second dose of PPV23 is recommended 5 years after the first dose for the following groups:○Persons with sickle cell disease, asplenia, renal disease, hematologic or generalized malignancy, or other immunocompromising condition○Persons Δ65 years of age who received PPV23 before age 65 years for an underlying medical condition, if at least 5 years have passed since their previous dose•Because of limited data regarding the duration of protection provided by PPV23 and the safety of multiple doses, only a single revaccination is recommended. Persons should receive one dose if they have an indication for polysaccharide vaccine and their vaccination history is unknown. #### Safety and Side Effects {#cesec219} •After receipt of PCV7, mild local reactions, such as redness, swelling, or tenderness, occur in 10%--23% of infants. Larger areas of redness or swelling or limitations in arm movement may occur in 1%--9% of infants. Low-grade fever can occur in up to 24% and fever higher than 102.2° F may occur in up to 2.5% of vaccinees.•After receipt of PPV23, self-limiting local side effects occur in approximately half of vaccine recipients and are more common after revaccination than with first dose. These reactions usually resolve within 48 hours. More severe local reactions and systemic symptoms, including myalgias and fever, are rare. #### Precautions and Contraindications {#cesec220} •PCV7 is contraindicated for children known to have hypersensitivity to any component of the vaccine.•Health-care providers may delay vaccination of children with moderate or severe illness until the child has recovered, although minor illnesses, such as mild upper-respiratory tract infection with or without low-grade fever, are not contraindications.•Revaccination with PPV23 is contraindicated for persons who had a severe reaction (e.g., anaphylactic reaction or localized arthus-type reaction) to the initial dose. ### Additional Preventive Measures {#cesec221} •The following may reduce the risk of pneumococcal disease:○improving control of chronic conditions that predispose to pneumococcal disease, such as diabetes and HIV,○stopping smoking, and○avoiding crowded living conditions.•Chemoprophylaxis is not routinely recommended for close contacts of pneumococcal meningitis or other cases of invasive disease or for travelers unless otherwise recommended by the health-care practitioner supervising their care. POLIOMYELITIS {#subchapter17} ============= Alexander James P. Wassilak Steven Infectious Agent {#cesec222} ---------------- •The infectious agent is poliovirus (genus Enterovirus) types 1, 2, and 3.•Polioviruses are small (27--30 nm), nonenveloped viruses with capsids enclosing a single-stranded, positive-sense RNA genome about 7,500 nucleotides long.•Most of the properties of polioviruses are shared with the other enteroviruses. Mode of Transmission {#cesec223} -------------------- Fecal--oral or oral transmission. Acute infection involves the gastrointestinal tract and occasionally the central nervous system. Occurrence {#cesec224} ---------- •In the prevaccine era, infection with poliovirus was common worldwide, with seasonal peaks and epidemics in the summer and fall in temperate areas.•The incidence of poliomyelitis in the United States declined rapidly after the licensure of inactivated polio vaccine (IPV) in 1955 and live oral polio vaccine (OPV) in the 1960s. The last cases of indigenously acquired polio in the United States occurred in 1979.•The Global Polio Eradication Initiative (GPEI) subsequently led to elimination of polio in the Americas, where the last wild poliovirus (WPV)-associated polio case was detected in 1991.•In 1999, a change in vaccination policy in the United States from use of OPV to exclusive use of IPV resulted in the elimination of the 8--10 vaccine-associated paralytic poliomyelitis (VAPP) cases that had occurred annually since the introduction of OPV in the 1960s.•In the United States, two events that occurred in 2005 highlighted the continuing but low risk for poliovirus infection for unvaccinated persons, whether residing in the United States or traveling.○A case of imported VAPP occurred in an unvaccinated U.S. adult who had traveled abroad, likely from contact with an infant recently vaccinated with OPV.○An unvaccinated immunocompromised infant and four children in two other families in the same small rural community were found to be asymptomatically infected with a vaccine-derived poliovirus, presumably originating outside the United States in a country that uses OPV.•The GPEI has built upon the success in the Americas and made great progress in eradicating wild polioviruses. There are only four countries where wild poliovirus circulation has never been interrupted: Afghanistan, India, Nigeria, and Pakistan. WPV type 2 has not been detected since October 1999.•During 2002--2006, 22 previously polio-free countries were affected by importations of WPV type 1 from the remaining polio-endemic countries, primarily Nigeria. In 2007--2008, polio cases occurred in 12 countries following importations of WPV originating from Nigeria or India.•In spite of recent WPV outbreaks and continued circulation in the four countries where WPV circulation has never been interrupted, the GPEI has reduced the number of reported polio cases worldwide by more than 99% since the mid-1980s. With intensified efforts, worldwide eradication of polio appears feasible in the future. Risk for Travelers {#cesec225} ------------------ •Because of polio eradication efforts, the number of countries where travelers are at risk for polio has decreased dramatically.•At the time of publication, most of the world\'s population resides in areas considered free of WPV circulation, including the Western Hemisphere, the Western Pacific region (which encompasses China), and the European region.•Vaccination is recommended for all travelers to polio-endemic or epidemic areas, including countries with recent proven WPV circulation and neighboring countries. As of September 2008, these areas include some but not all countries in Africa, South Asia, Southeast Asia, and the Middle East. For current information on the status of polio eradication efforts and vaccine recommendations, consult the Travel Notices on the CDC Travelers\' Health website ([www.cdc.gov/travel/](http://www.cdc.gov/travel/){#interref51}) or the GPEI website ([www.polioeradication.org/](http://www.polioeradication.org/){#interref52}). Clinical Presentation {#cesec226} --------------------- Clinical manifestations of poliovirus infection range from asymptomatic (most infections) to symptomatic, including acute flaccid paralysis of a single limb to quadriplegia, respiratory failure, and, rarely, death. Diagnosis {#cesec227} --------- The diagnosis is made by the identification of poliovirus in clinical specimens (usually stool) obtained from an acutely ill patient. Poliovirus may be detected from stool specimens for up to 4 weeks after onset of illness. Treatment {#cesec228} --------- Only symptomatic treatment is available, ranging from pain and fever relief to intubation and mechanical ventilation for those with respiratory insufficiency. Preventive Measures for Travelers {#cesec229} --------------------------------- •A person is considered to be fully immunized if he or she has received a primary series of at least three doses of IPV, three doses of OPV, or four doses of any combination of IPV and OPV.•To eliminate the risk for VAPP, OPV has not been recommended for routine immunization in the United States since January 1, 2000, and is no longer available in this country.•OPV continues to be used in the majority of countries and for global polio eradication activities. ### Vaccine {#cesec230} #### Infants and Children {#cesec231} •Because OPV is no longer recommended for routine immunization in the United States, all infants and children should receive four doses of IPV at 2, 4, and 6--18 months and 4--6 years of age. The fourth (booster) dose is not needed if the third dose of the primary series is administered on or after the fourth birthday.•If accelerated protection is needed, the minimum interval between doses is 4 weeks, although the preferred interval between the second and third doses is 2 months.•The minimum age for IPV administration is 6 weeks. Infants and children who have initiated the poliovirus vaccination series with one or more doses of OPV should receive IPV to complete the series. #### Adults {#cesec232} •Adults who are traveling to areas where poliomyelitis cases are still occurring and who are unvaccinated, incompletely vaccinated, or whose vaccination status is unknown should receive two doses of IPV administered at an interval of 4--8 weeks; a third dose should be administered 6--12 months after the second.•If three doses of IPV cannot be administered within the recommended intervals before protection is needed, the following alternatives are recommended:○If \>8 weeks is available before protection is needed, three doses of IPV should be administered at least 4 weeks apart.○If \<8 weeks but \>4 weeks is available before protection is needed, two doses of IPV should be administered at least 4 weeks apart.○If \<4 weeks is available before protection is needed, a single dose of IPV is recommended.•If fewer than three doses are administered, the remaining IPV doses to complete a three-dose series should be administered when feasible, at the intervals recommended above, if the person remains at increased risk for poliovirus exposure.•Adults (Δ18 years of age) who are traveling to areas where poliomyelitis cases are occurring and who have received a primary series with either IPV or OPV in childhood should receive another dose of IPV before departure.•For adults, available data do not indicate the need for more than a single lifetime booster dose with IPV. #### Allergy to Vaccine {#cesec233} •Minor local reactions (pain and redness) can occur following IPV. No serious adverse reactions to IPV have been documented.•IPV should not be administered to persons who have experienced a severe allergic (anaphylactic) reaction after a previous dose of IPV or after receiving streptomycin, polymyxin B, or neomycin which IPV contains in trace amounts; hypersensitivity reactions can occur following IPV among persons sensitive to these three antibiotics. #### Pregnancy and Breastfeeding {#cesec234} •If a pregnant woman is unvaccinated or incompletely vaccinated and requires immediate protection against polio because of planned travel to a country or area where polio cases are occurring, IPV can be administered as recommended for adults.•Breastfeeding is not a contraindication to immunization of an infant or mother against polio. #### Precautions and Contraindications {#cesec235} •IPV may be administered to persons with diarrhea.•Minor upper respiratory illnesses with or without fever, mild to moderate local reactions to a previous dose of IPV, current antimicrobial therapy, and the convalescent phase of acute illness are not contraindications for vaccination. #### Immunosuppression {#cesec236} •IPV may be administered safely to immunodeficient travelers and their household contacts. Although a protective immune response cannot be ensured, IPV might confer some protection to the immunodeficient person.•Persons with certain primary immunodeficiency diseases should avoid contact with excreted OPV virus (e.g., exposure to a child vaccinated with OPV within the previous 6 weeks); however, this situation no longer occurs in the United States unless a child receives OPV overseas. RUBELLA {#subchapter18} ======= Reef Susan E. Infectious Agent {#cesec237} ---------------- Rubella virus is a member of *Togaviridae* family and the only member of the genus *Rubivirus.* Mode of Transmission {#cesec238} -------------------- •Rubella virus is transmitted through person-to-person contact or droplets shed from the respiratory secretions of infected persons.•If a woman with rubella is infected during pregnancy, the virus can cross the placenta and infect the fetus. Occurrence {#cesec239} ---------- •Rubella occurs worldwide.•In the United States, endemic rubella has been eliminated. However, since 2005, an average of 10 cases is reported each year. Of these cases, approximately 33% are imported or linked to importations. Risk for Travelers {#cesec240} ------------------ •All susceptible persons are at risk for infection from exposure to rubella during travel outside the United States.•Because asymptomatic rubella infections are common, travelers may be unaware that they have been in contact with an infected person. Clinical Presentation {#cesec241} --------------------- •The average incubation period is 14 days, with a range of 12--23 days.•Rubella usually presents as a nonspecific, maculopapular, generalized rash lasting 3 days or fewer (hence the term "3-day measles") with generalized lymphadenopathy, particularly of the posterior auricular, suboccipital and posterior cervical lymph nodes.•Asymptomatic rubella virus infections are common, and up to 50% of infections occur without rash.•In adults and adolescents, the rash may be preceded by a 1- to 5-day prodrome of low-grade fever, malaise, anorexia, mild conjunctivitis, coryza, sore throat, and lymphadenopathy.•The most important and serious consequence of rubella is infection during early pregnancy. These consequences may include miscarriages, fetal deaths/stillbirths, and an infant born with constellation of severe birth defects known as congenital rubella syndrome (CRS). The most common congenital defects are cataracts, heart defects, and hearing impairment. Diagnosis {#cesec242} --------- •Many illnesses can mimic rubella, and up to 50% of rubella infections are asymptomatic. Therefore, the only reliable evidence of acute rubella virus infection is laboratory diagnosis.•Serologic testing for rubella-specific IgM antibody is the most commonly used for diagnosis of rubella.•Diagnosis can also be made by demonstration of seroconversion of rubella-specific IgG antibody titers and by detection of virus either through virus culture or PCR. Treatment {#cesec243} --------- There is no specific antiviral therapy for rubella; basic treatment consists of supportive care. Preventive Measures for Travelers {#cesec244} --------------------------------- ### Vaccine {#cesec245} •Before international travel, persons should be immune to rubella.•Acceptable presumptive evidence of immunity to rubella for international travelers includes---○Documentation of receipt of one or more doses of rubella-containing vaccine on or after the first birthday○Laboratory evidence of rubella immunity (a positive serologic test for rubella-specific IgG antibody) #### Adverse Reactions, Precautions, and Contraindications to Rubella Vaccine {#cesec246} •Refer to the Measles (Rubeola) section earlier in this chapter for information on reactions following MMR vaccine and additional precautions and contraindications. TETANUS {#subchapter19} ======= Joyce M. Patricia Infectious Agent {#cesec247} ---------------- •*Clostridium tetani*, the tetanus bacillus, is a spore-forming, anaerobic gram-positive bacterium.•Clinical disease is caused by a neurotoxin produced by anaerobic tetanus bacilli growing in contaminated wounds. Mode of Transmission {#cesec248} -------------------- •Tetanus is a global health problem because *C. tetani* spores are ubiquitous in the environment.•Lesions that are considered "tetanus prone" are wounds contaminated with dirt, feces, or saliva, deep wounds, burns, crush injuries, or those with necrotic tissue.•Tetanus has also been associated with apparently clean superficial wounds, surgical procedures, insect bites, dental infections, chronic sores and infections, and intravenous drug use.•A reservoir of tetanus bacteria exists in the intestines of horses and other animals, including humans, in which the organism is a harmless normal inhabitant. Soil or fomites contaminated with animal and human feces propagate transmission.•Tetanus has no direct person-to-person transmission. Occurrence {#cesec249} ---------- •In 2006, an estimated 290,000 people worldwide died of tetanus, most of them in Asia, Africa, and South America.•The disease occurs almost exclusively in persons who are inadequately immunized.•Worldwide, the disease is more common in agricultural regions and in areas where contact with animal excreta is more likely and immunization is inadequate.•In developing countries, tetanus in neonates born to unvaccinated mothers (neonatal tetanus) is the most common form of the disease.•In 10% of reported cases in the United States, no antecedent wound was identified. Risk for Travelers {#cesec250} ------------------ Tetanus can occur anywhere in the world in inadequately vaccinated persons. Clinical Presentation {#cesec251} --------------------- •Acute manifestations of tetanus are characterized by muscle rigidity and painful spasms, often starting in the muscles of the jaw and neck. Severe tetanus can lead to respiratory failure and death.•The incubation period is usually 3--21 days (average 10 days), although it may range from 1 day to several months, depending on the character, extent, and location of the wound. Most cases occur within 14 days. In general, shorter incubation periods are associated with more heavily contaminated wounds, more severe disease, and a worse prognosis. ### Clinical Syndromes {#cesec252} #### Generalized Tetanus {#cesec253} •Generalized tetanus is the most common form, accounting for more than 80% of cases.•Neonatal tetanus is generalized tetanus in neonates, usually due to umbilical stump infections.•The average incubation period from injury to symptom onset is 7--8 days (range 3 days--3 weeks).•The most common initial sign is trismus (spasm of the muscles of mastication or "lockjaw"). Trismus may be followed by painful spasms in other muscle groups in the neck, trunk, and extremities and by generalized, tonic, seizure-like activity or frank convulsions in severe cases.•Generalized tetanus can be accompanied by autonomic nervous system abnormalities, as well as a variety of complications related to severe spasm and prolonged hospitalization.•The clinical course of generalized tetanus is variable and depends on the degree of prior immunity, the amount of toxin present, and the age and general health of the patient.•Even with modern intensive care, generalized tetanus is associated with mortality rates of 10%--20%. #### Localized Tetanus {#cesec254} •Localized tetanus is an unusual form of the disease consisting of spasm of muscles in a confined area close to the site of the injury.•Although localized tetanus often occurs in persons with partial immunity and is usually mild, progression to generalized tetanus can occur. #### Cephalic Tetanus {#cesec255} •The rarest form, cephalic tetanus, is associated with lesions of the head or face and has been described in association with ear infections (i.e., otitis media).•The incubation period is short, usually 1--2 days.•Unlike generalized and localized tetanus, cephalic tetanus results in flaccid cranial nerve palsies rather than spasm. Trismus may also be present. Like localized tetanus, cephalic tetanus can progress to the generalized form. Diagnosis {#cesec256} --------- •The diagnosis is almost always made clinically.•The disease is characterized by painful muscular contractions, primarily of the masseter and neck muscles, secondarily of trunk muscles.•A common first sign suggestive of tetanus in older children and adults is abdominal rigidity, though rigidity is sometimes confined to the region of injury.•Generalized spasms occur, frequently induced by sensory stimuli; typical features of the tetanic spasm are the position of opisthotonos and the facial expression known as "risus sardonicus."•History of an injury or apparent portal of entry may be lacking.•The organism is rarely recovered from the site of infection, and usually there is no detectable antibody response. Treatment {#cesec257} --------- •Tetanus is a medical emergency requiring hospitalization, immediate treatment with human tetanus immune globulin (TIG) (or equine antitoxin if human immune globulin is not available), a tetanus toxoid booster, agents to control muscle spasm, and, if indicated, aggressive wound care and antibiotics.○Specific treatment: TIG administered intramuscularly in doses of 3000--6000 IU. If immunoglobulin is not available, tetanus antitoxin (equine origin) in a single large dose should be given intravenously, following testing for hypersensitivity.•Metronidazole is the most appropriate antibiotic. It is associated with the shortest recovery time and lowest case--fatality rate. It should be given for 7--14 days in large doses; this also allows for a reduction in the amount of muscle relaxants and sedatives required.•The wound should be debrided widely and excised if possible. Wide debridement of the umbilical stump in neonates is not indicated.•Depending on the severity of disease, mechanical ventilation and agents to control autonomic nervous system instability may be required.•An adequate airway should be maintained, and sedation should be used as indicated; muscle relaxant drugs, together with tracheostomy or nasotracheal intubation and mechanically assisted respiration, may be lifesaving.•Active immunization should be initiated concurrently with treatment. Preventive Measures for Travelers {#cesec258} --------------------------------- •Travelers should ensure they have adequate immunity to tetanus.○Active immunity is induced by tetanus toxoid and persists for at least 10 years after full immunization; transient passive immunity follows injection of TIG or tetanus antitoxin (equine origin).○Infants of actively immunized mothers acquire passive immunity that protects them from neonatal tetanus.○Recovery from tetanus may not result in immunity; second attacks can occur, and primary immunization is indicated after recovery.•Wounded travelers who received their most recent tetanus toxoid-containing vaccine more than 5 years previously or who have not received at least three doses of tetanus toxoid-containing vaccines may require a dose of tetanus toxoid-containing vaccine (Tdap, Td, or DTaP), depending on the nature of the wound.•Human tetanus immune globulin (TIG) is indicated in travelers with tetanus-prone wounds who have an unknown or incomplete history of primary tetanus vaccination. ### General Preventive Measures {#cesec259} •Universal active immunization with adsorbed tetanus toxoid, gives durable protection for at least 10 years; after the initial basic series has been completed, single booster doses elicit high levels of immunity.•In children \<7 years of age, the toxoid is generally administered together with diphtheria toxoid and pertussis vaccine as a triple (DTP or DTaP) antigen, or as a double (DT) antigen when contraindications to pertussis vaccine exist.•Td is used for children \>7 years of age.•For adolescents 11--12 years of age, a single dose of Tdap is recommended for routine booster, and for adolescents and adults 13--64 years of age, a single dose of Tdap is recommended to replace the next decennial Td booster or when indicated as part of wound prophylaxis.•In countries with incomplete immunization programs for children, all pregnant women should receive two doses of tetanus toxoid in the first pregnancy, with an interval of at least 1 month, and with the second dose at least 2 weeks prior to childbirth.•Vaccine-induced maternal immunity is important in preventing maternal and neonatal tetanus. Active protection should be maintained by administering booster doses of Td every 10 years, preferably before or between pregnancies.•For children and adults who are severely immunocompromised or infected with HIV, tetanus toxoid is indicated in the same schedule and dose as for immunocompetent persons even though the immune response may be suboptimal.•Minor local reactions following tetanus toxoid injections are relatively frequent; severe local and systemic reactions are infrequent but do occur, particularly after excessive numbers of prior doses have been given. ### Prophylaxis in Wound Management (see [Table 2-21](#cetable21){ref-type="table"}) {#cesec260} •Tetanus prophylaxis in patients with wounds is based on careful assessment of whether the wound is clean or contaminated, the immunization status of the patient, proper use of tetanus toxoid and/or TIG, wound cleaning and, where required, surgical debridement and the proper use of antibiotics.•Those who have been completely immunized and who sustain minor and uncontaminated wounds require a booster dose of toxoid only if more than 10 years have elapsed since the last dose was given. For major or contaminated wounds, a single booster injection of tetanus toxoid (preferably as Td or Tdap) should be administered promptly on the day of injury if the patient has not received tetanus toxoid within the preceding 5 years.•Persons who have not completed a full primary series of tetanus toxoid require a dose of toxoid as soon as possible following the wound and may require passive immunization with human TIG if the wound is major or if it is contaminated with soil containing animal excreta. DTP/DTaP, DT, or Td, as determined by the age of the patient and previous immunization history, should be used at the time of the wound and ultimately to complete the primary series.•Passive immunization with at least 250 IU of TIG intramuscularly (or 1,500 to 5,000 IU of antitoxin of animal origin, if globulin is not available), regardless of the patient\'s age, is indicated for patients with other than clean, minor wounds and a history of no, unknown, or fewer than three previous tetanus toxoid doses. When tetanus toxoid and TIG or antitoxin are given concurrently, separate syringes and separate sites must be used. VARICELLA (CHICKENPOX) {#subchapter20} ====================== Harriman Kathleen H. Chavez Gilberto F. Infectious Agent {#cesec261} ---------------- •Varicella-zoster virus (VZV) is a member of the herpesvirus family.•Humans are the only reservoir of the virus, and disease occurs only in humans. Mode of Transmission {#cesec262} -------------------- •VZV is transmitted from person to person by direct contact, inhalation of aerosols from vesicular fluid of skin lesions of acute varicella or zoster, or infected respiratory tract secretions that might also be aerosolized.•The virus enters the host through the upper respiratory tract or the conjunctiva.•In utero infection can also occur as a result of transplacental passage of virus during maternal varicella infection.•The period of contagiousness is estimated to begin 1--2 days before the onset of rash and to end when all lesions are crusted, typically 4--7 days after onset of rash in immunocompetent persons, but this period may be longer in immunocompromised persons. Occurrence {#cesec263} ---------- •Varicella occurs worldwide. In temperate climates, varicella tends to be a childhood disease, with peak incidence during late winter and early spring. In tropical climates, infection tends to occur at older ages, resulting in higher susceptibility among adults than in temperate climates.•Before introduction of varicella vaccine in the United States in 1995, varicella was endemic, and virtually all persons were infected by adulthood. Since implementation of the varicella vaccination program, the epidemiology and clinical characteristics of varicella have changed, with substantial declines in morbidity and mortality. The incidence of varicella has steadily declined in all age groups, with the greatest decline among children 1--4 years of age. Risk for Travelers {#cesec264} ------------------ •Varicella vaccine is routinely used for vaccination of healthy children in only some countries, including the United States, Uruguay, Qatar, Australia, Canada, Costa Rica, Germany, and South Korea.•The risk for varicella infection is higher for people traveling to most other parts of the world than it is in the United States. However, VZV is still widely circulating in the United States. Additionally, exposure to herpes zoster (shingles), while less common than varicella, poses a risk for varicella infection in susceptible travelers.•Travelers at highest risk for severe varicella disease are immunocompromised persons or pregnant women without a history of varicella disease or vaccination. Clinical Presentation {#cesec265} --------------------- •Varicella is generally a mild disease in children. It usually lasts 4--7 days and is characterized by a short (1- to 2-day) or absent prodromal period (low-grade fever, malaise) and by a pruritic rash consisting of crops of macules, papules, and vesicles (on average 250--500 lesions), which appear in three or more successive waves and resolve by crusting.•Serious complications are the exception but can occur, mainly in infants, adolescents, adults, and immunocompromised persons. They include secondary bacterial infections of skin lesions, pneumonia, cerebellar ataxia, and encephalitis.•The average incubation period for varicella is 14--16 days (range 10--21 days).•A modified varicella, known as breakthrough disease, can occur in some vaccinated persons, because the vaccine is 70%--90% effective in preventing disease. Breakthrough varicella is most commonly (∼70%--80% of cases) mild, with \<50 skin lesions, less fever, and shorter duration of rash. The rash may be atypical in appearance with fewer vesicles and predominance of maculopapular lesions. Nevertheless, breakthrough varicella is infectious (although less than varicella in unvaccinated persons). Persons with breakthrough varicella should be isolated for as long as lesions persist. Diagnosis {#cesec266} --------- •Skin lesions are the preferred specimen for laboratory confirmation of varicella disease.○Vesicular fluid or a scab can be used to identify VZV by using polymerase chain reaction (PCR). Rapid diagnostic tests (PCR, direct fluorescent antibody) are the methods of choice.○VZV can also be isolated from scrapings of a vesicle base during the first 3--4 days of the eruption.○Collecting skin lesion specimens from breakthrough cases can be challenging because the rash is often maculopapular with few or no vesicles. If lesions are not present, scraping of the lesion is recommended.•Serologic tests for confirmation of disease:○A significant rise in serum varicella IgG antibody from acute- and convalescent-phase samples by any standard serologic assay can confirm a diagnosis retrospectively, but may not be reliable in immunocompromised people.○Commercially available tests are not sufficiently sensitive to reliably demonstrate vaccine-induced immunity. Postexposure Prophylaxis {#cesec267} ------------------------ ### Vaccine {#cesec268} •Varicella vaccine is recommended for postexposure administration for healthy unvaccinated persons without other evidence of immunity.•Administration of varicella vaccine to exposed susceptible persons Δ12 months of age, as soon as possible within 72 hours and possibly up to 120 hours after exposure, may prevent or modify disease and is recommended if there are no contraindications to use. In several studies, protective efficacy was reported as Δ90% when children were vaccinated within 3 days of exposure. ### Use of Varicella Zoster Immune Globulin (VZIG) {#cesec269} •In certain circumstances, postexposure prophylaxis with VZIG is recommended.•The decision to administer VZIG to a person exposed to varicella should be based on 1) whether the person is susceptible, 2) whether the exposure is likely to result in infection, and 3) whether the person is at greater risk for complications than the general population.•Persons at greater risk for severe complications who are not candidates for varicella vaccination who may benefit from postexposure prophylaxis with VZIG include:○susceptible immunocompromised persons (including people being treated with chronic corticosteroids Δ2 mg/kg of body weight or a total of 20 mg/day of prednisone or equivalent)○susceptible pregnant women○newborns whose mothers had onset of varicella within 5 days before and 2 days after delivery○preterm infants at Δ28 weeks gestation whose mothers are susceptible to varicella○preterm infants at \<28 weeks gestation or £1,000 g birth weight, regardless of maternal history or serostatus.•VZIG provides maximum benefit when administered as soon as possible after exposure, but may be effective if administered as late as 96 hours after exposure.•The product currently in use in the United States, VariZIG, is available under an Investigational New Drug protocol and can be obtained from the sole authorized U.S. distributor, FFF enterprises (Temecula, California) (24-hour telephone, 800-843-7477 or [www.fffenterprises.com](http://www.fffenterprises.com){#interref54}).•If administration of VariZIG does not appear possible within 96 hours of exposure, administration of immune globulin intravenous (IGIV) should be considered as an alternative (IGIV should also be administered within 96 hours of exposure). Treatment {#cesec270} --------- •Oral acyclovir is not recommended for routine use in healthy children with varicella but should be considered for otherwise healthy people at increased risk for moderate to severe disease, e.g.: persons aged \>12 years; people with chronic cutaneous or pulmonary disorders; receiving long-term salicylate therapy; and receiving short, intermittent or aerosolized courses of corticosteroids.•Intravenous antiviral therapy, when administered within 24 hours of onset of rash is recommended for immunocompromised persons, including patients being treated with chronic corticosteroids. Preventive Measures for Travelers {#cesec271} --------------------------------- Although vaccination against varicella is not a requirement for entry into any country (including the United States), persons traveling or living abroad should ensure that they are immune. ### Vaccine {#cesec272} •Varicella vaccine contains live, attenuated VZV. It is available as a monovalent formulation and in combination formulation, as measles--mumps--rubella--varicella (MMRV) vaccine, which is licensed in the United States for children 1--12 years only.•Two doses of varicella-containing vaccine are now recommended for all susceptible persons older than one year without contraindications. The first dose should be administered at 12--15 months of age and the second dose at 4--6 years of age. A second catch-up dose of varicella vaccination is recommended for children, adolescents and adults who previously have received one dose. The minimum interval for children younger than 13 years is 3 months. The ACIP now recommends that all others at least 13 years of age without evidence of immunity be vaccinated with two doses of varicella vaccine at an interval of 4--8 weeks. In case of uncertainty, prior varicella disease is not a contraindication to varicella vaccination.•Evidence of immunity to varicella includes any of the following:○Documentation of age-appropriate vaccination:□Preschool-age children aged Δ12 months: 1 dose□School-age children, adolescents, and adults: 2 doses○Laboratory evidence of immunity or laboratory confirmation of disease○Birth in the United States before 1980 (not a criterion for health-care personnel, pregnant women, and immunocompromised persons)○A health-care provider diagnosis of varicella or a health-care provider verification of a history of varicella disease○A health-care provider diagnosis of herpes zoster or a health-care provider verification of a history of herpes zoster disease #### Adverse Reactions {#cesec273} •The most common adverse reactions following varicella vaccine are injection site complaints (pain, soreness, redness, and swelling) that are self-limited. Fever was reported in uncontrolled trials in 15% of children and 10% of adolescents and adults. A macular or vaccine rash usually consisting of a few lesions at the injection site was reported in 3% and 1% of persons receiving the first and second dose, respectively. A generalized rash with a small number of lesions may rarely occur, within 3 weeks of vaccination.•Varicella vaccine is a live-virus vaccine that induces latent infection similar to that caused by wild VZV. Consequently, zoster caused by vaccine virus has been reported. This appears to occur at a lower rate than following natural infection but longer term follow-up is needed. #### Contraindications {#cesec274} ##### Allergy {#cesec275} •Persons with severe allergy (hives, swelling of the mouth or throat, difficulty breathing, hypotension, and shock) to gelatin or neomycin or who have had a severe allergic reaction to a prior dose of vaccine should not be vaccinated.•Single-antigen varicella vaccine does not contain egg protein or preservative. For the combination MMRV vaccine, live measles and live mumps vaccine are produced in chick embryo culture. However, the risk for serious allergic reactions after administration of measles- or mumps-containing vaccines in persons who are allergic to eggs is low. ##### Altered Immunity {#cesec276} Persons with immunosuppression of cellular immune function resulting from leukemia, lymphomas of any type, generalized malignancy, immunodeficiency disease, or immunosuppressive therapy should not be vaccinated. Treatment with low-dose prednisone (e.g., \<2 mg/kg of body weight/day or \<20 mg/day) or aerosolized steroid preparations is not a contraindication to varicella vaccination. Persons whose immunosuppressive therapy with steroids has been stopped for 1 month (3 months for chemotherapy) may be vaccinated. In addition, persons with impaired humoral immunity may now be vaccinated. Because children infected with HIV are at greater risk for morbidity from varicella and herpes zoster than are healthy children, the ACIP recommends that varicella vaccine should be considered for HIV-infected children at least 12 months of age with CD4+ T-lymphocyte percentages Δ15% and without evidence of varicella immunity. Eligible children should receive two doses of single-antigen varicella vaccine, with a minimum 3-month interval between doses. Vaccination (two doses, administered 3 months apart) may be considered for HIV-infected older children, adolescents and adults with CD4+ T-lymphocyte count Δ200 cells/mL, after weighing the risks and benefits. ##### Pregnancy {#cesec277} Women known to be pregnant or attempting to become pregnant should not receive varicella vaccine. Pregnancy should be avoided for 1 month following varicella vaccination. Breastfeeding is not a contraindication to the varicella vaccination. #### Precautions {#cesec278} ##### Illness {#cesec279} Vaccination of persons who have acute severe illness, including untreated, active tuberculosis, should be postponed until recovery. ##### Recent Administration of Blood, Plasma, or Immune Globulin {#cesec280} The effect of the administration of immune globulin (IG) on the response to varicella virus vaccine is unknown. Because of the potential inhibition of the antibody response by passively transferred antibodies, varicella vaccines should not be administered for 3--11 months, depending on the dosage, after administration of blood (except washed red cells), plasma, or IG. ##### Use of Salicylates {#cesec281} No adverse events following varicella vaccination related to the use of salicylates (e.g., aspirin) have been reported to date. However, the manufacturer recommends that vaccine recipients avoid the use of salicylates for 6 weeks after receiving varicella vaccine because of the association between aspirin use and Reye syndrome following varicella.   {#cesec282} = Malaria MALARIA {#subchapter21} ======= Arguin Paul M. Steele Stefanie F. Infectious Agent {#cesec283} ---------------- Malaria in humans is caused by one of four protozoan species of the genus *Plasmodium*: *P. falciparum, P. vivax, P. ovale*, or *P. malariae*. Recently, *P. knowlesi*, a parasite of Old World monkeys, has been documented as a cause of human infections and some fatalities in Southeast Asia. Investigations are ongoing to determine the extent of its transmission to humans. Mode of Transmission {#cesec284} -------------------- All species are transmitted by the bite of an infected female *Anopheles* mosquito. Occasionally, transmission occurs by blood transfusion, organ transplantation, needle sharing, or congenitally from mother to fetus. Occurrence {#cesec285} ---------- •Each year malaria causes 350--500 million infections worldwide and approximately 1 million deaths.•Transmission occurs in large areas of Central and South America, parts of the Caribbean, Africa, Asia (including South Asia, Southeast Asia, and the Middle East), Eastern Europe, and the South Pacific ([Maps 2-7](#f10){ref-type="fig"} and [2-8](#f11){ref-type="fig"} ).Map 2-7Malaria-endemic countries in the Western Hemisphere.Map 2-8Malaria-endemic countries in the Eastern Hemisphere.•Information about malaria transmission in specific countries (see the [Malaria Risk Information and Prophylaxis, by Country](#subchapter22){ref-type="sec"}, section later in this chapter) is derived from various sources, including WHO.•Tools such as the interactive malaria map can assist in locating more unusual destinations and determining if malaria transmission occurs there (see [www.cdc.gov/malaria/risk_map/](http://www.cdc.gov/malaria/risk_map/){#interref56}). Risk for Travelers {#cesec286} ------------------ •The risk for a traveler acquiring malaria differs substantially from region to region and from traveler to traveler, even within a single country.•From 1997 through 2006, 10,745 cases of malaria among U.S. residents were reported to CDC. Of these, 6,376 (59.3%) were acquired in sub-Saharan Africa; 1,498 (13.9%) in Asia; 1,427 (13.3%) in the Caribbean and Central and South America; and 278 (0.03%) in Oceania. During this period, 54 fatal malaria infections occurred among U.S. residents; 46 (85.2%) were caused by *P. falciparum*, of which 33 (71.1%) were acquired in sub-Saharan Africa.•These absolute numbers of cases should be considered within the context of the volume of travel to these locations. Regions with the highest estimated relative risk for infection for travelers are West Africa and Oceania. Regions with moderate estimated relative risk for infection are the other parts of Africa, South Asia, and South America. Regions with lower estimated relative risk are Central America and other parts of Asia. There is considerable country-by-country variation, as well as variable transmission within countries and sometimes seasonal variation.•Prevention of malaria involves striking a balance between ensuring that all people who will be at risk for infection use the appropriate prevention measures, while preventing adverse effects of those interventions among people using them unnecessarily. An individual risk assessment should be conducted for every traveler, taking into account not only the destination country, but also the detailed itinerary, including specific cities, types of accommodation, season, and style of travel. In addition, conditions such as pregnancy or the presence of antimalarial drug resistance at the destination may modify the risk assessment.•Depending on level of risk, it may be appropriate to recommend no specific interventions, mosquito avoidance measures only, or mosquito avoidance measures plus chemoprophylaxis.•For areas of intense transmission, such as West Africa, exposure for even short periods of time can result in transmission, so this area should be considered high risk.•Malaria risk is not distributed homogeneously throughout all countries. Some destinations have malaria transmission occurring throughout the whole country, while in others it occurs in defined pockets. If travelers are going to the high-risk pockets during peak transmission times, even though the country as a whole may be low risk, this destination for this individual may be high risk.•Geography is just one part of determining a traveler\'s risk for infection. Risk can differ substantially for different travelers if their behaviors and circumstances differ. For example, travelers staying in air-conditioned hotels may be at lower risk than backpackers or adventure travelers. Similarly, long-term residents living in screened and air-conditioned housing are less likely to be exposed than are persons living without such amenities.•The highest risk is associated with first- and second-generation immigrants living in nonendemic countries who return to their countries of origin to visit friends and relatives (VFRs). VFR travelers often consider themselves to be at no risk because they grew up in a malarious country and consider themselves immune. However, acquired immunity is lost very quickly, and VFRs should be considered as having the same risk as otherwise nonimmune travelers.•Travelers should also be reminded that even if one has had malaria before, one can get it again and preventive measures are still necessary. All travelers going to malaria-endemic countries, even for short periods of time, such as cruise ship passengers, may be at risk for becoming infected with malaria.•Persons who have been in an area where malaria transmission occurs, either during daytime or nighttime hours, are not permitted to donate blood in the United States for a period of time after returning from the malarious area. Persons who are residents of nonmalarious countries are not permitted to donate blood for 1 year after they have returned from a malarious area. Persons who are residents of malarious countries are not permitted to donate blood for 3 years after leaving a malarious area. Persons who have had malaria are not allowed to donate blood for 3 years after treatment for malaria.•Risk assessments may differ between travel medicine providers and blood banks. A travel medicine provider advising a traveler going to a relatively low-risk country for a short period of time and engaging in behaviors that place them at lower risk for exposure may choose insect avoidance only and no chemoprophylaxis for the traveler. However, upon the traveler\'s return, a blood bank may still choose to defer that traveler for 1 year because of the travel to an area where transmission occurs. Clinical Presentation {#cesec287} --------------------- •Malaria is characterized by fever and influenza-like symptoms, including chills, headache, myalgias, and malaise; these symptoms can occur at intervals.•Uncomplicated disease may be associated with anemia and jaundice. In severe disease, most commonly caused by *P. falciparum*, seizures, mental confusion, kidney failure, acute respiratory disease syndrome (ARDS), coma, and death may occur.•Malaria symptoms can develop as early as 7 days (usually at least 14 days) after initial exposure in a malaria-endemic area and as late as several months or more after departure. Diagnosis {#cesec288} --------- •Travelers who have symptoms of malaria should be advised to seek medical evaluation **as soon as possible**.•Smear microscopy remains the gold standard for malaria diagnosis. Microscopy can also be used to determine the species of malaria parasite and quantify the parasitemia---both of which are necessary pieces of information for providing the most appropriate treatment.•Various test kits are available to detect antigens derived from malaria parasites. Such immunologic (immunochromatographic) tests most often use a dipstick or cassette format and provide results in 2--15 minutes. These rapid diagnostic tests (RDTs) offer a useful alternative to microscopy in situations where reliable microscopic diagnosis is not available. The U.S. Food and Drug Administration (FDA) has approved one RDT for use in the United States by hospital and commercial laboratories, not by individual clinicians or by patients themselves. This RDT, called BinaxNOW Malaria test, is produced by Inverness Medical Professional Diagnostics, located in Scarborough, Maine.•Polymerase chain reaction (PCR) tests are also available for detecting malaria parasites; however, none are FDA-approved. Although these tests are slightly more sensitive than routine microscopy, results are not usually available as quickly as microscopy results should be, thus limiting the clinical utility of this test. PCR testing can be used to determine the species of the parasite if the microscopic results are ambiguous.•In sub-Saharan Africa, the rate of false-positive blood films for malaria may be very high. Travelers to this region should be warned they may be diagnosed with malaria incorrectly, even though they are taking a reliable antimalarial regimen. In such cases, acutely ill travelers should be advised to seek the best available medical services and follow the treatment offered locally (except the use of halofantrine which is not recommended; see below), but **not** to stop their chemoprophylaxis regimen. Treatment {#cesec289} --------- •Malaria can be treated effectively early in the course of the disease, but delay of appropriate therapy can have serious or even fatal consequences.•Travelers who have symptoms of malaria should be advised to seek medical evaluation **as soon as possible**.•Specific treatment options depend on the species of malaria, the likelihood of drug resistance (based on the location of acquisition of infection), the age of the patient, pregnancy status, and the severity of infection. If possible, it is advisable to consult with a provider who has specialized travel/tropical medicine expertise or with an infectious disease physician.•CDC recommendations for malaria treatment can be found at [www.cdc.gov/malaria/diagnosis_treatment/treatment.htm](http://www.cdc.gov/malaria/diagnosis_treatment/treatment.htm){#interref57}.•Medications that are not used in the United States for the treatment of malaria, such as halofantrine (Halfan), are widely available overseas. CDC does not recommend halofantrine for treatment because of cardiac adverse events, including deaths, which have been documented following treatment doses. These adverse events have occurred in persons with and without pre-existing cardiac problems and both in the presence and absence of other antimalarial drugs (e.g., mefloquine). ### Self-Treatment ([Table 2-22](#cetable22){ref-type="table"}) {#cesec290} •Travelers who reject the advice to take prophylaxis, who choose a suboptimal drug regimen (e.g., chloroquine in an area with chloroquine-resistant *P. falciparum*), or who require a less-than-optimal drug regimen for medical reasons are at greater risk for acquiring malaria and needing prompt treatment.•Travelers who are taking effective prophylaxis but who will be in very remote areas may decide, in consultation with their health-care provider, to take along a full course of an approved malaria treatment regimen for self-treatment. This should occur very rarely.•Travelers should be advised to take their presumptive self-treatment promptly if they have fever, chills, or other influenza-like illness and if professional medical care is not available within 24 hours. **Travelers should be advised that this self-treatment of a possible malarial infection is only a temporary measure and that prompt medical evaluation is imperative**.•Atovaquone/proguanil may be used for presumptive self-treatment for travelers NOT taking atovaquone/proguanil for prophylaxis. If taking atovaquone/proguanil for prophylaxis, the use of the same drug at therapeutic doses is not recommended to empirically treat fever (suspected malaria). The CDC Malaria Branch (Malaria Hotline 770-488-7788) can provide consultation to health-care providers on other potential options for self-treatment if atovaquone/proguanil cannot be used. Table 2-22Presumptive self-treatment of malariaDrugAdult DosePediatric DoseCommentsAtovaquone/proguanil (Malarone). Self-treatment drug to be used if professional medical care is not available within 24 hours. Medical care should be sought immediately after treatment.4 tablets (each dose contains 1,000 mg atovaquone and 400 mg proguanil) orally as a single daily dose for 3 consecutive daysDaily dose to be taken for 3 consecutive days: 5--8 kg: 2 pediatric tablets; 9--10 kg: 3 pediatric tablets; 11--20 kg: 1 adult tablet; 21--30 kg: 2 adult tablets; 31--40 kg: 3 adult tablets; \>41 kg: 4 adult tabletsContraindicated in persons with severe renal impairment (creatinine clearance \<30 mL/min). Not recommended for self-treatment in persons on atovaquone/proguanil prophylaxis. Not currently recommended for children \<5 kg, pregnant women, and women breastfeeding infants weighing \<5 kg ### Malaria Hotline {#cesec291} •Health-care professionals who require assistance with the diagnosis or treatment of malaria should call the CDC Malaria Hotline (770-488-7788) from 8:00 am to 4:30 pm Eastern time. After hours or on weekends and holidays, health-care providers requiring assistance should call the CDC Emergency Operations Center at 770-488-7100 and ask the operator to page the person on call for the Malaria Branch.•Information on diagnosis and treatment is available at [www.cdc.gov/malaria](http://www.cdc.gov/malaria){#interref58}. Preventive Measures for Travelers {#cesec292} --------------------------------- Malaria prevention consists of a combination of mosquito avoidance measures and chemoprophylaxis. Although very efficacious, none of the recommended interventions are 100% effective. ### Mosquito Avoidance Measures {#cesec293} •Because of the nocturnal feeding habits of *Anopheles* mosquitoes, malaria transmission occurs primarily between dusk and dawn.•Contact with mosquitoes can be reduced by remaining in well-screened areas, using mosquito bed nets (preferably insecticide-treated nets), using a pyrethroid-containing flying-insect spray in living and sleeping areas during evening and nighttime hours, and wearing clothes that cover most of the body.•All travelers should use an effective mosquito repellent.•The most effective repellent against a wide range of vectors is DEET (*N*,*N*-diethylmetatoluamide), an ingredient in many commercially available insect repellents. The actual concentration of DEET varies widely among repellents. DEET formulations as high as 50% are recommended for both adults and children older than 2 months of age (see the [Protection Against Mosquitoes, Ticks, and Other Insects and Arthropods](#subchapter29){ref-type="sec"} section later in this chapter). DEET should be applied to the exposed parts of the skin when mosquitoes are likely to be present.•In addition to using a topical insect repellent, a permethrin-containing product may be applied to bed nets and clothing for additional protection against mosquitoes. ### Chemoprophylaxis {#cesec294} •All currently recommended primary chemoprophylaxis regimens involve taking a medicine before travel, during travel, and for a period of time after leaving the malaria endemic area. Beginning the drug before travel allows the antimalarial agent to be in the blood before the traveler is exposed to malaria parasites.•Presumptive antirelapse therapy (also known as terminal prophylaxis) uses a medication towards the end of the exposure period (or immediately thereafter) to prevent relapses or delayed-onset clinical presentations of malaria caused by hypnozoites (dormant liver stages) of *P. vivax* or *P. ovale*. Because most malarious areas of the world (except the Caribbean) have at least one species of relapsing malaria, travelers to these areas have some risk for acquiring either *P. vivax* or *P. ovale*, although the actual risk for an individual traveler is difficult to define. Presumptive anti-relapse therapy is generally indicated only for persons who have had prolonged exposure in malaria-endemic areas (e.g., missionaries, volunteers).•In choosing an appropriate chemoprophylactic regimen before travel, the traveler and the health-care provider should consider several factors. The travel itinerary should be reviewed in detail and compared with the information on where malaria transmission occurs within a given country (see the [Malaria Risk Information and Prophylaxis, by Country](#subchapter22){ref-type="sec"}, section later in this chapter) to determine whether the traveler will actually be traveling in a part of the country where malaria occurs and if significant antimalarial drug resistance has been reported in that location.•The resistance of *P. falciparum* to chloroquine has been confirmed in all areas with *P. falciparum* malaria except the Caribbean, Central America west of the Panama Canal, and some countries in the Middle East. In addition, resistance to sulfadoxine--pyrimethamine (e.g., Fansidar) is widespread in the Amazon River Basin area of South America, much of Southeast Asia, other parts of Asia, and in large parts of Africa. Resistance to mefloquine has been confirmed on the borders of Thailand with Burma (Myanmar) and Cambodia, in the western provinces of Cambodia, in the eastern states of Burma (Myanmar), on the border between Burma and China, along the borders of Laos and Burma, and the adjacent parts of the Thailand--Cambodia border, as well as in southern Vietnam ([Map 2-9](#f12){ref-type="fig"} ).Map 2-9Geographic distribution of mefloquine-resistant malaria.•Additional factors to consider are the patient\'s other medical conditions, medications being taken (to assess potential drug--drug interactions), the cost of the medicines, and the potential side effects.•The medications recommended for chemoprophylaxis of malaria may also be available at overseas destinations. However, combinations of these medications and additional drugs that are not recommended may be commonly prescribed and used in other countries. Travelers should be strongly discouraged from obtaining chemoprophylactic medications while abroad. The quality of these products is not known, and they may not be protective and may be dangerous. These medications may have been produced by substandard manufacturing practices, may be counterfeit, or may contain contaminants. Additional information on this topic can be found in *Perspectives:* Counterfeit Drugs later in this chapter and in an FDA document Purchasing Medications Outside the United States ([www.fda.gov/ora/import/purchasing_medications.htm](http://www.fda.gov/ora/import/purchasing_medications.htm){#interref59}). #### Medications Used for Chemoprophylaxis {#cesec295} ##### Atovaquone/Proguanil (Malarone) {#cesec296} •Atovaquone/proguanil is a fixed combination of the two drugs, atovaquone and proguanil.•Prophylaxis should begin 1--2 days before travel to malarious areas and should be taken daily, at the same time each day, while in the malarious areas, and daily for 7 days after leaving the area (see [Table 2-23](#cetable23){ref-type="table"} for recommended dosages).Table 2-23Drugs used in the prophylaxis of malariaDrugUsageAdult DosePediatric DoseCommentsAtovaquone/proguanil (Malarone)Prophylaxis in all areasAdult tablets contain 250 mg atovaquone and 100 mg proguanil hydrochloride. 1 adult tablet orally, dailyPediatric tablets contain 62.5 mg atovaquone and 25 mg proguanil hydrochloride. 5-8 kg: frac12; pediatric tablet daily; \>8--10 kg: frac34; pediatric tablet daily; \>10-20 kg: 1 pediatric tablet daily; \>20-30 kg: 2 pediatric tablets daily; \>30-40 kg: 3 pediatric tablets daily; \>40 kg: 1 adult tablet dailyBegin 1-2 days before travel to malarious areas. Take daily at the same time each day while in the malarious area and for 7 days after leaving such areas. Contraindicated in persons with severe renal impairment (creatinine clearance \<30 mL/min). Atovaquone/proguanil should be taken with food or a milky drink. Not recommended for prophylaxis for children \<5 kg, pregnant women, and women breastfeeding infants weighing \<5 kg. Partial tablet dosages may need to be prepared by a pharmacist and dispensed in individual capsules, as described in the text.Chloroquine phosphate (Aralen and generic)Prophylaxis only in areas with chloroquine-sensitive malaria300 mg base (500 mg salt) orally, once/week5 mg/kg base (8.3 mg/kg salt) orally,once/week, up to maximum adult dose of 300 mg baseBegin 1-2 weeks before travel to malarious areas. Take weekly on the same day of the week while in the malarious area and for 4 weeks after leaving such areas. May exacerbate psoriasisDoxycycline (many brand names and generic)Prophylaxis in all areas100 mg orally, daily\>8 years of age: 2 mg/kg up to adult dose of 100 mg/dayBegin 1-2 days before travel to malarious areas. Take daily at the same time each day while in the malarious area and for 4 weeks after leaving such areas. Contraindicated in children \<8 years of age and pregnant womenHydroxychloroquine sulfate (Plaquenil)An alternative to chloroquine for prophylaxis only in areas with chloroquine-sensitive malaria310 mg base (400 mg salt) orally, once/week5 mg/kg base (6.5 mg/kg salt) orally,once/week, up to maximum adult dose of 310 mg baseBegin 1-2 weeks before travel to malarious areas. Take weekly on the same day of the week while in the malarious area and for 4 weeks after leaving such areas.Mefloquine (Lariam and generic)Prophylaxis in areas with mefloquine-sensitive malaria228 mg base (250 mg salt) orally, once/week\<9 kg: 4.6 mg/kg base (5 mg/kg salt) orally, once/week; \>9--19 kg: frac14; tablet once/week; \>19-30 kg: frac12; tablet once/week;Begin 1-2 weeks before travel to malarious areas. Take weekly on the same day of the week while in the malarious area and for 4 weeks after leaving such areas. Contraindicated in persons allergic to mefloquine or related compounds (e.g., quinine, quinidine) and in persons with active depression, a recent history of depression, generalized anxiety disorder,Mefloquine (Lariam and generic)\>31--45 kg: frac34; tablet once/week; \>45 kg: 1 tablet once/weekpsychosis, schizophrenia, other major psychiatric disorders, or seizures. Use with caution in persons with psychiatric disturbances or a previous history of depression. Not recommended for persons with cardiac conduction abnormalitiesPrimaquineProphylaxis for short-duration travel to areas with principally *P. vivax*30 mg base (52.6 mg salt) orally, daily0.5 mg/kg base (0.8 mg/kg salt) up to adult dose orally, dailyBegin 1-2 days before travel to malarious areas. Take daily at the same time each day while in the malarious area and for 7 days after leaving such areas.Contraindicated in persons with G6PD deficiency. Also contraindicated during pregnancy and lactation unless the infant being breastfed has a documented normal G6PD levelPrimaquineUsed for presumptive antirelapse therapy (terminal prophylaxis) to decrease the risk for relapses of *P. vivax and P. ovale*30 mg base (52.6 mg salt) orally, once/day for 14 days after departure from the malarious area0.5 mg/kg base (0.8 mg/kg salt) up to adult dose orally, once/day for 14 days after departure from the malarious areaIndicated for persons who have had prolonged exposure to *P. vivax* and *P. ovale* or both. Contraindicated in persons with G6PD deficiency. Also contraindicated during pregnancy and lactation unless the infant being breastfed has a documented normal G6PD level[^79]•Malarone is very well tolerated, and side effects are rare. The most common adverse effects reported in persons using atovaquone/proguanil for prophylaxis or treatment are abdominal pain, nausea, vomiting, and headache. Malarone should not be used for prophylaxis in children weighing \<5 kg, pregnant women, or patients with severe renal impairment (creatinine clearance \<30 mL/min). It should be used with caution by patients taking coumadin (warfarin) for anticoagulation. ##### Chloroquine (Aralen) and Hydroxychloroquine (Plaquenil) {#cesec297} •Chloroquine phosphate or hydroxychloroquine sulfate can be used for prevention of malaria only in destinations where chloroquine resistance is not present (see [Maps 2-7](#f10){ref-type="fig"} and [2-8](#f11){ref-type="fig"} or the next section in this chapter, [Malaria Risk Information and Prophylaxis, by Country](#subchapter22){ref-type="sec"}).•Prophylaxis should begin 1--2 weeks before travel to malarious areas. It should be continued by taking the drug once a week, on the same day of the week, during travel in malarious areas and for 4 weeks after a traveler leaves these areas (see [Table 2-23](#cetable23){ref-type="table"} for recommended dosages).•Reported side effects include gastrointestinal disturbance, headache, dizziness, blurred vision, insomnia, and pruritus, but generally these effects do not require that the drug be discontinued. High doses of chloroquine, such as those used to treat rheumatoid arthritis, have been associated with retinopathy; this serious side effect appears to be extremely unlikely when chloroquine is used for routine weekly malaria prophylaxis. Chloroquine and related compounds have been reported to exacerbate psoriasis. Persons who experience uncomfortable side effects after taking chloroquine may tolerate the drug better by taking it with meals. As an alternative, the related compound hydroxychloroquine sulfate may be better tolerated. Box 2-2Clinical pearls•Overdose of antimalarial drugs, particularly chloroquine, can be fatal. Medication should be stored in childproof containers out of the reach of infants and children.•Chemoprophylaxis can be started earlier if there are particular concerns about tolerating one of the medications. For example, mefloquine can be started 3--4 weeks in advance to allow potential adverse events to occur before travel. If unacceptable side effects develop, there would be time to change the medication before the traveler\'s departure.•The drugs used for antimalarial chemoprophylaxis are generally well tolerated. However, side effects can occur. Minor side effects usually do not require stopping the drug. Travelers who have serious side effects should see a health-care provider who can determine if their symptoms are related to the medicine and make an appropriate medication change.•In comparison with drugs with short half-lives, which are taken daily, drugs with longer half-lives, which are taken weekly, offer the advantage of a wider margin of error if the traveler is late with a dose. For example, if a traveler is 1--2 days late with a weekly drug, prophylactic blood levels can remain adequate; if the traveler is 1--2 days late with a daily drug, protective blood levels are less likely to be maintained.•In those who are G6PD deficient, primaquine can cause hemolysis, which can be fatal. Be sure to document a normal G6PD level before prescribing primaquine.•Travelers should be informed that malaria can be fatal if treatment is delayed. Medical help should be sought promptly if malaria is suspected, and a blood sample should be taken and examined for malaria parasites on one or more occasions.•Malaria smear results or an RDT test must be available immediately. Sending specimens to offsite laboratories where results are not available for extended periods of time (days) is not acceptable. If a patient has an illness suggestive of severe malaria and a compatible travel history in an area where malaria transmission occurs, it is advisable to start treatment as soon as possible, even before the diagnosis is established. CDC recommendations for malaria treatment can be found at [www.cdc.gov/malaria/diagnosis_treatment/treatment.htm](http://www.cdc.gov/malaria/diagnosis_treatment/treatment.htm){#interref60}. ##### Doxycycline (Many Brand Names and Generic) {#cesec298} •Doxycycline prophylaxis should begin 1--2 days before travel to malarious areas. It should be continued once a day, at the same time each day, during travel in malarious areas and daily for 4 weeks after the traveler leaves such areas.•Insufficient data exist on the antimalarial prophylactic efficacy of related compounds such as minocycline (commonly prescribed for the treatment of acne). Persons on a long-term regimen of minocycline who are in need of malaria prophylaxis should stop taking minocycline 1--2 days before travel and start doxycycline instead. The minocycline can be restarted after the full course of doxycycline is completed (see [Table 2-23](#cetable23){ref-type="table"} for recommended dosages).•Doxycycline can cause photosensitivity, usually manifested as an exaggerated sunburn reaction. The risk for such a reaction can be minimized by avoiding prolonged, direct exposure to the sun and by using sunscreens. In addition, doxycycline use is associated with an increased frequency of vaginal yeast infections. Gastrointestinal side effects (nausea or vomiting) may be minimized by taking the drug with a meal. To reduce the risk for esophagitis, travelers should be advised not to take doxycycline before going to bed. Doxycycline is contraindicated in persons with an allergy to tetracyclines, during pregnancy, and in infants and children \<8 years of age.•Vaccination with the oral typhoid vaccine Ty21a should be delayed for at least 24 hours after taking a dose of doxycycline. ##### Mefloquine (Lariam) {#cesec299} •Mefloquine prophylaxis should begin 1--2 weeks before travel to malarious areas. It should be continued once a week, on the same day of the week, during travel in malarious areas and for 4 weeks after a traveler leaves such areas (see [Table 2-23](#cetable23){ref-type="table"} for recommended dosages).•Mefloquine has been associated with rare serious adverse reactions (e.g., psychoses, seizures) at prophylactic doses; these reactions are more frequent with the higher doses used for treatment. Other side effects that have occurred in chemoprophylaxis studies include gastrointestinal disturbance, headache, insomnia, abnormal dreams, visual disturbances, depression, anxiety disorder, and dizziness. Other more severe neuropsychiatric disorders occasionally reported during postmarketing surveillance include sensory and motor neuropathies (including paresthesia, tremor, and ataxia), agitation or restlessness, mood changes, panic attacks, forgetfulness, confusion, hallucinations, aggression, paranoia, and encephalopathy. On occasion, psychiatric symptoms have been reported to continue long after mefloquine has been stopped. Mefloquine is contraindicated for use by travelers with a known hypersensitivity to mefloquine or related compounds (e.g., quinine, quinidine) and in persons with active depression, a recent history of depression, generalized anxiety disorder, psychosis, schizophrenia, other major psychiatric disorders, or seizures. It should be used with caution in persons with psychiatric disturbances or a previous history of depression. A review of available data suggests that mefloquine may be used in persons concurrently on beta blockers, if they have no underlying arrhythmia. However, mefloquine is not recommended for persons with cardiac conduction abnormalities.•Any traveler receiving a prescription for mefloquine must also receive a copy of the FDA Medication Guide, which can be found at the following website: [www.fda.gov/cder/foi/label/2003/19591s19lbl_Lariam.pdf](http://www.fda.gov/cder/foi/label/2003/19591s19lbl_Lariam.pdf){#interref61}. ##### Primaquine {#cesec300} •Primaquine phosphate has two distinct uses for malaria prevention: primary prophylaxis and presumptive antirelapse therapy (also called terminal prophylaxis).•When taken for primary prophylaxis, primaquine should be taken 1--2 days before travel to malarious areas, daily, at the same time each day, while in the malarious areas, and daily for 7 days after leaving the areas (see [Table 2-23](#cetable23){ref-type="table"} for recommended dosages). Primary prophylaxis with primaquine obviates the need for presumptive antirelapse therapy.•When used for presumptive antirelapse therapy, primaquine is administered for 14 days after the traveler has left a malarious area. When chloroquine, doxycycline, or mefloquine is used for primary prophylaxis, primaquine is usually taken during the last 2 weeks of postexposure prophylaxis. When atovaquone/proguanil is used for prophylaxis, primaquine may be taken during the final 7 days of atovaquone/proguanil, and then for an additional 7 days. It is preferable that primaquine be given concurrently with the primary prophylaxis medication. However, if that is not feasible, the primaquine course should still be administered after the primary prophylaxis medication has been completed.•The most common adverse event in glucose-6-phosphate dehydrogenase (G6PD) in normal persons is gastrointestinal upset if primaquine is taken on an empty stomach. This problem is minimized or eliminated if primaquine is taken with food.•In G6PD-deficient persons, primaquine can cause hemolysis that can be fatal. **Before primaquine is used, G6PD deficiency MUST be ruled out by appropriate laboratory testing**. #### Travel to Areas with Limited Malaria Transmission {#cesec301} For destinations (see the next section in this chapter, [Malaria Risk Information and Prophylaxis, by Country](#subchapter22){ref-type="sec"}) where malaria cases occur sporadically and risk for infection to travelers is assessed as being very low, it is recommended that travelers use mosquito avoidance measures only, and no chemoprophylaxis should be prescribed. #### Travel to Areas with Mainly P. vivax *Malaria* {#cesec302} •For destinations where the main species of malaria present is *P. vivax*, in addition to mosquito avoidance measures, primaquine is a good choice for primary prophylaxis for travelers who are not G6PD-deficient. Its use for this indication is considered off-label use in the United States.•The predominant species of malaria and the recommended chemoprohylaxis medicines are listed in the following section in this chapter, [Malaria Risk Information and Prophylaxis, by Country](#subchapter22){ref-type="sec"}.•For persons unable to take primaquine, other drugs can be used as described below, depending on the presence of antimalarial drug resistance. #### Travel to Areas with Chloroquine-Sensitive Malaria {#cesec303} •For destinations where chloroquine-sensitive malaria is present, in addition to mosquito avoidance measures, the many effective chemoprophylaxis alternatives include chloroquine, atovaquone/proguanil, doxycycline, mefloquine, and in some instances primaquine for travelers who are not G6PD-deficient.•Longer-term travelers may prefer the convenience of weekly chloroquine, while shorter-term travelers may prefer the shorter course of atovaquone/proguanil or primaquine. #### Travel to Areas with Chloroquine-Resistant Malaria {#cesec304} For destinations where chloroquine-resistant malaria is present, in addition to mosquito avoidance measures, chemoprophylaxis options are limited to atovaquone/proguanil, doxycycline, and mefloquine. #### Travel to Areas with Mefloquine-Resistant Malaria {#cesec305} For destinations where mefloquine-resistant malaria is present, in addition to mosquito avoidance measures, chemoprophylaxis options are reduced to either atovaquone/proguanil or doxycycline. #### Chemoprophylaxis for Infants, Children, and Adolescents {#cesec306} •Infants of any age or weight or children and adolescents of any age can contract malaria. Therefore, all children traveling to malaria-risk areas should take an antimalarial drug.•In the United States, antimalarial drugs are available only in tablet form and may taste quite bitter. Pediatric dosages should be carefully calculated according to body weight but should never exceed adult dosage. Pharmacists can pulverize tablets and prepare gelatin capsules for each measured dose. If the child is unable to swallow the capsules or tablets, parents should prepare the child\'s dose of medication by breaking open the gelatin capsule and mixing the drug with a small amount of something sweet, such as applesauce, chocolate syrup, or jelly, to ensure the entire dose is delivered to the child. Giving the dose on a full stomach may minimize stomach upset and vomiting.•Chloroquine and mefloquine are options for use in infants and children of all ages and weights, depending on the presence of drug resistance at their destination.•Primaquine can be used for children who are not G6PD-deficient traveling to areas with principally *P. vivax*.•Doxycycline may be used for children who are at least 8 years of age.•Atovaquone/proguanil may be used for prophylaxis for infants and children weighing at least 5 kg (11 lbs). Providers should note that this prophylactic dosing for children weighing \<11 kg constitutes off-label use in the United States.•Pediatric dosing regimens are contained in [Table 2-23](#cetable23){ref-type="table"}. #### Chemoprophylaxis during Pregnancy and Breastfeeding {#cesec307} •Malaria infection in pregnant women can be more severe than in nonpregnant women. Malaria can increase the risk for adverse pregnancy outcomes, including prematurity, abortion, and stillbirth. For these reasons, and because no chemoprophylactic regimen is completely effective, women who are pregnant or likely to become pregnant should be advised to avoid travel to areas with malaria transmission if possible (see the Traveling while Pregnant section in Chapter 8). If travel to a malarious area cannot be deferred, use of an effective chemoprophylaxis regimen is essential.•Pregnant women traveling to areas where chloroquine-resistant *P. falciparum* has not been reported may take chloroquine prophylaxis. Chloroquine has not been found to have any harmful effects on the fetus when used in the recommended doses for malaria prophylaxis; therefore, pregnancy is not a contraindication for malaria prophylaxis with chloroquine phosphate or hydroxychloroquine sulfate.•For travel to areas where chloroquine resistance is present, mefloquine is currently the only medication recommended for malaria chemoprophylaxis during pregnancy. A review of mefloquine use in pregnancy from clinical trials and reports of inadvertent use of mefloquine during pregnancy suggests that its use at prophylactic doses during the second and third trimesters of pregnancy is not associated with adverse fetal or pregnancy outcomes. More limited data suggest it is also safe to use during the first trimester.•Because of insufficient data regarding the use during pregnancy, atovaquone/proguanil is not currently recommended for the prevention of malaria in pregnant women.•Doxycycline is contraindicated for malaria prophylaxis during pregnancy because of the risk for adverse effects seen with tetracycline, a related drug, on the fetus, which include discoloration and dysplasia of the teeth and inhibition of bone growth.•Primaquine should not be used during pregnancy because the drug may be passed transplacentally to a G6PD-deficient fetus and cause hemolytic anemia in utero.•Health-care professionals who require additional assistance with the management of pregnant travelers who are unable to take mefloquine chemoprophylaxis should call the CDC Malaria Hotline (770-488-7788).•Very small amounts of antimalarial drugs are excreted in the breast milk of lactating women. Because the quantity of antimalarial drugs transferred in breast milk is insufficient to provide adequate protection against malaria, infants who require chemoprophylaxis must receive the recommended dosages of antimalarial drugs listed in [Table 2-23](#cetable23){ref-type="table"}.•Because chloroquine and mefloquine may be safely prescribed to infants, it is also safe for infants to be exposed to the small amounts excreted in breast milk.•Although data are very limited about the use of doxycycline in lactating women, most experts consider the theoretical possibility of adverse events to the infant to be remote.•Although no information is available on the amount of primaquine that enters human breast milk, the mother and infant should be tested for G6PD deficiency before primaquine is given to a woman who is breastfeeding.•Because data are not yet available on the safety of atovaquone/proguanil prophylaxis in infants weighing \<5 kg (\<11 lbs), CDC does not currently recommend it for the prevention of malaria in women breastfeeding infants weighing \<5 kg. However, it can be used for treatment of women who are breastfeeding infants of any weight when the potential benefit outweighs the potential risk to the infant (e.g., treating a breastfeeding woman who has acquired *P. falciparum* malaria in an area of multidrug-resistant strains and who cannot tolerate other treatment options). #### Changing Medications during Chemoprophylaxis as a Result of Side Effects {#cesec308} •Medications recommended for prophylaxis against malaria have different modes of action that affect the parasites at different stages of the life cycle. Thus, if the medication needs to be changed because of side effects before a full course has been completed, there are some special considerations.•If a traveler starts prophylaxis with a medication such as mefloquine or doxycycline and then changes to atovaquone/proguanil during or after travel, the standard duration of prophylaxis for atovaquone/proguanil would be insufficient.•If the switch occurs 3 weeks or more before departure from the risk area, atovaquone/proguanil should be taken for the remainder of the stay in the risk area and for 1 week thereafter.•If the switch occurs \<3 weeks before departure from the risk area, atovaquone/proguanil should be taken for 4 weeks after the switch.•If the switch occurs following departure from the risk area, atovaquone/proguanil should be continued until 4 weeks after the date of departure from the risk area.•Due to their pharmacokinetics, switching from a daily medicine such as doxycycline to a weekly medicine such as mefloquine should be avoided.•Health-care professionals who require additional assistance with the management of travelers who need to change medications during prophylaxis should call the CDC Malaria Hotline (770-488-7788). MALARIA RISK INFORMATION AND PROPHYLAXIS, BY COUNTRY {#subchapter22} ==================================================== Tan Kathrine R. Mali Sonja Arguin Paul M. Table 2-24**Malaria risk information and prophylaxis, by country**[1](#cetablefn80){ref-type="table-fn"}CountryAreas with MalariaDrug Resistance[2](#cetablefn81){ref-type="table-fn"}Malaria Species[3](#cetablefn82){ref-type="table-fn"}Recommended Chemoprophylaxis[4](#cetablefn83){ref-type="table-fn"}**Afghanistan**April--December in all areas at altitudes \<2,000 m (\<6,561 ft)Chloroquine*P. vivax* 80%--90%\ *P. falciparum* 10%--20%Atovaquone/proguanil, doxycycline, or mefloquine**Albania**NoneNot applicableNot applicableNot applicable**Algeria**NoneNot applicableNot applicableNot applicable**Andorra**NoneNot applicableNot applicableNot applicable**Angola**AllChloroquine*P. falciparum* 90%\ *P. ovale* 5%\ *P. vivax* 5%Atovaquone/proguanil, doxycycline, or mefloquine**Anguilla** (U.K.)NoneNot applicableNot applicableNot applicable**Antarctica**NoneNot applicableNot applicableNot applicable**Antigua and Barbuda**NoneNot applicableNot applicableNot applicable**Argentina**Rural areas of Salta and Jujuy provinces (along Bolivian border) and Misiones and Corrientes provinces (along border of Paraguay). Malaria present in Iguassu FallsNone*P. vivax* 100%Atovaquone/proguanil, chloroquine, doxycycline, mefloquine, or primaquine[5](#cetablefn84){ref-type="table-fn"}**Armenia**Previously limited to the Ararat Valley in the Ararat and Artashat regions and Masis district. No cases reported since 2006NoneHistorically *P. vivax* 100%Mosquito avoidance only**Aruba**NoneNot applicableNot applicableNot applicable**Australia**, including **Cocos** (**Keeling**) **Islands**NoneNot applicableNot applicableNot applicable**Austria**NoneNot applicableNot applicableNot applicable**Azerbaijan**Rural areas \<1,500 m (4,921 ft). None in BakuNone*P. vivax* 100%Atovaquone/proguanil, chloroquine, doxycycline, mefloquine, or primaquine[5](#cetablefn84){ref-type="table-fn"}**Azores** (Portugal)NoneNot applicableNot applicableNot applicable**Bahamas, The**Present only in Great Exuma IslandNone*P. falciparum* 100%Atovaquone/proguanil, chloroquine, doxycycline, or mefloquine**Bahrain**NoneNot applicableNot applicableNot applicable**Bangladesh**All areas, except in city of DhakaChloroquine*P. falciparum* 77%\ *P. vivax* 23%Atovaquone/proguanil, doxycycline, or mefloquine**Barbados**NoneNot applicableNot applicableNot applicable**Belarus**NoneNot applicableNot applicableNot applicable**Belgium**NoneNot applicableNot applicableNot applicable**Belize**All areas, except in Belize CityNone*P. vivax* 95%\ *P. falciparum* 5%Atovaquone/proguanil, chloroquine, doxycycline, mefloquine, or primaquine[5](#cetablefn84){ref-type="table-fn"}**Benin**AllChloroquine*P. falciparum* 85%\ *P. ovale* 5%--10%\ *P. vivax* rareAtovaquone/proguanil, doxycycline, or mefloquine**Bermuda** (U.K.)NoneNot applicableNot applicableNot applicable**Bhutan**Rural areas \<1,700 m (\<5,577 ft) of the southern belt districts along the border with India: Chirang, Geylegphug, Samchi, Samdrup Jongkhar, Sarpang and ShemgangChloroquine*P. falciparum* 50%\ *P. vivax* 50%Atovaquone/proguanil, doxycycline, or mefloquine**Bolivia**All areas \<2,500 m (\<8,202 ft) in the following departments: Beni, Chuquisaca, Cochabamba, La Paz, Pando, Santa Cruz, and Tarija. None in city of La PazChloroquine*P. vivax* 70%-95%\ *P. falciparum* 5%-30%Atovaquone/proguanil, doxycycline, or mefloquine**Bosnia and Herzegovina**NoneNot applicableNot applicableNot applicable**Botswana**North of 22° S in the northern provinces of Central, Chobe, Ghanzi, and Ngamiland, including safaris to the Okavango Delta area. None in the city of GaboroneChloroquine*P. falciparum* 90%\ *P. vivax* 5%\ *P. ovale* 5%Atovaquone/proguanil, doxycycline, or mefloquine**Brazil**States of Acre, Rondônia, Amapá, Amazonas, Roraima, and Tocantins. Parts of states of Maranhaõ (western part), Mato Grosso (northern part), and Pará (except Belem City). Also present in urban areas, including large cities such as Porto Velho, Boa Vista, Macapa, Manaus, Santarem, and Maraba, where the transmission occurs on the periphery of these citiesMalaria in Iguassu FallsChloroquine*P. vivax* 75%\ *P. falciparum* 25%Atovaquone/proguanil, doxycycline, or mefloquine**British Indian Ocean Territory**, includes **Diego Garcia** (U.K.)NoneNot applicableNot applicableNot applicable**Brunei**NoneNot applicableNot applicableNot applicable**Bulgaria**NoneNot applicableNot applicableNot applicable**Burkina Faso**AllChloroquine*P. falciparum* 80%\ *P. ovale* 5%-10%\ *P. vivax* rareAtovaquone/proguanil, doxycycline, or mefloquine**Burma (Myanmar)**Rural areas throughout the country at altitudes \<1,000 m (\<3,281 ft). None in cities of Rangoon (Yangon) and MandalayChloroquine Mefloquine (see [Map 2-9](#f12){ref-type="fig"})*P. falciparum* 80%\ *P. vivax* 20%In the provinces of Bago, Kayah, Kachin, Kayin, Shan, and Tanintharyi: Atovaquone/proguanil or doxycyclineAll other areas: Atovaquone/proguanil, doxycycline, or mefloquine**Burundi**AllChloroquine*P. falciparum* \>85%\ *P. malariae, P. ovale,* and *P. vivax \<15%*Atovaquone/proguanil, doxycycline, or mefloquine**Cambodia**Present throughout country, including the temple complex at Angkor Wat, except none in Phnom Penh and around Lake Tonle SapChloroquine Mefloquine (see [Map 2-9](#f12){ref-type="fig"})*P. falciparum* 86%\ *P. vivax* 12%\ *P. malariae 2%*In the provinces of Preah Vihear, Siemreap, Oddar Meanchey, Banteay Meanchey, Battambang, Pailin, Kampot, Koh Kong, and Pursat bordering Thailand:Atovaquone/proguanil or doxycyclineAll other areas: Atovaquone/proguanil, doxycycline, or mefloquine**Cameroon**AllChloroquine*P. falciparum* 80%\ *P. ovale* 5%--10%\ *P. vivax* rareAtovaquone/proguanil, doxycycline, or mefloquine**Canada**NoneNot applicableNot applicableNot applicable**Canary Islands** (Spain)NoneNot applicableNot applicableNot applicable**Cape Verde**Limited to Sao Tiago IslandChloroquine*P. falciparum primarily*Atovaquone/proguanil, doxycycline, or mefloquine**Cayman Islands** (U.K.)NoneNot applicableNot applicableNot applicable**Central African Republic**AllChloroquine*P. falciparum* 85%\ *P. malariae, P. ovale,* and *P. vivax* 15%Atovaquone/proguanil, doxycycline, or mefloquine**Chad**AllChloroquine*P. falciparum* 85%\ *P. malariae, P. ovale,* and *P. vivax* 15%Atovaquone/proguanil, doxycycline, or mefloquine**Chile**NoneNot applicableNot applicableNot applicable**China**Rural parts of Anhui, Yunnan, Hainan provinces. Rare cases occur in other rural parts of the country \<1,500m(\<4,921 ft) during May-December. None in major river cruises and urban areasChloroguine (Vlefloguine (see [Map 2-9](#f12){ref-type="fig"})*P. falciparum* primarily in Hainan and Yunnan.\ *P. vivax* primarily elsewhereAlong China-Burma border in the western part of Yunnan province:Atovaguone/proguanil or doxycyclineHainan and the other parts of Yunnan province:Atovaguone/proguanil, doxycycline or mefloguineAnhui province: Atovaguone/proguanil, chloroguine, doxycycline, or mefloguineAll other areas with malaria transmission: Mosguito avoidance**Christmas Island** (Australia)NoneNot applicableNot applicableNot applicable**Colombia**All rural areas at altitudes \< 1,800 m (\<5,906 ft). None in Bogota and CartagenaChloroguine*P. falciparum* 50% *P. vivax* 50%Atovaguone/proguanil, doxycycline, or mefloguine**Comoros**AllChloroguine*P. falciparum* primarilyAtovaguone/proguanil, doxycycline, or mefloguine**Congo, Republic of the (Congo-Brazzavi1le)**AllChloroguine*P. falciparum* primarilyAtovaguone/proguanil, doxycycline, or mefloguine**Cook Islands** (New Zealand)NoneNot applicableNot applicableNot applicable**Costa Rica**Limon province, but not in Limon city (Puerto Limon). Rare cases in Puntarenas, Alajuela, Guanacaste, and Heredia provincesNone*P. vivax* 90%\ *P. falciparum* 10%Limon province: Atovaguone/proguanil, chloroguine, doxycycline, mefloguine, or primaquine[5](#cetablefn84){ref-type="table-fn"}All other areas with malaria transmission: Mosguito avoidance**Côte d\'Ivoire (Ivory Coast)**AllChloroguine*P. falciparum* 85% *P. ovale 5%-10%*\ *P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloguine**Croatia**NoneNot applicableNot applicableNot applicable**Cuba**NoneNot applicableNot applicableNot applicable**Cyprus**NoneNot applicableNot applicableNot applicable**Czech Republic**NoneNot applicableNot applicableNot applicable**Democratic Republic of the Congo (Congo-Kinshasa)**AllChloroguine*P. falciparum* 90%\ *P. ovale* 5% *P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloguine**Denmark**NoneNot applicableNot applicableNot applicable**Djibouti**AllChloroguine*P. falciparum* 90%\ *P. vivax* 5%-10%Atovaguone/proguanil, doxycycline, or mefloguine**Dominica**NoneNot applicableNot applicableNot applicable**Dominican Republic**All areas (including resort areas), except not present in the cities of Santo Domingo and SantiagoNone*P. falciparum* 100%Atovaguone/proguanil, chloroguine, doxycycline, or mefloguine**Easter Island** (Chile)NoneNot applicableNot applicableNot applicable**Ecuador,** including the **Galapagos Islands**All areas at altitudes \<1,500m(\<4,921 ft). Not present in the cities of Guayaguil, Quito, and the Galapagos IslandsChloroguine*P. vivax* 75%\ *P. falciparum* 25%Atovaguone/proguanil, doxycycline, or mefloguine**Egypt**NoneNot applicableNot applicableNot applicable**El Salvador**Rural areas of Santa Ana, Ahuachapãn, La Paz, and La Union departmentsNone*P. vivax* 99% *P. falciparum \<* 1 %Atovaguone/proguanil, chloroguine, doxycycline, mefloguine or primaquine[5](#cetablefn84){ref-type="table-fn"}**Equatorial Guinea**AllChloroguine*P. falciparum* 85% *P. malariae, P. ovale,* and *P. vivax* 15%Atovaguone/proguanil, doxycycline, or mefloguine**Eritrea**All areas at altitudes \<2,200m(\<7,218ft). None in AsmaraChloroguine*P. falciparum* 85% *P.* vivax 10%-15% *P. ovale* rareAtovaguone/proguanil, doxycycline, or mefloguine**Estonia**NoneNot applicableNot applicableNot applicable**Ethiopia**All areas at altitudes \<2,500m (\<8,202 ft), except none in Addis AbabaChloroguine*P. falciparum* 85% *P* vivax 10%-15%\ *P. malariae* and *P. ovale* \<5%Atovaguone/proguanil, doxycycline, or mefloguine**Falkland, South Georgia & South Sandwich Islands (U.K.)**NoneNot applicableNot applicableNot applicable**Faroe Islands** (Denmark)NoneNot applicableNot applicableNot applicable**Fiji**NoneNot applicableNot applicableNot applicable**Finland**NoneNot applicableNot applicableNot applicable**France**NoneNot applicableNot applicableNot applicable**French Guiana**All areas, except none in city of Cayenne or Devil\'s Island (Ile du Diable)Chloroguine*P. falciparum* \>50% *P. vivax,* \<50%Atovaguone/proguanil, doxycycline, or mefloguine**French Polynesia,** includes the island groups of **Society Islands (Tahiti, Moorea,** and **Bora-Bora), Marquesas Islands (Hiva Oa** and **Ua Huka),** and **Austral Islands (Tubuai** and **Rurutu)**NoneNot applicableNot applicableNot applicable**Gabon**AllChloroguine*P. falciparum* 95%\ *P. malariae, P. ovale, P. vivax* 5%Atovaguone/proguanil, doxycycline, or mefloguine**Gambia, The**AllChloroguine*P. falciparum* 85% *P. malariae, P. ovale, P. vivax 15%*Atovaguone/proguanil, doxycycline, or mefloguine**Georgia**Present in the southeastern part of the country near the Azerbaijan border, mainly in the Kakheti and Kveno Kartli regions. None in TblisiNone*P. vivax 100%*Atovaguone/proguanil, chloroguine, doxycycline, mefloguine, or primaquine[5](#cetablefn84){ref-type="table-fn"}**Germany**NoneNot applicableNot applicableNot applicable**Ghana**AllChloroguine*P. falciparum* 85% *P. ovale 5%-10%*\ *P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloguine**Gibraltar (U.K.)**NoneNot applicableNot applicableNot applicable**Greece**NoneNot applicableNot applicableNot applicable**Greenland** (Denmark)NoneNot applicableNot applicableNot applicable**Grenada**NoneNot applicableNot applicableNot applicable**Guadeloupe** (France)NoneNot applicableNot applicableNot applicable**Guam (U.S.)**NoneNot applicableNot applicableNot applicable**Guatemala**Rural areas only at altitudes \< 1,500 m (\<4,921ft). None in Guatemala City, Antigua or Lake AtitlánNone*P. vivax* 97%\ *P. falciparum* 3%Atovaguone/proguanil, chloroguine, doxycycline, mefloguine, or primaquine[5](#cetablefn84){ref-type="table-fn"}**Guinea**AllChloroguine*P. falciparum* 85% *P. ovale 5%-10%*\ *P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloquine**Guinea-Bissau**AllChloroguine*P. falciparum* 85% *P. ovale 5%-10%*\ *P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloquine**Guyana**All rural areas \<900 m (\<2,953 ft)Chloroguine*P. falciparum* 60% *P. vivax 40% P. malariae \<* 1 %Atovaguone/proguanil, doxycycline, or mefloquine**Haiti**All (including Port Labadee)None*P. falciparum* 100%Atovaguone/proguanil, chloroguine, doxycycline, or mefloguine**Holy See**NoneNot applicableNot applicableNot applicable**Honduras**Present throughout the country at altitudes \<1000m(\<3,281 ft) and in Roatãn and other Bay Island. None in Tegucigalpa and San Pedro SulaNone*P. vivax* 50%-95%\ *P. falciparum* 5%-50%Atovaguone/proguanil, chloroguine, doxycycline, or mefloquine**Hong Kong SAR** (China)NoneNot applicableNot applicableNot applicable**Hungary**NoneNot applicableNot applicableNot applicable**Iceland**NoneNot applicableNot applicableNot applicable**India**All areas throughout country **except** no malaria in areas \>2,000 m (\>6,561 ft) in Himachal Pradesh, Jammu, Kashmir, and Sikkim. Present in cities of Delhi and Bombay (Mumbai)Chloroguine*P. vivax* 40%\ *P. falciparum* 20%-40%\ *P. malariae* and *P. ovale* 20%-40%Atovaguone/proguanil, doxycycline, or mefloquine**Indonesia**Present in rural areas of Sumatra, Sulawesi, Kalimantan (Borneo) and Nusa Tenggara Barat (includes the island of Lombok)All areas of eastern Indonesia (provinces of Papua Indonesia, Irian Jaya Barat, Nusa Tenggara Timur, Maluku, and Maluku Utara)None in Jakarta, resort areas of Bali and the island of Java, except for the Menoreh Hills in central Java. None in urban areas in Sumatra, Kalimantan, Nusa Tenggara Barat and SulawesiChloroguine*P. falciparum* 66%\ *P. vivax* 34%Atovaguone/proguanil, doxycycline, or mefloquine**Iran**Rural areas of Sistan-1 Baluchestan, the southern tropical part of Kerman, and Hormozgan provinceArdebil and East Azerbijan provinces north of the Zagros mountains during March through NovemberChloroguine*P. vivax* 88% *P. falciparum* 11 %Atovaguone/proguanil, doxycycline, or mefloquine**Iraq**Present in areas at altitudes \< 1,500 m (\<4,921 ft) in provinces of Duhok, Erbil, Ninawa, Sulaimaninya, and Ta\'mim. None in Baghdad, Tikrit, and RamadiNone*P. vivax 100%*Atovaguone/proguanil, chloroguine, doxycycline, mefloguine, or primaquine[5](#cetablefn84){ref-type="table-fn"}**Ireland**NoneNot applicableNot applicableNot applicable**Israel**NoneNot applicableNot applicableNot applicable**Italy**NoneNot applicableNot applicableNot applicable**Jamaica**Rare local cases in KingstonNone*P. falciparum* 100%(Mosquito avoidance only)**Japan**NoneNot applicableNot applicableNot applicable**Jordan**NoneNot applicableNot applicableNot applicable**Kazakhstan**NoneNot applicableNot applicableNot applicable**Kenya**Present in all areas (including game parks) at altitudes \<2,500 m (\<8,202 ft). None in NairobiChloroguine*P. falciparum* 85% *P. vivax5%-10% P. ovale* up to 5%Atovaguone/proguanil, doxycycline, or mefloquine**Kiribati** (formerly **Gilbert Islands),** includes **Tarawa, Tabuaeran (Fanning Island),** and **Banaba (Ocean Island)**NoneNot applicableNot applicableNot applicable**Korea, North**Present in southern provincesNonePresumed to be *P. vivax 100%*Atovaguone/proguanil, chloroguine, doxycycline, mefloguine, or primaquine[5](#cetablefn84){ref-type="table-fn"}**Korea, South**Limited to rural areas in the northern parts of Kyonggi and Kangwon provinces including the demilitarized zone (DMZ)None*P. vivax 100%*Atovaguone/proguanil, chloroguine, doxycycline, mefloguine, or primaquine[5](#cetablefn84){ref-type="table-fn"}**Kosovo**NoneNot applicableNot applicableNot applicable**Kuwait**NoneNot applicableNot applicableNot applicable**Kyrgyzstan**Frequent border crossings between neighboring countries with malaria poses a small risk of malaria transmission in the southern and western parts of the country along the borders of Tajikistan and Uzbekistan No malaria transmission reported in BishkekNone*P. vivax* 99% *P. falciparum* rare imported casesMosquito avoidance only**Laos**All, except none in the city of VientianeChloroquine (Vlefloquine (see [Map 2-9](#f12){ref-type="fig"})*P. falciparum* 95%\ *P. vivax* 4%\ *P. malariae* and *P. ovale 1%*Along the Laos-Burma border in the provinces of Bokéoand Louang Namtha and along the Laos-Thailand border in the province of Saravane and Champassack: Atovaquone/proguanil or doxycyclineAll other areas: Atovaquone/proguanil, doxycycline, or mefloquine**Latvia**NoneNot applicableNot applicableNot applicable**Lebanon**NoneNot applicableNot applicableNot applicable**Lesotho**NoneNot applicableNot applicableNot applicable**Liberia**AllChloroquine*P. falciparum* 85%\ *P. ovale* 5%-10%\ *P. vivax* rareAtovaquone/proguanil, doxycycline, or mefloquine**Libya**NoneNot applicableNot applicableNot applicable**Liechtenstein**NoneNot applicableNot applicableNot applicable**Lithuania**NoneNot applicableNot applicableNot applicable**Luxembourg**NoneNot applicableNot applicableNot applicable**Macau SAR** (China)NoneNot applicableNot applicableNot applicable**Macedonia**NoneNot applicableNot applicableNot applicable**Madagascar**AllChloroquine*P. falciparum* 85%\ *P. vivax* 5%-10%\ *P. ovale* 5%Atovaquone/proguanil, doxycycline, or mefloquine**Madeira Islands** (Portugal)NoneNot applicableNot applicableNot applicable**Malawi**AllChloroquine*P. falciparum* 90%\ *P. malariae, P. ovale,* a nd *P. vivax* 10%Atovaquone/proguanil, doxycycline, or mefloquine**Malaysia**Present in rural areas of Malaysian Borneo, and to a lesser extent in rural areas of peninsular MalaysiaChloroquine*P. falciparum* 40%\ *P. vivax* 50% *IP ovale \<1% P. knowlesi* reported to cause some human infections hereAtovaquone/proguanil, doxycycline, or mefloquine**Maldives**NoneNot applicableNot applicableNot applicable**Mali**AllChloroquine*P. falciparum* 85%\ *P. ovale* 5%-10% *P. vivax* rareAtovaquone/proguanil, doxycycline, or mefloquine**Malta**NoneNot applicableNot applicableNot applicable**Marshall Islands**NoneNot applicableNot applicableNot applicable**Martinique** (France)NoneNot applicableNot applicableNot applicable**Mauritania**Present in southern provinces. None in Dakhlet-Nouadhibou, Inchiri, Adrar and Tiris-Zemmour regionsChloroquine*P. falciparum* 85%\ *P. ovale* 5%-10% *P. vivax* rareAtovaquone/proguanil, doxycycline, or mefloquine**Mauritius**NoneNot applicableNot applicableNot applicable**Mayotte** (French territorial collectivity)AllChloroquine*P. falciparum* 40%-50%\ *P. vivax* 35%-40% *P ovale \<1%*Atovaquone/proguanil, doxycycline, or mefloquine**Mexico**Limited to areas infrequently visited by travelers, including small foci along the Guatemala and Belize borders in the states of Chiapas, Quintana Roo, and Tabasco; rural areas in the states of Nayarit, Oaxaca, and Sinaloa; and in an area between 24° N and 28° N latitude, and 106°W and 110°W longitude, which lies in parts of Sonora, Chihuahua, and Durango. No malaria along the United States-Mexico border and in the major resorts along the Pacific and Gulf coastsNone*P. vivax* 99%\ *P. falciparum* 1 %Atovaquone/proguanil, chloroquine, doxycycline, mefloquine, or primaquine[5](#cetablefn84){ref-type="table-fn"}**Micronesia, Federated States of;** includes **Yap Islands, Pohnpei, Chuuk,** and **Kosrae**NoneNot applicableNot applicableNot applicable**Moldova**NoneNot applicableNot applicableNot applicable**Monaco**NoneNot applicableNot applicableNot applicable**Mongolia**NoneNot applicableNot applicableNot applicable**Montenegro**NoneNot applicableNot applicableNot applicable**Montserrat (U.K.)**NoneNot applicableNot applicableNot applicable**Morocco**NoneNot applicableNot applicableNot applicable**Mozambique**AllChloroguine*P. falciparum* 95%\ *P. malariae* and *P. ovale* 5%\ *P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloquine**Namibia**Present in the provinces of Kunene, Ohangwena, Okavango, Omaheke, Omusati, Oshana. Oshikoto, Otjozondjupa, and the Caprivi StripChloroguine*P. falciparum* 90%\ *P. malariae, P. ovale,* and *P. vivax* 10%Atovaguone/proguanil, doxycycline, or mefloquine**Nauru**NoneNot applicableNot applicableNot applicable**Nepal**Present throughout country at altitudes \<1,200m(\<3,937ft). None in Kathmandu and on typical Himalayan treksChloroguine*P. vivax* 88% *P. falciparum* 12%Atovaguone/proguanil, doxycycline, or mefloquine**Netherlands**NoneNot applicableNot applicableNot applicable**Netherlands Antilles (Bonaire, Curaçao, Saba, St. Eustasius,** and **St. Maarten)**NoneNot applicableNot applicableNot applicable**New Caledonia** (France)NoneNot applicableNot applicableNot applicable**New Zealand**NoneNot applicableNot applicableNot applicable**Nicaragua**Present in rural areas. None in ManaguaNone*P. vivax* 95%\ *P. falciparum* 5%Atovaguone/proguanil, chloroguine, doxycycline, mefloguine, or primaquine[5](#cetablefn84){ref-type="table-fn"}**Niger**AllChloroguine*P. falciparum* 85% *P. ovale 5%-10% P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloquine**Nigeria**AllChloroguine*P. falciparum* 85% *P. ovale 5%-10% P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloquine**Niue** (New Zealand)NoneNot applicableNot applicableNot applicable**Norfolk Island** (Australia)NoneNot applicableNot applicableNot applicable**Northern Mariana Islands** (U.S.), includes **Saipan, Tinian,** and **Rota Island**NoneNot applicableNot applicableNot applicable**Norway**NoneNot applicableNot applicableNot applicable**Oman**NoneNot applicableNot applicableNot applicable**Pakistan**All areas (including all cities) at altitudes \<2,500 m (\<8202 ft)Chloroguine*P. falciparum* 70%\ *P. vivax* 30%Atovaguone/proguanil, doxycycline, or mefloquine**Palau**NoneNot applicableNot applicableNot applicable**Panama**Present in rural areas of the provinces of Bocas Del Toro, Darién, Veragaus, San Blas and San Blas Islands. None in Panama City or in the former Canal ZoneChloroguine*P. vivax* 90%-95%\ *P. falciparum* 5%-10%Bocas Del Toro: Atovaguone/proguanil, chloroguine, doxycycline, mefloguine, or primaquine[5](#cetablefn84){ref-type="table-fn"}Darién, San Bias, and Veragaus provinces: Atovaguone/proguanil, doxycycline, mefloguine, or primaquine[5](#cetablefn84){ref-type="table-fn"}**Papua New Guinea**Present throughout at altitudes \< 1,800 m (\<5,906 ft)Chloroguine (both *P. falciparum* and *P. vivax)P. falciparum* 65%-80% *P. vivax 10%-30% P. malariae* and *P. ovale* rareAtovaguone/proguanil, doxycycline, or mefloquine**Paraguay**Present in the departments of Alto Parana, Caaguazu, and CanendiyuNone*P. vivax* 95%\ *P. falciparum* 5%Atovaguone/proguanil, chloroguine, doxycycline, mefloguine, or primaquine[5](#cetablefn84){ref-type="table-fn"}**Peru**All departments \<2000 m (6,561 ft) except none in Areguipa, (Vloguegua), Puno, and Tacna. Present in Puerto (Vlaldonado)Chloroguine*P. vivax* 70%\ *P. falciparum* 30%\ *P. malariae \<* 1 %Lima, coastal areas south of Lima, or the highland tourist areas (Cuzco, (Vlachu Picchu, and LakeTiticaca): (Mosquito avoidance only)Other areas: Atovaguone/proguanil, doxycycline, or mefloquine**Philippines**Present in rural areas \<600 m (1,969 ft), on islands of Luzon, Palawan, and Mindanao. None in urban areasChloroguine*P. falciparum* 70%-80%\ *P. vivax* 20%-30%Atovaguone/proguanil, doxycycline, or mefloquine**Pitcairn Islands (U.K.)**NoneNot applicableNot applicableNot applicable**Poland**NoneNot applicableNot applicableNot applicable**Portugal**NoneNot applicableNot applicableNot applicable**Puerto Rico (U.S.)**NoneNot applicableNot applicableNot applicable**Qatar**NoneNot applicableNot applicableNot applicable**Réunion** (France)NoneNot applicableNot applicableNot applicable**Romania**NoneNot applicableNot applicableNot applicable**Russia**Rare local cases by border with AzerbaijanNone*P. vivax 100%*By border with Azerbaijan: Mosguito avoidance only**Rwanda**AllChloroguine*P. falciparum* \>85%\ *P. vivax* 5%\ *P. ovale* 5%Atovaguone/proguanil, doxycycline, or mefloquine**Saint Barthélémy** (France)NoneNot applicableNot applicableNot applicable**Saint Helena (U.K.)**NoneNot applicableNot applicableNot applicable**Saint Kitts (Saint Christopher) and Nevis (U.K.)**NoneNot applicableNot applicableNot applicable**Saint Lucia**NoneNot applicableNot applicableNot applicable**Saint Martin** (France)NoneNot applicableNot applicableNot applicable**Saint Pierre** and **Miquelon** (France)NoneNot applicableNot applicableNot applicable**Saint Vincent and the Grenadines**NoneNot applicableNot applicableNot applicable**Samoa** (formerly **Western Samoa)**NoneNot applicableNot applicableNot applicable**Samoa, American (U.S.)**NoneNot applicableNot applicableNot applicable**San Marino**NoneNot applicableNot applicableNot applicable**São Tomé and Principé**AllChloroguine*P. falciparum* 85%\ *P. malariae, P. ovale* 5%\ *P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloquine**Saudi Arabia**Provinces of AI (Vladinah, Asir (excluding high-altitude areas \>2,000 m), Jazan, and Mecca. None in cities of Jeddah, Mecca, Medina, Riyadh, and Ta\'ifChloroguine*P. falciparum* predominantly *P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloquine**Senegal**AllChloroguine*P. falciparum* \>85% *P. ovale 5%-10% P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloquine**Serbia**NoneNot applicableNot applicableNot applicable**Seychelles**NoneNot applicableNot applicableNot applicable**Sierra Leone**AllChloroguine*P. falciparum* 85%\ *P. malariae, P. ovale,* and *P. vivax* 5%Atovaguone/proguanil, doxycycline, or mefloquine**Singapore**NoneNot applicableNot applicableNot applicable**Slovakia**NoneNot applicableNot applicableNot applicable**Slovenia**NoneNot applicableNot applicableNot applicable**Solomon Islands**AllChloroguine*P. falciparum* 60%\ *P. vivax* 35%-40% *P. ovale \<1%*Atovaguone/proguanil, doxycycline, or mefloquine**Somalia**AllChloroguine*P. falciparum* 95% *P. vivax, P. malariae,* and *P. ovale* 5%Atovaguone/proguanil, doxycycline, or mefloquine**South Africa**Present in the (Vlpumalanga Province, Limpopo (Northern) Province, and northeastern KwaZulu-Natal as far south as the Tugela River. Present in Kruger National ParkChloroguine*P. falciparum* 90%\ *P. vivax* 5%\ *P. ovale* 5%Atovaguone/proguanil, doxycycline, or mefloquine**Spain**NoneNot applicableNot applicableNot applicable**Sri Lanka**All areas, except none in the districts of Colombo, Galle, Gampaha, Kalutara, (Vlatara, and Nuvvara Eliya)Chloroguine*P. vivax* 88%\ *P. falciparum* 12%Atovaguone/proguanil, doxycycline, or mefloquine**Sudan**AllChloroguine*P. falciparum* 90%\ *P. malariae, P. vivax,* and *P. ovale* 10%Atovaguone/proguanil, doxycycline, or mefloquine**Suriname**All areas, except none in ParamariboChloroguine*P. falciparum* 70% *P. vivax 15%-20%*Atovaguone/proguanil, doxycycline, or mefloquine**Swaziland**Present in the northern and eastern areas bordering (Vlozambigue and Zimbabwe, including all of Lubombo district)Chloroguine*P. falciparum* 90%\ *P. vivax* 5%\ *P. ovale* 5%Atovaguone/proguanil, doxycycline, or mefloquine**Sweden**NoneNot applicableNot applicableNot applicable**Switzerland**NoneNot applicableNot applicableNot applicable**Syria**Rare cases in the northern border in El Hassaka provinceNone*P. vivax* predominantly(Mosquito avoidance only)**Taiwan**NoneNot applicableNot applicableNot applicable**Tajikistan**All areas \<2,000 m (6562 ft)Chloroguine*P. vivax* 90%\ *P. falciparum* 10%Atovaguone/proguanil, doxycycline, mefloguine, or primaguine[5](#cetablefn84){ref-type="table-fn"}**Tanzania**All areas at altitudes \<1,800m(\<5,906ft)Chloroguine*P. falciparum* \>85%\ *P. malariae,* and *P. ovale \>* 10%\ *P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloquine**Thailand**Rural, forested areas that border Cambodia, Laos, and Myanmar (Burma). Rare local cases in Phang Nga and Phuket. None in cities and in major tourist resorts. None in cities of Bangkok, Chiang Mai, Chiang Rai, Pattaya, Koh Samui, and Koh PhanganChloroguine Mefloguine (see [Map 2-9](#f12){ref-type="fig"})*P. falciparum* 50% (up to 75% some areas)\ *P. vivax* 50% (up to 60% some areas)\ *P. ovale,* rarePhang Nga and Phuket: Mosguito avoidance onlyAll other areas: Atovaguone/proguanil or doxycycline**Timor-Leste (East Timor)**AllChloroguine*P. falciparum* 50%\ *P. vivax* 50% *P. ovale \<1% P. malariae \<* 1 %Atovaguone/proguanil, doxycycline, or mefloquine**Togo**AllChloroguine*P. falciparum* 85% *P. ovale 5%-10% P. vivax* rareAtovaguone/proguanil, doxycycline, or mefloquine**Tokelau** (NewZealand)NoneNot applicableNot applicableNot applicable**Tonga**NoneNot applicableNot applicableNot applicable**Trinidad and Tobago**NoneNot applicableNot applicableNot applicable**Tunisia**NoneNot applicableNot applicableNot applicable**Turkey**Present by border with Syria. None on the Incerlik U.S. Air Force base and on typical cruise itinerariesNoneP. v/Vaxand *P. falciparum* presentAtovaguone/proguanil, chloroguine, doxycycline, or mefloquine**Turkmenistan**Rare local cases by Afghanistan borderNone*P. vivax 100%*Mosguito avoidance only**Turks and Caicos Islands (U.K.)**NoneNot applicableNot applicableNot applicable**Tuvalu**NoneNot applicableNot applicableNot applicable**Uganda**AllChloroguine*P. falciparum* \>85%\ *P. malariae, P. ovale,* and *P. vivax* \<15%Atovaguone/proguanil, doxycycline, or mefloquine**Ukraine**NoneNot applicableNot applicableNot applicable**United Arab Emirates**NoneNot applicableNot applicableNot applicable**United Kingdom** (with **Channel Islands** and **Isle of Man)**NoneNot applicableNot applicableNot applicable**United States**NoneNot applicableNot applicableNot applicable**Uruguay**NoneNot applicableNot applicableNot applicable**Uzbekistan**Rare cases along the Afghanistan and Tajikistan borderNoneP *vivax 100%*Mosguito avoidance only**Vanuatu**AllChloroguine*P. falciparum* 60%\ *P. vivax* 35%-40% *P. ovale \<1%*Atovaguone/proguanil, doxycycline, or mefloquine**Venezuela**Rural areas of the following states: Apure, Amazonas, Barinas, Bolivar, Sucre, Tachira, and Delta Amacuro. Present in Angel Falls. None in Margarita IslandChloroguine*P. vivax* 80%-90%\ *P. falciparum* 10%-20%Atovaguone/proguanil, doxycycline, or mefloquine**Vietnam**Rural, forested areas, **except** none in the Red River delta and the coast north of Nha Trang. None in Can Tho, Da Nang, Haiphong, Hanoi, Ho Chi Minh City (Saigon), Hue, Nha Trang, and Qui NhonChloroguine (Vlefloguine (see [Map 2-9](#f12){ref-type="fig"})*P. falciparum* 50%-80%\ *P. vivax* 20%-50%Southern part of the country in the provinces of Dae Lac, Gia Lai, Khan h Hoa, Kon Turn, Lam Dong, Ninh Thuan, Song Be, Tay Ninh: Atovaguone/proguanil or doxycyclineAll other areas: Atovaguone/proguanil, doxycycline, or mefloquine**Virgin Islands, British**NoneNot applicableNot applicableNot applicable**Virgin Islands, U.S.**NoneNot applicableNot applicableNot applicable**Western Sahara**Rare casesNoneUnknownMosguito avoidance only**Yemen**All areas at altitudes \<2,000 m (\<6,561 ft). None in Sana\'aChloroguine*P. falciparum* 95%\ *P. malariae, P. vivax,* and *P. ovale 5%*Atovaguone/proguanil, doxycycline, or mefloquine**Zambia**AllChloroguine*P. falciparum* \>90% *P. vivax* up to 5%\ *P. ovale* up to 5 %Atovaguone/proguanil, doxycycline, or mefloquine**Zimbabwe**AllChloroguine*P. falciparum* \>90%\ *P. vivax* up to 5%\ *P. ovale* up to 5 %Atovaguone/proguanil, doxycycline, or mefloquine[^80][^81][^82][^83][^84]   {#cesec309} - Self-Treatable Diseases SELF-TREATABLE DISEASES {#cesec310} ----------------------- **Alan J. Magill** Despite our best efforts at helping to prevent illness, travelers will often become ill while traveling. Obtaining reliable and timely medical care can be problematic in many destinations. As a result, prescribing certain medications in advance can empower the traveler to self-diagnose and treat common health problems. During an activity in a remote setting, such as trekking, the only alternative to self-treatment would be no treatment. In some developing countries, appropriate pre-travel counseling may result in a more accurate diagnosis and treatment than relying on local medical care. In addition, the increasing awareness of counterfeit drugs in pharmacies in the developing world (as many as 20%--30% of the drugs on the shelves) makes it more important for travelers to carry reliable drugs from their own country. Providing education and prescriptions is part of the pre-travel consultation. The key aspect to this strategy is to recognize which travelers may be at risk and to educate them as to the diagnosis and treatment of the particular illness. The keys to successful self-treatment strategies are providing a simple disease definition, providing one choice of treatment, and educating the traveler about the expected outcome of treatment. Using travelers\' diarrhea as an example, one could provide the following advice:•Travelers\' diarrhea is defined as "the sudden onset of relatively uncomfortable diarrhea."•The treatment is ciprofloxacin 500 mg every 12 hours for 1 day (two doses).•The traveler should feel better within 6 to 24 hours. To minimize the potential negative effects of a self-treatment strategy, the recommendations should follow a few key points:•Drugs used must be safe, well tolerated, and effective for use as self-treatment.•A drug\'s toxicity or potential for harm, if used incorrectly or in an overdose situation, should be minimal.•Good directions are critical. Consider providing simple but clear handouts describing how to use the drugs. Keeping the directions simple will greatly increase the effectiveness of the strategy. Following are some of the most common situations in which people would find self-treatment useful. The extent of self-treatment recommendations offered to the traveler should reflect the remoteness and difficulty of travel and the availability of reliable medical care at the particular destination. The recommended self-treatment options for each disease are provided in the designated section of the Yellow Book. **Travelers\' diarrhea (TD)** is perhaps most frequent indication for self-treatment. The success of this strategy is based on the epidemiologic evidence that bacterial pathogens account for more than 90% of TD in short-term travelers. The recognition of antibiotic resistance for certain organisms in specific destinations has made the empiric choice of treatment somewhat more problematic in recent times (see the [Travelers\' Diarrhea section](#cesec311){ref-type="sec"} next in this chapter). **Altitude illness** or acute mountain sickness (AMS) is a risk for travelers who ascend rapidly to altitudes \>8,000 ft (2,440 m). Certain common travel destinations, such as Cuzco, Peru, or Lhasa, Tibet, involve flying to altitudes of 11,300 ft (3,445 m) or 12,700 ft (3,870 m). The symptoms of headache, anorexia, nausea, fatigue, lassitude, and poor sleep can largely be prevented or treated with acetazolamide (see the [Altitude Illness](#subchapter24){ref-type="sec"} section later in this chapter). **Jet lag** affects almost everyone who crosses three or more time zones. There is no consensus on the optimal pharmacologic treatment or prevention of the symptoms of jet lag, but sleeping medication taken at the destination may help regularize sleep patterns (see the [Jet Lag](#subchapter25){ref-type="sec"} section later in this chapter). **Motion sickness** can be a major deterent to enjoyment for any susceptible person on a boat or a winding road. Premedication may help alleviate or ameliorate this bothersome syndrome (see the [Motion Sickness](#subchapter26){ref-type="sec"} section later in this chapter). The self-treatment of suspected **respiratory infections** with empiric antibiotics is controversial. Almost all upper respiratory tract infections are initially caused by viruses. However, these viral infections, under the stress of travel, can lead to bacterial sinusitis, bronchitis or pneumonia. Respiratory infections that last longer than a week without signs of improvement may require empiric antibiotics for recovery. Prolonged respiratory infections may have more of a negative impact on a trip than diarrheal disease (see the [Respiratory Infections](#subchapter27){ref-type="sec"} section later in this chapter). **Bacterial skin infections** are not common among travelers, but when they occur, they can be particularly distressing. Bacterial abscesses or cellulitis can worsen rapidly and be very painful. If the traveler is in a remote area or even more than a day\'s travel from medical care, the use of empiric antibiotic treatment can be extremely beneficial (see the Skin and Soft Tissue Infections in Returned Travelers section in Chapter 4). **Urinary tract infections** are common among many women, and carrying an antibiotic for empiric treatment of this condition may be valuable in many circumstances. **Vaginal yeast infections** in women can be a very annoying and debilitating problem. For women who know they are prone to infections, all sexually active women, and those who may be receiving antibiotics for other reasons, including doxycycline for antimalarial prophylaxis, a self-treatment course of their preferred antifungal medication can be prescribed. **Occupational/HIV needlestick** is a particular risk to those participating in medical-related activities. Every year thousands of such individuals are now working in areas of sub-Saharan Africa, where the HIV prevalence may be higher than 15%--20%. A significant needlestick in this setting should prompt immediate wound care and the possible use of antiretroviral medications (see the [Occupational Exposure to HIV](#subchapter28){ref-type="sec"} section later in this chapter). **Malaria self-treatment** is often referred to as stand-by emergency treatment (SBET). This strategy asks the traveler to use a therapeutic dose of an appropriate antimalarial drug when the traveler has a significant fever accompanied by systemic illness, and then proceed to reliable medical care within 24 hours. The goal is to prevent death or severe malaria. Since most travelers at risk of malaria should be advised to use prophylactic medication, this strategy is usually discouraged and reserved for a specific type of traveler under certain defined circumstances (see the [Malaria](#subchapter21){ref-type="sec"} section earlier in this chapter). TRAVELERS\' DIARRHEA {#cesec311} -------------------- **Bradley A. Connor** ### Description {#cesec312} Travelers\' diarrhea (TD) is the most predictable travel-related illness. Attack rates range from 30% to 70% of travelers, depending on the destination. Traditionally, it was thought that TD could be prevented by following eating rules, but studies have found that people who follow the rules still get ill. Poor hygiene practice in local restaurants is likely the largest contributor to the risk for TD. TD itself is a clinical syndrome that can result from a variety of intestinal pathogens. Bacterial pathogens are the predominant risk, thought to account for 80%--90% of TD. Intestinal viruses have been isolated in studies of TD, but they usually account for 5%--8% of illnesses. Protozoal pathogens are slower to manifest symptoms, and collectively account for about 10% of diagnoses in longer-term travelers. What is commonly known as "food poisoning" involves the ingestion of preformed toxins in food. In this syndrome, vomiting and diarrhea may both be present, but symptoms usually resolve spontaneously within 12 hours. ### Infectious Agent {#cesec313} •Bacteria are the most common cause of TD. The most common pathogen is enterotoxigenic *Escherichia coli*, followed by *Campylobacter jejuni*, *Shigella* sp., and *Salmonella* sp. Enteroadherent and other *E. coli* species have been found to also be common pathogens in bacterial diarrhea.•Viral diarrhea can be caused by a number of viral pathogens, including norovirus, rotavirus, and astrovirus.•*Giardia* is the main protozoal pathogen found in travelers. *Entamoeba histolytica* is a relatively uncommon pathogen in travelers. *Cryptosporidium* is also relatively uncommon. The risk for *Cyclospora* is highly geographic and seasonal, with the most well-known risks in Nepal, Peru, Haiti, and Guatemala. *Dientamoeba fragilis* is a low-grade but persisent pathogen that is occasionally diagnosed in travelers.•The individual pathogens are each discussed in their own sections in Chapter 5, and persistent diarrhea is discussed in Chapter 4. ### Occurrence {#cesec314} •The most important determinant of risk is travel destination, and there are regional differences in both the risk for and etiology of diarrhea.•The world is generally divided into three grades of risk: low, intermediate, and high.○Low-risk countries include the United States, Canada, Australia, New Zealand, Japan, and countries in Northern and Western Europe.○Intermediate-risk countries include those in Eastern Europe, South Africa, and some of the Caribbean islands.○High-risk areas include most of Asia, the Middle East, Africa, Mexico, and Central and South America. ### Risk for Travelers {#cesec315} Travelers\' diarrhea occurs equally in male and female travelers and is more common in young adults than in older people. In short-term travelers, bouts of TD do not appear to protect against future attacks, and more than one episode of TD may occur during a single trip. A cohort of expatriates taking up residence in Kathmandu, Nepal, experienced an average of 3.2 episodes of TD per person in their first year. In more temperate regions, there may be seasonal variations in diarrhea risk. In South Asia, for example, during the hot months preceding the monsoon, much higher TD attack rates are commonly reported. In environments where large numbers of people do not have access to plumbing or outhouses, the amount of stool contamination in the environment will be higher and more accessible to flies. Inadequate electrical capacity may lead to frequent blackouts or poorly functioning refrigeration, which can result in unsafe food storage and an increased risk for disease. Inadequate water supplies can lead to the absence of sinks for handwashing by restaurant staff. Poor training in handling and preparation of food may lead to cross-contamination from meat and inadequate sterilization of food preparation surfaces and utensils. In destinations in which effective food handling courses have been provided, the risk for TD has been demonstrated to decrease. It should be noted, however, that pathogens that cause TD are not unique to developing countries. The risk of TD is associated with the hygiene practices in specific destinations and the handling and preparation of food in restaurants in developed countries as well. ### Clinical Presentation {#cesec316} •Bacterial diarrhea presents with the sudden onset of bothersome symptoms that can range from mild cramps and urgent loose stools, to severe abdominal pain, fever, vomiting, and bloody diarrhea.•Viral enteropathogens present in a similar fashion to bacterial pathogens, although with norovirus vomiting may be more prominent.•Protozoal diarrhea, such as that caused by *Giardia intestinalis*, or *Entamoeba histolytica*, generally has a more gradual onset of low-grade symptoms, with 2--5 loose stools per day.•The incubation period of the pathogens can be a clue to the etiology of TD.○Bacterial and viral pathogens have an incubation period of 6--48 hours.○Protozoal pathogens generally have an incubation period of 1--2 weeks and rarely present in the first few weeks of travel. An exception can be *Cyclospora cayetanensis*, which can present quickly in areas of high risk.•Untreated bacterial diarrhea lasts 3--5 days. Viral diarrhea lasts 2--3 days. Protozoal diarrhea can persist for weeks to months without treatment.•An acute bout of gastroenteritis can lead to persistent gastrointestinal symptoms, even in the absence of continued infection (see the Persistent Travelers\' Diarrhea section in Chapter 4). Other postinfectious sequelae include reactive arthritis and Guillain--Barré syndrome. ### Preventive Measures for Travelers {#cesec317} •For travelers to high-risk areas, several approaches may be recommended that can reduce but never completely eliminate the risk for TD. These include---○Instruction regarding food and beverage selection○Use of agents other than antimicrobial drugs for prophylaxis○Use of prophylactic antibiotics•Carrying small containers of hand-sanitizing solutions or gels (containing at least 60% alcohol) may make it easier for travelers to clean their hands before eating. #### Food and Beverage Selection {#cesec318} Care in selecting food and beverages for consumption might minimize the risk for acquiring TD. Travelers should be advised that foods that are freshly cooked and served piping hot are safer than foods that may have been sitting for some time in the kitchen or in a buffet. Care should be taken to avoid beverages diluted with nonpotable water (reconstituted fruit juices, ice, and milk) and foods washed in nonpotable water, such as salads. Other risky foods include raw or undercooked meat and seafood, and unpeeled raw fruits and vegetables. Safe beverages include those that are bottled and sealed, or carbonated. Boiled beverages and those appropriately treated with iodine or chlorine may also be safely consumed. Although food and water precautions continue to be recommended, travelers may not always be able to always adhere to the advice. Furthermore, many of the factors that ensure food safety, such as restaurant hygiene, are out of the traveler\'s control. #### Nonantimicrobial Drugs for Prophylaxis {#cesec319} The primary agent studied for prevention of TD, other than antimicrobial drugs, is bismuth subsalicylate (BSS), which is the active ingredient in Pepto-Bismol. Studies from Mexico have shown this agent (taken daily as either 2 oz of liquid or two chewable tablets four times per day) reduces the incidence of TD from 40% to 14%. BSS commonly causes blackening of the tongue and stool and may cause nausea, constipation, and rarely tinnitus. BSS should be avoided by travelers with aspirin allergy, renal insufficiency, and gout, and by those taking anticoagulants, probenecid, or methotrexate. In travelers taking aspirin or salicylates for other reasons, the use of BSS may result in salicylate toxicity. Caution should be used in administering BSS to children with viral infections, such as varicella or influenza, because of the risk for Reye syndrome. BSS is not recommended for children \<3 years of age. Studies have not established the safety of BSS use for periods \>3 weeks. The use of probiotics, such as Lactobacillus GG and *Saccharomyces boulardii*, has been studied in the prevention of TD in limited numbers of subjects. Results are inconclusive, partially because standardized preparations of these bacteria are not reliably available. #### Prophylactic Antibiotics {#cesec320} Prophylactic antibiotics have been demonstrated to be quite effective in the prevention of TD. Controlled studies have shown that diarrhea attack rates are reduced from 40% to 4% by the use of antibiotics. The prophylactic antibiotic of choice has changed over the past few decades as resistance patterns have evolved. Agents such as trimethoprim-sulfamethoxazole and doxycycline are no longer considered effective antimicrobial agents against enteric bacterial pathogens. The fluoroquinolones have been the most effective antibiotics for the prophylaxis and treatment of bacterial TD pathogens, but increasing resistance to these agents, mainly among *Campylobacter* species, may limit their benefit in the future. A nonabsorbable antibiotic, rifaximin, is being investigated for its potential use in TD prophylaxis. In the only study published to date, rifaximin reduced the risk for TD in travelers to Mexico by 77%. At this time, prophylactic antibiotics should not be recommended for most travelers. In addition to affording no protection against nonbacterial pathogens, the use of antibiotics may be associated with allergic or adverse reactions in a certain percentage of travelers. The use of prophylactic antibiotics should be weighed against the result of using prompt, early self-treatment with antibiotics when TD occurs, which can limit the duration of illness to 6--24 hours in most cases. Prophylactic antibiotics may be considered for short-term travelers who are high-risk hosts (such as those who are immunosuppressed) or are taking critical trips during which even a short bout of diarrhea could impact the purpose of the trip. ### Treatment {#cesec321} Antibiotics are the principal element in the treatment of TD. Adjunctive agents used for symptomatic control may also be recommended. #### Antibiotics {#cesec322} As bacterial causes of TD far outnumber other microbial etiologies, empiric treatment with an antibiotic directed at enteric bacterial pathogens remains the best therapy for TD. The benefit of treatment of TD with antibiotics has been proven in numerous studies. The effectiveness of a particular antimicrobial depends on the etiologic agent and its antibiotic sensitivity. Both as empiric therapy or for treatment of a specific bacterial pathogen, first-line antibiotics include those of the fluoroquinolone class, such as ciprofloxacin or levofloxacin. Increasing microbial resistance to the fluoroquinolones, especially among *Campylobacter* isolates, may limit their usefulness in some destinations such as Thailand, where *Campylobacter* is prevalent. Isolated anecdotal case reports of resistant *Campylobacter* diarrhea occur periodically from other destinations. An alternative to the fluoroquinolones in this situation is azithromycin. Rifaximin has been approved for the treatment of TD caused by noninvasive strains of *E. coli*. However, since it is often difficult for travelers to distinguish between invasive and noninvasive diarrhea and since they would have to carry a back-up drug in the event of invasive diarrhea, the overall usefulness of rifaximin as empiric self-treatment remains to be determined. Single-dose or 1-day therapy for TD with a fluoroquinolone is well established, both by clinical trials and clinical experience. The best regimen for azithromycin treatment is not yet established. One study used a single dose of 1,000 mg, but side effects (mainly nausea) may limit the acceptability of this large dose. Azithromycin, 500 mg per day for 1--2 days, appears to be effective in most cases of TD. #### Antimotility Agents {#cesec323} Antimotility agents provide symptomatic relief and serve as useful adjuncts to antibiotic therapy in TD. Synthetic opiates, such as loperamide and diphenoxylate, can reduce bowel movement frequency and enable travelers to ride on an airplane or bus while awaiting the effects of antibiotics. Loperamide appears to have antisecretory properties as well. The safety of loperamide when used along with an appropriate antibiotic has been well established, even in cases of invasive pathogens. Loperamide can be used in children, and liquid formulations are available. In practice, however, these drugs are rarely given to small children. #### Oral Rehydration Therapy {#cesec324} Fluids and electrolytes are lost in cases of TD, and replenishment is important, especially in young children or adults with chronic medical illness. In adult travelers who are otherwise healthy, severe dehydration resulting from TD is unusual unless prolonged vomiting is present. Nonetheless, replacement of fluid losses remains an important adjunct to other therapy and helps the traveler feel better more quickly. Travelers should remember to use only beverages that are sealed or carbonated, or otherwise known to be purified. For more severe fluid loss, replacement is best accomplished with oral rehydration solutions (ORS), such as the WHO ORS solutions, which are widely available at stores and pharmacies in most developing countries (see [Table 2-25](#cetable24){ref-type="table"} for details). ORS is prepared by adding one packet to the appropriate volume of boiled or treated water. Travelers may find most ORS formulations to be relatively unpalatable, due to their saltiness. In most cases, rehydration can be maintained with any palatable liquid.Table 2-25Composition of WHO oral rehydration solution (ORS) for diarrheal illnessIngredientAmountMeasurementSodium chloride3.5 g/L½ tspPotassium chloride1.5 g/L1¼ tspGlucose20.0 g/L2 tbspTrisodium citrate (or sodium bicarbonate)2.9 g/L (or 2.5 g/L)½ tspWater1,000 g1 liter #### Treatment of TD Caused by Protozoa {#cesec325} The most common parasitic cause of TD is *Giardia intestinalis*, and treatment options include metronidazole, tinidazole, and nitazoxanide. Although cryptosporidiosis is usually a self-limited illness in immunocompetent persons, nitazoxanide can be considered as a treatment option. Cyclosporiasis is treated with trimethoprim--sulfamethoxazole. Treatment of amebiasis is with metronidazole or tinidazole, followed by treatment with a luminal agent such as paromomycin. #### Treatment for Children {#cesec326} Children who accompany their parents on trips to high-risk destinations may be expected to have TD as well. There is no reason to withhold antibiotics from children who contract TD. In older children and teenagers, treatment recommendations for TD follow those for adults, with possible adjustments in the dose of medication. Macrolides such as azithromycin are considered first-line antibiotic therapy in children, although some experts now use short-course fluoroquinolone therapy for travelers \<18 years of age. Rifaximin is approved for use starting at 12 years of age. Infants and younger children are at higher risk for developing dehydration from TD, which is best prevented by the early use of ORS solutions. Breastfed infants should continue to nurse on demand, and bottle-fed infants can continue to drink their formula. Older infants and children may eat a regular diet, depending on the level of their appetite while they are ill. Infants in diapers are at risk for developing a painful, ecxematous rash on their buttocks in response to the liquid stool. Hydrocortisone cream will quickly improve this rash. More information about diarrhea and dehydration are discussed in the Traveling Safely with Infants and Children section in Chapter 7. *PERSPECTIVES:* GLOBAL IMPACT OF DIARRHEAL DISEASE {#subchapter23} ================================================== Pawlowski Sean W. Guerrant Richard L. Travelers frequently acquire the enteric pathogens that are present in food and water at their destinations and develop travelers\' diarrhea. Mercifully, this is usually a nonfatal nuisance for travelers from developed nations, who, due to their well-nourished state, lack of other co-morbidities, antibiotic availability, and lack of repetitive infections, recover from their illness (more often than not) with few lasting effects. In the short term, for those living in developing countries, infections due to enteric organisms are potentially life threatening, particularly in children and when combined with other illnesses, such as measles. Oral rehydration is the mainstay of treatment, but children suffering repeated, malnourishing illnesses at weaning require critical nutrient and micronutrient therapy as well. While those traveling to aid in humanitarian efforts, including missionaries and volunteers, may feel they are well prepared to meet the challenges they face, many may not fully understand the profound impact that the lack of availability of clean water and sanitation has on indigenous populations. Travel health advisors, as well, should understand the serious implications and consequences of the global burden of these diarrheal illnesses as their occurrence is centered in the destinations of many U.S. travelers. Recent figures estimate 1.6 million deaths per year worldwide in children \<5 years of age in these areas. Fortunately, this rate is dramatically down from 4.6 million per year (estimates from 1955 to 1979), in large part due to the implementation of oral rehydration therapy. However, the morbidity rates either have not fallen or have slightly increased. In addition, as the AIDS pandemic continues to spread, complications associated with endemic enteric pathogens will likely increase both mortality and morbidity. More difficult to quantify is the global burden of repeated or persistent diarrheal illnesses. Unfortunately, these contribute significantly to the impairment of physical and cognitive development of children and to long-term disability, which ultimately result in substantial national economic losses. ALTITUDE ILLNESS {#subchapter24} ================ Hackett Peter H. Shlim David R. Occurrence {#cesec327} ---------- The stresses of the high-altitude environment include cold, low humidity, increased ultraviolet (UV) radiation, and decreased air pressure, all of which can cause problems for travelers. The greatest concern, however, is hypoxia. At 10,000 ft (3,000 m), for example, the inspired PO~2~ is only 69% of sea-level value. The degree of hypoxic stress depends upon altitude, rate of ascent, and duration of exposure. Sleeping at high altitude produces the greatest hypoxia; day trips to high altitude with return to low altitude are much less stressful on the body. Acclimatization {#cesec328} --------------- The human body adjusts very well to moderate hypoxia, but requires time to do so ([Box 2-3](#cetextbox1){ref-type="boxed-text"} ). The process of acute acclimatization to high altitude takes 3--5 days; therefore, acclimatizing for a few days at 8,000--9,000 ft before proceeding to higher altitude is ideal. Acclimatization prevents altitude illness, improves sleep, and increases comfort and well-being, although exercise performance will always be reduced compared with low altitude. Increase in ventilation is the most important factor in acute acclimatization; therefore, respiratory depressants must be avoided. Increased red-cell production does not play a role in acute acclimatization.Box 2-3Tips for acclimatizationThe following are helpful tips for people traveling to high altitude destinations.•Ascend gradually, if possible. Try not to go directly from low altitude to \>9,000 ft (2,750 m) sleeping altitude in one day.•Consider using acetazolamide (Diamox) to speed acclimatization if abrupt ascent is unavoidable.•Avoid alcohol for the first 48 hours.•Participate in only mild exercise for the first 48 hours.•Having a high-altitude exposure at \>9,000 ft (2,750 m), for 2 nights or more within 30 days prior to the trip is useful.•Treat an altitude headache with simple analgesics. Risk for Travelers {#cesec329} ------------------ Inadequate acclimatization may lead to altitude illness in any traveler going to 8,000 ft (2,500 m) or higher. Susceptibility and resistance to altitude illness are genetic traits, and no screening tests are available to predict risk. Risk is not affected by training or physical fitness. Children are equally susceptible as adults; persons \>50 years of age have slightly lower risk. How a traveler has responded to high altitude previously is the most reliable guide for future trips but is not infallible. However, given certain baseline susceptibility, risk is greatly influenced by rate of ascent and exertion. Determining an itinerary that will avoid any occurrence of altitude illness is difficult because of variations in individual susceptibility, as well as in starting points and terrain. Itineraries with a high risk for altitude illness include flying directly to \>9,000 ft or rapid hiking ascents, such as climbing Mt. Kilimanjaro. It is best to average no more than 1,000 ft (300 m) ft per day in altitude gain above 12,000 ft (3,660 m). Examples of high-altitude cities with airports are Cuzco, Peru (11,000 ft; 3,326 m); La Paz, Bolivia (12,000 ft; 3,660 m); and Lhasa, Tibet (12,500 ft; 3,810 m). Travelers flying into these locations may require a period of acclimatization before proceeding higher, and drug prophylaxis may be indicated. Clinical Presentation {#cesec330} --------------------- Altitude illness is divided into three syndromes:•Acute mountain sickness (AMS)•High-altitude cerebral edema (HACE)•High-altitude pulmonary edema (HAPE) ### Acute Mountain Sickness (AMS) {#cesec331} AMS is the most common form of altitude illness, striking, for example, 25% of all visitors sleeping above 8,000 ft (2,500 m) in Colorado. Symptoms are those of an alcohol hangover: headache is the cardinal symptom, sometimes accompanied by fatigue, loss of appetite, nausea, and, occasionally, vomiting. Headache onset is usually 2--12 hours after arrival at a higher altitude, and often during or after the first night. Preverbal children may develop loss of appetite, irritability, and pallor. AMS generally resolves with 24--72 hours of acclimatization. ### High-Altitude Cerebral Edema (HACE) {#cesec332} HACE is a severe progression of AMS and is rare; it is most often associated with pulmonary edema. In addition to AMS symptoms, lethargy becomes profound, with drowsiness, confusion, and ataxia on tandem gait test. A person with HACE requires immediate descent; death from HACE can ensue within 24 hours of developing ataxia if the person fails to descend. ### High-Altitude Pulmonary Edema (HAPE) {#cesec333} HAPE can occur by itself or in conjunction with AMS and HACE; incidence is 1/10,000 skiers in Colorado and up to 1 of 100 climbers at \>14,000 ft (4,270 m). Initial symptoms are increased breathlessness with exertion, and eventually increased breathlessness at rest, associated with weakness and cough. Oxygen or descent of 1,000 m or more is life-saving. HAPE can be more rapidly fatal than HACE. ### Pre-Existing Medical Problems {#cesec334} •Travelers with medical conditions, such as heart failure, myocardial ischemia (angina), sickle cell disease, or any form of pulmonary insufficiency, should be advised to consult a physician familiar with high-altitude medical issues before undertaking high-altitude travel.•The risk for new ischemic heart disease in previously healthy travelers does not appear to be increased at high altitudes.•Diabetics can travel safely to high altitude, but they must be accustomed to exercise and carefully monitor their blood glucose. Diabetic ketoacidosis may be triggered by altitude illness and may be more difficult to treat in those on acetazolamide. Not all glucose meters may read accurately at high altitudes.•Most people do not have visual problems at high altitude. However, at very high altitudes some persons who have had radial keratotomy may develop acute farsightedness and be unable to climb by themselves. LASIK and other newer procedures may produce only minor visual disturbances at high altitudes.•There are no studies or case reports of harm to a fetus if the mother travels briefly to high altitude during pregnancy. However, it may be prudent to recommend that pregnant women stay at sleeping altitudes of 12,000 ft (3,658 m) if possible. The dangers of having a pregnancy complication in remote, mountainous terrain should also be discussed. Treatment {#cesec335} --------- ### Acetazolamide {#cesec336} Acetazolamide (Diamox) prevents AMS when taken before ascent and can speed recovery if taken after symptoms have developed. The drug works by acidifying the blood, which causes an increase in respiration and thus aids acclimatization. An effective dose that minimizes the common side effects of increased urination and paresthesias of the fingers and toes is 125 mg every 12 hours, beginning the day before ascent and continuing the first 2 days at altitude, or longer if ascent continues. Allergic reactions to acetazolamide are uncommon, but the drug is related to sulfonamides and should not be used by sulfa-allergic persons with history of anaphylaxis. A trial dose taken in a safe environment before travel may be useful for those with a more mild allergic history to sulfonamides. People with history of severe penicillin allergy have occasionally had allergic reactions to acetazolamide. ### Dexamethasone {#cesec337} Dexamethasone is very effective for prevention and treatment of AMS and HACE, and perhaps HAPE as well. Unlike acetazolamide, rebound can occur if the drug is discontinued at altitude prior to acclimatization. Acetazolamide is preferable to prevent AMS while ascending, with dexamethasone reserved for treatment during descent. Adult dosage is 4 mg every 6 hours. HAPE is always associated with increased pulmonary artery pressure, and pulmonary vasodilators are useful for preventing and treating HAPE. ### Nifedipine {#cesec338} Nifedipine prevents and ameliorates HAPE in persons who are particularly susceptible to the condition. The adult dosage is 20 mg of extended release every 8--12 hours. PDE-5 inhibitors can also selectively lower pulmonary artery pressure, with less effect on systemic blood pressure. ### Other Medications {#cesec339} Tadalafil (Cialis), 10 mg twice a day, during ascent can prevent HAPE and is being studied for treatment. When taken before ascent, gingko biloba, 100--120 mg twice daily, was shown to reduce AMS in adults in some trials, but it was not effective in others, probably due to variation in ingredients. Gingko biloba has not yet been compared directly with acetazolamide. Preventive Measures for Travelers {#cesec340} --------------------------------- The main point of instructing travelers about altitude illness is not to prevent any possibility of altitude illness, but to prevent death from altitude illness. The onset of symptoms and clinical course is sufficiently slow and predictable that there is no reason for someone to die from altitude illness unless trapped by weather or geography in a situation in which descent is impossible. The three rules that travelers should be made aware of to prevent death from altitude illness are---•Know the early symptoms of altitude illness and be willing to acknowledge when they are present.•Never ascend to sleep at a higher altitude when experiencing symptoms of altitude illness, no matter how minor they seem.•Descend if the symptoms become worse while resting at the same altitude. For trekking groups and expeditions going into remote high-altitude areas, where descent to a lower altitude could be problematic, a pressurization bag (such as the Gamow bag) can prove extremely beneficial. A foot pump produces an increased pressure of 2 lbs. per in[@bib250], mimicking a descent of 5,000--6,000 ft (1,500--1,800 m), depending on the starting altitude. The total packed weight of bag and pump is 6.5 kg. For most travelers, the best way to avoid altitude illness is by gradual ascent, with extra rest days at intermediate altitudes every 3,000 ft (900 m) or less. If ascent must be rapid, acetazolamide may be used prophylactically, and dexamethasone and pulmonary artery pressure-lowering drugs, such as nifedipine or sildenafil, may be carried for emergencies. JET LAG {#subchapter25} ======= Yanni Emad Description {#cesec341} ----------- Jet lag is a temporary disorder among air travelers who rapidly travel across three or more time zones. Jet lag results from the slow adjustment of the body clock to the destination time, so that daily rhythms and the internal drive for sleep and wakefulness are out of synchrony with the new environment. The intrinsic body clock resides in the suprachiasmatic nuclei at the base of the hypothalamus, which contains melatonin receptors. The body clock receives information about light from the eyes and is also thought to receive input via the intergeniculate leaflet that carries information about physical activities and general excitement. Melatonin is manufactured in the pineal gland from tryptophan, and its synthesis and release are stimulated by darkness and suppressed by light; consequently, the secretion of melatonin is responsible for setting our sleep--wake cycle. The body clock is adjusted to the solar day by rhythmic cues in the environment known as zeitgebers (time-givers). The main zeitgebers are the light--dark cycle and this rhythmic secretion of melatonin. Exercise might also exert a weaker effect on the body clock than other zeitgebers. Although incompletely understood, the body clock is partly responsible for the daily rhythms in core temperature and plasma hormone concentrations as well. Occurrence {#cesec342} ---------- •Eastward travel is associated with difficulty in falling asleep at the destination bedtime and difficulty arising in the morning.•Westward travel is associated with early evening sleepiness and predawn awakening.•Travelers flying within the same time zone typically experience the fewest problems.•Crossing more time zones or traveling eastward generally increases the time required for adaptation.•Jet lag lasts for several days, roughly equal to two-thirds the number of time zones crossed for eastward flights, and about half the number of time zones crossed after westward flights. Risk for Travelers {#cesec343} ------------------ •Individual responses to crossing time zones and the ability to adapt to new time zones vary. The intensity and duration of jet lag are related to the following:○Number of time zones crossed○Direction of travel○Ability to sleep while traveling○Availability and intensity of local circadian time cues at the destination○Individual differences in phase tolerance•Although more data are needed, risk factors cited by the American Academy of Sleep Medicine include the following:○Older individuals tend to experience fewer jet lag symptoms than those who are younger.○Exposure to local (natural) light--dark cycle usually accelerates adaptation after jet travel over 2 to 10 time zones Clinical Presentation {#cesec344} --------------------- Signs of jet lag include the following:•Poor sleep, including delayed sleep onset (after eastward flight), early awakening (after westward flight), and fractionated sleep (after flights in either direction).•Poor performance in both physical and mental tasks during the new daytime.•Negative subjective changes, such as increased fatigue, frequency of headaches and irritability, and decreased ability to concentrate.•Gastrointestinal disturbances (indigestion, frequency of defecation, and the altered consistency of stools) and decreased interest in and enjoyment of meals. Preventive Measures for Travelers {#cesec345} --------------------------------- ### Prior to Travel {#cesec346} •Stay healthy by continuing to exercise, eating a nutritious diet, and getting plenty of rest.•Consider timed bright light exposure prior to and during travel (although it requires high motivation and strict compliance with the prescribed light--dark schedules).•Break up the journey with a stop-over. ***Note:*** The use of the nutritional supplement melatonin is controversial for the prevention of jet lag. Some clinicians advocate the use of 0.5 mg to 5 mg of melatonin during the first few days of travel, and there are data to suggest its efficacy. However, the quality control of its production is not regulated by the U.S. Food and Drug Administration, and contaminants have been found in commercially available products. Current information does not support the use of special diets to ameliorate jet lag. ### During Travel {#cesec347} Travelers should be advised to---•Avoid large meals, alcohol, and caffeine.•Drink plenty of water.•Move around on the plane to promote mental and physical acuity.•Wear comfortable shoes and clothing.•Sleep, if possible, during long flights. ### On Arrival at the Destination {#cesec348} Travelers should be advised to---•Avoid situations requiring critical decision-making, such as important meetings, on the first day after arrival.•Adapt to the local schedule as soon as possible. However, if the travel period is 2 days or less, travelers should remain on home time.•Optimize exposure to sunlight following arrival in either direction.•Eat meals appropriate to the local time. Treatment {#cesec349} --------- •The 2008 American Academy of Sleep Medicine (AASD) recommendations include promoting sleep with hypnotic medication, although the effects of hypnotics on daytime symptoms of jet lag have not been well studied.•The prescription of nonaddictive sedative hypnotics (nonbenzodiazepines), such as zolpidem, has been shown in some studies to promote longer periods of high-quality sleep. If a benzodiazepine is preferred, a short-acting one, such as temazepam, is recommended to minimize oversedation the following day.•Because alcohol intake is often high during international travel, the risk for interaction with hypnotics should be emphasized with patients. MOTION SICKNESS {#subchapter26} =============== Carroll I. Dale Occurrence {#cesec350} ---------- Motion sickness is the result of a conflict between the various senses in regard to motion. The semicircular canals and otoliths in the inner ear sense angular and vertical motion, while the eyes and the proprioceptors determine the body\'s position in space. When signals received by the eyes or the proprioceptors do not match those being transmitted by the inner ear, motion sickness occurs. It can occur in either the presence or absence of actual motion, such as when viewing a slide through a microscope. Symptoms include nausea, vomiting, pallor, sweating, and often a sense of impending doom. Motion sickness is likely to occur when there is movement simultaneously in multiple planes, such as on amusement rides, on board ships, or during air travel. Risk for Travelers {#cesec351} ------------------ •All individuals, given sufficient stimulus, will develop motion sickness.•Children 2--12 years of age are especially susceptible, while infants and toddlers seem relatively immune.•Women, especially when pregnant, menstruating, or on hormones, are more likely to have motion sickness.•Persons with migraine are more prone to either migraine or motion sickness at the same time as the other malady.•Those who expect to be sick are more apt to experience symptoms. Treatment {#cesec352} --------- There are both nonpharmacologic and pharmacologic interventions for the prevention or management of motion sickness. None are ideal, and the medications typically cause drowsiness or similar adverse effects. Some feel that permitting continued exposure to motions that induce motion sickness will decondition the response and diminish the symptoms; however, most persons traveling for a limited time will understandably not be willing to endure the symptoms in the hope of deconditioning and will instead want to avail themselves of some of the suggestions that follow or the medications listed in [Table 2-26](#cetable25){ref-type="table"} .Table 2-26Pharmacologic interventions for motion sickness (adult dosing)MedicationDoseCaution/Safety InformationAdverse EffectsDrug Interactions**Vitamin supplements** Vitamin B~6~Various**Pyridoxine-doxylamine** (Example of brand: Diclectin)Fixed combination available by Rx. in Canada. Sold separately in U.S.More than 200,000 participants in controlled studies of nausea in pregnancy**Anticholinergic** Scopolamine (Examples of brands: Scopace, Transderm-scop)Patch: 1.5 mg q 3 days, apply behind ear at least 4 hrs before travelOral: 0.4-0.8 mg q 8 hrs beginning 1 hr before travelContraindicated in narrow-angle glaucoma, urinary retention, Gl obstruction, myasthenia gravis. Wash hands after patch application to prevent transfer to eyes.Caution in hot environment or with thyroid, cardiopulmonary, GE reflux, liver, or kidney disease, seizure or psychotic disorder*Common:* dry mouth/nose/throat, blurred vision, drowsiness*Less common:* palpitations, urinary retention, bloating, constipation, headache, confusion, hyperexcitability, insomnia, toxic psychosisAdditive effects with alcohol and other CNS depressants. Antacids impair absorption of oral scopolamine. May impair Gl motility when used with antidiarrheal drugs. May impair absorption of oral medications**Antihistamines** Dimenhydrinate (Examples of brands: Calm X, Dramamine, Triptone)Tablets 50 mg, syrup 12.5 mg/5 ml\_. Take 30 min before travel. Adults 50-100 mg q 4-6 hoursCaution in glaucoma, urinary retention, GI obstruction, liver or kidney disease, chronic obstructive pulmonary disease (COPD), seizure disorder. Should not be used in children \<2 yrs. Take with food or milk to reduce nausea.*Common:* drowsiness, anticholinergic symptoms (dry mouth/nose/throat, blurred vision, urinary retention), thick respiratory secretions*Less common:* dizziness, weakness, hypotension or hypertension, cardiac arrhythmia, wheezing, sweating, nausea, vomiting, bloating, diarrhea, constipation, jaundice, anorexia, headache, confusion, tinnitus, paradoxical hyperexcitability, seizures, psychosis, acute dystonic reaction, paresthesias, photosensitivity, anaphylaxisAdditive effects with alcohol and other CNS depressants. Antihistamine effects may be potentiated by monamine oxidase inhibitors. Antacids may impair absorption.**Diphenhydramine** (Brands: multiple)Available in oral capsules and tablets (25 mg, 50 mg),As aboveAs aboveAs above**Diphenhydramine**elixir (12.5 mg/5 mL). Adults 10-50 mg q 4-6 hours 12.5 mg, 25 mg, 50 mg**Meclizine** (Brands: Antivert \[Rx\], Bonine \[OTC\], Dramamine II \[OTC\], Meclicot \[Rx\], Medivert \[Rx\])Adult dose 25-50 mg q 24 hoursAs aboveAs aboveAs above**Cyclizine** (Brand: Marezine \[OTC\])50 mg tablets. Adult dose 50 mg q 4-6 hrsAs aboveAs aboveAs above**Antidopaminergic** Promethazine (Brands: Phenergan, Promacot)Available in oral tablets (12.5 mg, 25 mg, 50 mg), syrup (6.25 mg/5 mL, 25 mg/5 mL), rectal suppositories, and intramuscular injection. Adults: 25 mg every 8-12 hrs 5 mg and 10 mg tabletsCaution in sulfite allergy (some formulations contain sulfite), cardiovascular disease, peptic ulcer diseasePronounced sedation, postural hypotension, skin rash, body temperature dysregulation, extrapyramidal symptoms, delirium, neuroleptic malignant syndromeMay interact with other neurologic drugs**Metoclopramide** (Brands: Reglan)Adults: 10-15 mg q 6 hrsUnproven benefit as antinausea agent with motion sickness, but may help by hastening gastric emptyingSedation, insomnia, extrapyramidal symptomsMay decrease absorption of medications from stomach while increasing absorption from intestine. May necessitate change in insulin dose or timing in diabetics**Sympathomimetics** PseudoephedrineAdults: 60 mg q 6 hoursSometimes used to counteract sedating effect of other medicationsDifficult urination, dry mouth, restlessness, headache**Benzodiazepines** Diazepam (Brand: Valium)2 mg, 5 mg and 10 mg tablets. Adult dose 2-10 mg q 6 hrsVery sedating; perhaps of value when added to other medications**Other antiemetics** Prochlorperazine (Brand: Compazine)5 mg and 10 mg tablets. Adult dose 5-10 mg q 6 hrsEffective against nausea but not specific for motion sickness.May cause photosensitization, extrapyramidal symptoms**Ondansetron** (Brand: Zofran)4 mg and 8 mg tablets; Adult dose 4-8 mg q 8-12 hrsOrally disintegrating tablets contain phenylalanineContraindicated with apomorphine. Effect may be decreased with some anticonvulsants (carbamazepine, phenytoin) and rifamycin (rifampin, rifabutin) ### Medications {#cesec353} •Antihistamines are the most commonly used and available medications, although nonsedating ones appear to be the least effective.•Pyridoxine hydrochloride (vitamin B~6~) plus doxylamine succinate (an antihistamine) is prescribed under the brand name of Diclectin in Canada and often recommended in their separate forms by clinicians in the United States.•Sedation is the primary side effect of all the efficacious drugs.○Sedation is problematic when treating patients who perform essential tasks such as flying a plane or acting as crew on a ship, or in travelers who wish to participate in activities such as scuba diving or hang gliding.•Some common prescription medications used by travelers may aggravate the nausea of motion sickness (see [Table 2-27](#cetable26){ref-type="table"} ).Table 2-27Medications that may increase nauseaMedication ClassExamplesAntibioticsAzithromycin, metronidazole, erythromycin, trimethoprim-sulfamethoxazoleAntiparasiticsAlbendazole, thiabendazole, iodoquinol, chloroquine, mefloquineEstrogensOral contraceptives, estradiolCardiovascularDigoxin, levodopaNarcotic analgesicsCodeine, morphine, meperidineNonsteroidal analgesicsIbuprophen, naproxen, indomethacinAntidepressantsFluoxetine, paroxitene, sertralineAsthma medicationAminophyllineBisphosphonatesAlendronate sodium, ibandronate sodium, risedronate sodium #### Medications in Children {#cesec354} •For symptomatic treatment of children 2--12 years of age, dimenhydrinate, 1--1.5 mg/kg per dose, or diphenhydramine, 0.5--1 mg/kg per dose up to 25 mg, can be given 1 hour before travel and every 6 hours during the trip.•Because some children have paradoxical agitation with these medicines, a test dose should be given at home before departure.•Scopalamine causes potentially dangerous adverse effects in children and should not be used; prochlorperazine and metoclopramide should be used with caution in children.•Antihistamines are not FDA approved for use for the prevention or treatment of motion sickness in children. Caregivers should be reminded to always ask a physician, pharmacist, or other health-care professional if they have any questions about how to use or dose antihistamines in children before they administer the medication. Oversedation of young children with antihistamines can lead to life-threatening side effects. #### Medications in Pregnancy {#cesec355} •Drugs with the most safety data regarding the treatment of the nausea of pregnancy would seem to be the logical first choice.•Letter scoring of the safety of medications in pregnancy may not be helpful, and practitioners should review the actual safety data or call the patient\'s obstetrical provider for suggestions.•Web-based information may be found at [www.Motherisk.org](http://www.Motherisk.org){#interref66} and [www.Reprotox.org](http://www.Reprotox.org){#interref67}. Preventive Measures for Travelers {#cesec356} --------------------------------- Nonpharmacologic interventions include---•Being aware of those situations which tend to trigger symptoms.•**Optimizing positioning**---Driving a vehicle instead of riding in it, as well as sitting in the front seat of a car or bus, sitting over the wing of an aircraft or being in the central cabin on a ship can help reduce symptoms.•**Eating or drinking**---Eating before the onset of symptoms may hasten gastric emptying, but in some individuals, can aggravate motion sickness. Drinking caffeinated beverages along with taking one of the medications suggested can help manage motion sickness.•**Reducing sensory input**---The reduction of aggravating stimuli (e.g., lying prone, looking at the horizon, or shutting eyes) can help alleviate symptoms.•**Adding distractions**---Aromatherapy using mint, lavender, or ginger (oral) helps some; flavored lozenges may help as well. They may function as placebos or, in the case of oral ginger, may hasten gastric emptying.•**Using acupressure or magnets**---Advocated by some to prevent or treat nausea (not specifically for motion sickness), although scientific data are lacking. RESPIRATORY INFECTIONS {#subchapter27} ====================== LaRocque Regina C. Ryan Edward T. Respiratory infections are an underappreciated risk for travel. Respiratory infection is a leading cause of seeking medical care in returning travelers and has been reported to occur in up to 20% of all travelers. Thus, respiratory infections may be almost as common as travelers\' diarrhea. Upper respiratory infection is more common than lower respiratory infection. In general, the types of respiratory infections that affect travelers are similar to those in nontravelers, and exotic causes are rare. Travelers may be exposed to respiratory tract pathogens while in transit, while in close contact with other individuals, and while at their final destination. Infectious Agent {#cesec357} ---------------- •Viral pathogens are the most common cause of respiratory infection in travelers; causative agents include coronavirus, adenovirus, rhinovirus, influenza virus, parainfluenza virus, human metapneumovirus, and respiratory syncytial virus.•Bacterial pathogens are less common but include *Streptococcus pneumoniae*, *Mycoplasma pneumoniae*, *Haemophilus influenzae*, *Chlamydophila pneumoniae*, and *Legionella* species. Viral pathogens may set the stage for subsequent bacterial sinusitis or bronchitis. Occurrence {#cesec358} ---------- •Outbreaks are usually associated with common exposure in hotels and cruise ships or among tour groups.•A few specific pathogens have been associated with outbreaks in travelers, including influenza, *Legionella pneumophila*, severe acute respiratory syndrome (SARS), and histoplasmosis.•The peak influenza season in the temperate northern hemisphere is December through February. In the temperate southern hemisphere, the peak influenza season is June through August. Travelers to tropical zones are at risk year round.•Exposure to an infected individual from another hemisphere, such as on a cruise ship or package tour, can lead to an outbreak of influenza at any time or place. Risk for Travelers {#cesec359} ------------------ Factors contributing to respiratory infection in travelers include---•Air-pressure changes during ascent and descent of aircraft. These baropressure changes can facilitate the development of sinusitis and otitis media.•Intermingling of large numbers of people in airports, travel hubs, transport vehicles, cruise ships, and hotels can facilitate transmission.•Direct air-borne transmission of respiratory tract pathogens aboard aircraft is unusual because of frequent air recirculation and filtration, although sporadic cases of SARS, influenza, tuberculosis, and other agents have occurred in modern aircraft. Transmission of infection may occur between passengers who are seated in proximity to one another, usually through direct contact or droplets.•Air quality at many travel destinations may not be optimal, and exposure to sulfur dioxide, nitrogen dioxide, carbon monoxide, ozone, and particulate matter in air is associated with a number of health risks, including increased risk for respiratory tract inflammation, exacerbations of asthma and chronic obstructive pulmonary disease, and increased risks of bronchitis and pneumonia.•Certain epidemiologic characteristics of travelers that have been associated with a higher risk for respiratory tract infection include children, the elderly, and individuals with co-morbid pulmonary conditions, such as asthma and chronic obstructive pulmonary disease.•The risk for tuberculosis among travelers is very low (see the Tuberculosis section in Chapter 5). Clinical Presentation {#cesec360} --------------------- •Most respiratory tract infections, especially those of the upper respiratory tract, are mild and not incapacitating.•Lower respiratory tract infections, particularly pneumonia, can be more severe.•Individuals with influenza commonly have acute onset of fever, myalgia, headache, and cough.•Travelers with a viral upper respiratory infection may have persistent symptoms and should consider the possibility of subsequent bacterial sinusitis or bronchitis with symptoms that worsen after one week. Diagnosis {#cesec361} --------- •Identifying a specific etiologic agent, especially in the absence of pneumonia, is often difficult and not clinically necessary.•If indicated, the following methods of diagnosis can be used:○Molecular methods are available for the diagnosis of a number of respiratory viruses, including influenza virus, parainfluenza virus, adenovirus, human metapneumovirus, and respiratory syncytial virus, and for certain nonviral pathogens such as *Legionella pneumophila*.○Rapid tests are also available for detecting group A streptococcal pharyngitis.○Microbiologic culturing of sputum and blood, although insensitive, can assist in identifying a causative respiratory pathogen in persons with pneumonia. Treatment {#cesec362} --------- •Affected travelers are usually managed similarly to nontravelers, although travelers with progressive or severe illness should be evaluated for illnesses specific to their travel destinations and exposure history.•Most respiratory infections of travelers are due to viruses, are mild, and do not require specific treatment or antibiotics. No systematic study of self-treatment of travelers with respiratory infections has been reported.•Self-treatment usually involves supportive measures and may include the use of analgesics, decongestants, increased fluid intake, and inhaled moisture.•Self-treatment with antibiotics can be considered for upper respiratory infections that are worsening after 7 days of symptoms, particularly if specific symptoms of sinusitis or bronchitis are present. A respiratory-spectrum fluoroquinolone such as levofloxacin or a macrolide such as azithromycin may be prescribed to the traveler for this purpose prior to travel.•The rate of influenza infection among travelers is not known. The difficulty in self-diagnosing influenza makes it problematic to decide whether to provide travelers with a self-treatment dose of a neuraminidase inhibitor. This practice should probably be limited to travelers with a specific underlying condition that may predispose them to severe influenza. ### Medical Interventions {#cesec363} Specific situations that may require medical intervention include---•Pharyngitis without rhinorrhea, cough, or other symptoms that may indicate infection with group A streptococcus.•Sudden onset of cough, chest pain, and fever that may indicate pneumonia, resulting in a situation where the traveler may be sick enough to seek medical care right away.•Travelers with underlying medical conditions, such as asthma, pulmonary disease, or heart disease, who may need to seek medical care earlier than otherwise healthy travelers. Preventive Measures for Travelers {#cesec364} --------------------------------- •Vaccines are available for the prevention of a number of respiratory tract pathogens, including influenza, *S. pneumoniae*, *H. influenzae* type B (in young children), pertussis, diphtheria, varicella, and measles. Unless contraindicated, travelers should be vaccinated against influenza.•The prevention of respiratory illness while traveling may not be possible, but common-sense preventive measures include---○Trying to minimize close contact with persons who are coughing and sneezing○Frequent handwashing, either with soap and water or alcohol-based hand sanitizers (containing at least 60% alcohol)○Using a vasoconstricting nasal spray immediately prior to air travel, if the traveler has a pre-existing eustachean tube dysfunction. OCCUPATIONAL EXPOSURE TO HIV {#subchapter28} ============================ Warnock Eli W. III Chosewood L. Casey Risk for Health-Care Workers in International Locations {#cesec365} ------------------------------------------------------- The safety practices and facility standards in health-care settings of developing countries may be less stringent than those in developed settings. The health-care resources and training of health-care workers in these settings may also be limited. These conditions have the potential to increase the risk for occupational HIV exposure to visiting health-care workers in developing countries. Lack of access to personal protective equipment may also increase risk. Additionally, the prevalence of HIV infection in some developing countries is higher than that in the United States. Due to limited access to adequate treatment, infected source material in some developing countries may have higher viral loads than source material in the United States. Occupational exposure to source material with higher viral loads increases the risk for acquiring HIV occupationally. Infectious Agent {#cesec366} ---------------- Human immunodeficiency virus (HIV) is one of the pathogens, along with hepatitis C virus (HCV) and hepatitis B virus (HBV), that may be transmitted occupationally to health-care workers. Mode of Transmission {#cesec367} -------------------- Occupational transmission of HIV and transmission of other blood-borne pathogens typically occur via percutaneous exposure to contaminated sharps, including needles, lancets, scalpels, and broken glass. It can also occur when mucous membranes or nonintact skin comes into contact with infected blood or other body fluids. Occurrence {#cesec368} ---------- •The estimated annual number of health-care workers worldwide exposed to sharps injuries contaminated with HIV was 327,000 in 2005.•The risk of HIV infection following percutaneous exposure with a contaminated sharp is estimated to be 0.3%, or approximately 3 infections per 1,000 exposures.•Worldwide, the total number of HIV infections attributable to sharps injuries has been estimated to be 1,000 (range 200--5,000).•A 2005 study estimated that these infections would result in the worldwide premature deaths of 736 (range 129--3,578) health-care workers during the years 2000 to 2030. Preventive Measures for Travelers {#cesec369} --------------------------------- Health-care providers working internationally who will be engaging in high-risk occupational activities, such as drawing blood or the other use of sharps during patient care, should---•Consistently follow standard precautions to reduce the risk of occupational exposure to HIV and other blood-borne pathogens. Standard precautions involve the use of protective barriers such as gloves, gowns, aprons, masks, or protective eyewear, which can reduce the risk of exposure of the health-care worker\'s skin or mucous membranes to potentially infective materials. Additional information about occupational health and safety standards for blood-borne pathogens can be found on the Occupational Safety and Health Administration (OSHA) website at [www.osha.gov/pls/oshaweb/owadisp.show_document?p_table=STANDARDS&p_id=10051](http://www.osha.gov/pls/oshaweb/owadisp.show_document?p_table=STANDARDS&p_id=10051){#interref69}.•Always be mindful of the hazards posed by sharps injuries.•Maintain strict safety standards while working in environments that may have less stringent standards.•Use devices with safety features and improved work practices as recommended by the National Institute for Occupational Safety and Health (NIOSH) to prevent injuries caused by needles, scalpels, and other sharp instruments or devices. Additional information about preventing needlestick injuries in health-care settings can be found on the NIOSH website: [www.cdc.gov/niosh/2000-108.html\#8](http://www.cdc.gov/niosh/2000-108.html#8){#interref70}.•Consider bringing their own protective equipment if they are unsure of its availability at their destination.•Consider bringing postexposure prophylaxis (PEP) for HIV with them for use in the event that they experience a sharps injury with a contaminated or potentially contaminated needle. Postexposure Management {#cesec370} ----------------------- Health-care providers who have been occupationally exposed to HIV or have been exposed to potentially infectious material from a source person who is likely to be infected with HIV should immediately---•Wash the exposed area with soap and water thoroughly. If mucous membrane exposure has occurred, flush the area with copious amounts of water or saline.•Seek qualified medical evaluation as soon as possible to guide decisions on postexposure treatment and testing.•Contact the National Clinicians\' Postexposure Prophylaxis Hotline (PEPline) at 1-888-448-4911 (24 hours/7 days a week) for assistance in assessing risk and advice on managing occupational exposures to HIV, hepatitis, and other blood-borne pathogens. Additional information about PEPline can be found on the National HIV/AIDS Clinicians\' Consultation Center website at [www.ucsf.edu/hivcntr/Hotlines/PEPline.html](http://www.ucsf.edu/hivcntr/Hotlines/PEPline.html){#interref71}.•Consider beginning postexposure prophylaxis (PEP) for HIV. ### Postexposure Prophylaxis {#cesec371} •A number of medication combinations are available for PEP.•Refer to MMWR\'s Updated U.S. Public Health Service Guidelines for the Management of Occupational Exposures to HIV, Recommendations for Postexposure Prophylaxis, Updated Information Regarding Antiretroviral Agents Used as HIV Postexposure Prophylaxis for Occupational HIV Exposures ([www.aidsinfo.nih.gov/Guidelines/GuidelineDetail.aspx?MenuItem=Guidelines&Search=Off&GuidelineID=10&ClassID=3](http://www.aidsinfo.nih.gov/Guidelines/GuidelineDetail.aspx?MenuItem=Guidelines&Search=Off&GuidelineID=10&ClassID=3){#interref72}) and the PEPline for more information about PEP recommendations.•Specific regimens should be individually determined for those travelers at risk by health-care providers familiar with the medications and the traveler\'s medical history.•If the exposed person chooses to initiate PEP, they must do so within hours, as delays lead to a significant decline in PEP effectiveness.•If indicated, arrange for procurement or shipment of additional postexposure prophylaxis from a credible source to complete the recommended 4-week course of treatment.•Consider other potential infectious disease exposures from the source material as well, to include HBV or HCV, and manage if appropriate. Postexposure Testing {#cesec372} -------------------- •Persons with occupational exposure to HIV should receive baseline postexposure HIV-antibody testing by enzyme immunoassay, postexposure counseling, and medical evaluation, whether or not they receive PEP. In addition to baseline HIV-antibody testing, persons occupationally exposed should receive follow-up HIV-antibody testing by enzyme immunoassay for 6 months following exposure, at 6 weeks, 12 weeks, and 6 months (aidsinfo.nih.gov/contentfiles/HealthCareOccupExpoGL.pdf).•The U.S. Public Health Service also recommends that health-care workers occupationally exposed to a source co-infected with HIV and HCV and who acquire HCV infection receive extended HIV postexposure surveillance for up to 12 months following exposure.•Exposed health-care providers should be advised to use precautions (e.g., avoid blood or tissue donations, breastfeeding, or pregnancy) to prevent secondary transmission, especially during the first 6--12 weeks postexposure.•For exposures for which PEP is prescribed, health-care providers should be informed about---○possible drug toxicities and the need for monitoring○possible drug interactions○the need for adherence to PEP regimens•Consider re-evaluation of exposed health-care providers, if possible, 72 hours postexposure, especially after additional information about the exposure or source person becomes available and when adverse events of any medication can be assessed.   {#cesec373} = Counseling and Advice for Travelers PROTECTION AGAINST MOSQUITOES, TICKS, AND OTHER INSECTS AND ARTHROPODS {#subchapter29} ====================================================================== Zielinski-Gutierrez Emily Wirtz Robert A. Nasci Roger S. Although vaccines or chemoprophylactic drugs are available to protect against some important vector-borne diseases such as yellow fever and malaria, travelers still should be advised to use repellents and other general protective measures against biting arthropods. The effectiveness of malaria chemoprophylaxis is variable, depending on patterns of drug resistance, bio-availability, and compliance with medication, and no similar preventive measures exist for other mosquito-borne diseases such as dengue or chikungunya. CDC recommends the use of products containing active ingredients that have been registered by the U.S. Environmental Protection Agency (EPA) for use as repellents applied to skin and clothing (see below). EPA registration of active ingredients indicates the materials have been reviewed and approved for efficacy and human safety when applied according to the instructions on the label. General Protective Measures {#cesec374} --------------------------- •**Avoid outbreaks:** To the extent possible, travelers should avoid known foci of epidemic disease transmission. The CDC Travelers\' Health webpage provides alerts and information on regional disease transmission patterns and outbreak alerts ([www.cdc.gov/travel](http://www.cdc.gov/travel){#interref77}).•**Be aware of peak exposure times and places:** Exposure to arthropod bites may be reduced if travelers modify their patterns of activity or behavior. Although mosquitoes may bite at any time of day, peak biting activity for vectors of some diseases (e.g., dengue, chikungunya) is during daylight hours. Vectors of other diseases (e.g., malaria) are most active in twilight periods (i.e., dawn and dusk) or in the evening after dark. Avoiding the outdoors or focusing preventive actions during peak hours may reduce risk. Place also matters; ticks are often found in grasses and other vegetated areas. Local health officials or guides may be able to point out areas with greater arthropod activity.•**Wear appropriate clothing:** Travelers can minimize areas of exposed skin by wearing long-sleeved shirts, long pants, boots, and hats. Tucking in shirts and wearing socks and closed shoes instead of sandals may reduce risk. Repellents or insecticides such as permethrin can be applied to clothing and gear for added protection; this measure is discussed in detail below.•**Check for ticks:** Travelers should be advised to inspect themselves and their clothing for ticks during outdoor activity and at the end of the day. Prompt removal of attached ticks can prevent some infections.•**Bed nets:** When accommodations are not adequately screened or air conditioned, bed nets are essential to provide protection and to reduce discomfort caused by biting insects. If bed nets do not reach the floor, they should be tucked under mattresses. Bed nets are most effective when they are treated with an insecticide or repellent such as permethrin. Pretreated, long-lasting bed nets can be purchased prior to traveling, or nets can be treated after purchase. The permethrin will be effective for several months if the bed net is not washed. (Long-lasting pretreated nets may be effective for much longer.)•**Insecticides:** Aerosol insecticides, vaporizing mats and mosquito coils can help to clear rooms or areas of mosquitoes; however, some products available internationally may contain pesticides that are not registered in the United States. Insecticides should always be used with caution, avoiding direct inhalation of spray or smoke. **Optimum protection can be provided by applying the repellents described in the following sections to clothing and to exposed skin**. Repellents for Use on Skin and Clothing {#cesec375} --------------------------------------- CDC has evaluated information published in peer-reviewed scientific literature and data available from EPA to identify several EPA-registered products that provide repellent activity sufficient to help people avoid the bites of disease-carrying mosquitoes. Products containing the following active ingredients typically provide reasonably long-lasting protection:•**DEET** (chemical name: *N*,*N*-diethyl-*m*-toluamide or *N*,*N*-diethly-3-methyl-benzamide). Products containing DEET include but are not limited to Off!, Cutter, Sawyer, and Ultrathon.•**Picaridin** (KBR 3023, aka Bayrepel, and icaridin outside the United States; chemical name 2-(2-hydroxyethyl)-1-piperidinecarboxylic acid 1-methylpropyl ester). Products containing picaridin include but are not limited to Cutter Advanced, Skin So Soft Bug Guard Plus and Autan (outside the United States).•**Oil of lemon eucalyptus** [\*](#fn2){ref-type="fn"} or **PMD** (chemical name: *para*-menthane-3,8-diol) the synthesized version of oil of lemon eucalyptus. Products containing OLE and PMD include but are not limited to Repel.•**IR3535** (chemical name: 3-\[*N*-butyl-*N*-acetyl\]-aminopropionic acid, ethyl ester) Products containing IR3535 include but are not limited to Skin so Soft Bug Guard Plus Expedition. EPA characterizes the active ingredients DEET and picaridin as "conventional repellents" and oil of lemon eucalyptus, PMD, and IR3535 as "biopesticide repellents," which are derived from natural materials. ### Repellent Efficacy {#cesec376} •Published data indicate that repellent efficacy and duration of protection vary considerably among products and among mosquito species.•Product efficacy and duration of protection are also markedly affected by ambient temperature, amount of perspiration, exposure to water, abrasive removal, and other factors.•In general, **higher concentrations of active ingredient provide longer duration of protection**, regardless of the active ingredient. Products with £10% active ingredient may offer only limited protection, often from 1--2 hours.•Products that offer **sustained release or controlled release (i.e., micro-encapsulated) formulations, even with lower active ingredient concentrations, may provide longer protection times**.•Studies suggest that concentrations of DEET above ∼50% do not offer a marked increase in protection time against mosquitoes (i.e., DEET efficacy tends to plateau at around 50%).•Regardless of what product is used, if travelers start to get mosquito bites they should reapply the repellent according to the label instructions or leave the area with biting insects if possible. Repellents should be purchased before traveling and can be found in hardware stores, drug stores and supermarkets. A wider variety of repellents can be found in camping, sporting goods, and military surplus stores. When purchasing repellents overseas, look for the EPA-registered active ingredients on the product labels; some names of products available internationally have been specified above. ### Repellents and Sunscreen {#cesec377} Repellents that are applied according to label instructions may be used with sunscreen with no reduction in repellent activity. Products that combine sunscreen and repellent are not recommended, because sunscreen may need to be reapplied with greater frequency and in greater amounts than are needed to provide protection from biting insects. **In general, the recommendation is to apply sunscreen first, before applying the repellent**. Repellents/Insecticides for Use On Clothing {#cesec378} ------------------------------------------- •**Clothing, shoes, bed nets, mesh jackets, and camping gear can be treated with permethrin for added protection**.•Products such as Permanone and Sawyer permethrin are registered with EPA specifically for this use.•Permethrin is a highly effective insecticide and repellent. Permethrin-treated clothing repels and kills ticks, mosquitoes, and other arthropods. Clothing and other items must be treated several days in advance of travel to allow them to dry. As with all pesticides, follow the label instructions when using permethrin clothing treatments. Alternatively, clothing pretreated with permethrin is commercially available (e.g., products from Buzz Off/Insect Shield).•Permethrin-treated materials retain repellency/insecticidal activity after repeated laundering but should be retreated as described on the product label to provide continued protection. Clothing treated with the other repellent products described above (e.g., DEET) provides protection from biting arthropods but will not last through washing and will require more frequent reapplications. Precautions when Using Insect Repellents {#cesec379} ---------------------------------------- •Apply repellents only to exposed skin and/or clothing, as directed on the product label. Do not use repellents under clothing.•Never use repellents over cuts, wounds or irritated skin.•Do not apply repellents to eyes or mouth, and apply sparingly around ears. When using sprays, do not spray directly on face-spray on hands first and then apply to face. Wash hands after application to avoid accidental exposure to eyes.•Do not allow children to handle repellents. When using on children, adults should apply repellents to their hands first, and then put it on the child. It may be advisable to avoid applying to children\'s hands.•Use just enough repellent to cover exposed skin and/or clothing. Heavy application and saturation are generally unnecessary for effectiveness. If biting insects do not respond to a thin film of repellent, apply a bit more.•After returning indoors, wash treated skin with soap and water or bathe. This is particularly important when repellents are used repeatedly in a day or on consecutive days. Also, wash treated clothing before wearing it again. (This precaution may vary with different repellents---check the product label.)•If anyone experiences a rash or other bad reaction from an insect repellent, the repellent use should be discontinued, the repellent should be washed off with mild soap and water, and a local poison control center should be called for further guidance. If seeking health care because of the repellent, take the repellent to the doctor\'s office and show the doctor.•Permethrin should never be applied to skin, but only to clothing, bed nets, or other fabrics as directed on the product label. ### Children {#cesec380} •Most repellents can be used on children \>2 months of age.•Protect infants \<2 months of age from biting mosquitoes by using an infant carrier draped with mosquito netting with an elastic edge for a tight fit.•Products containing oil of lemon eucalyptus specify that they should not be used on children \<3 years of age.•Other than the safety tips listed above, EPA does not recommend any additional precautions for using registered repellents on children or on pregnant or lactating women. Useful Links {#cesec381} ------------ •U.S. Environmental Protection Agency. How to Use Insect Repellents Safely; \[updated 2007 July 5; cited 2008 Nov 29\]. Available from: [www.epa.gov/pesticides/health/mosquitoes/insectrp.htm](http://www.epa.gov/pesticides/health/mosquitoes/insectrp.htm){#interref78}.•Centers for Disease Control and Prevention. Insect Repellent Use and Safety; \[updated 2008 May 14; cited 2008 Nov 29\]. Available from: [www.cdc.gov/ncidod/dvbid/westnile/qa/insect_repellent.htm](http://www.cdc.gov/ncidod/dvbid/westnile/qa/insect_repellent.htm){#interref79}.•Health Canada\'s Pest Management Regulatory Agency. Safety Tips on Using Personal Insect Repellents; \[updated 2004 September 17; cited 2008 Nov 29\]. Available from: [www.pmra-arla.gc.ca/english/consum/insectrepellents-e.html](http://www.pmra-arla.gc.ca/english/consum/insectrepellents-e.html){#interref80}. WATER DISINFECTION FOR TRAVELERS {#subchapter30} ================================ Backer Howard D. Risk for Travelers {#cesec382} ------------------ Waterborne disease is a risk for international travelers who visit countries that have poor hygiene and inadequate sanitation, and for wilderness users relying on surface water in any country, including the United States. Worldwide, more than one billion people have no access to potable water and 2.4 billion do not have adequate sanitation. In developing countries, the influence of high-density population and rampant pollution, along with absent, overwhelmed, or insufficient sanitation and water treatment systems, means that surface water may be highly polluted with human waste and even urban tap water may become contaminated. Primarily humans, but also animals, are the source of microorganisms that contaminate water sources and cause intestinal infections. The list of potential waterborne pathogens is extensive and includes bacteria, viruses, protozoa, and parasitic helminths. Most of the organisms that can cause travelers\' diarrhea can be waterborne, although the majority of travelers\' intestinal infections are probably transmitted by food. Microorganisms with small infectious doses can even cause illness through recreational water exposure, via inadvertent water ingestion. Bottled water has become the convenient solution for most travelers, but in some places, it may not be superior to tap water. Moreover, the plastic bottles create a huge ecological problem, since most developing countries do not recycle plastic bottles. All international travelers, especially long-term travelers or expatriates, should become familiar with and utilize simple methods to ensure safe drinking water. Disinfection, the desired result of field water treatment, means the removal or destruction of harmful microorganisms. The goal of disinfection is to reduce the risk of gastrointestinal infection and diarrheal illness. [Table 2-28](#cetable27){ref-type="table"} compares benefits and limitations of different methods.Table 2-28Comparison of water disinfection techniquesTechniqueAdvantagesDisadvantagesHeat• Does not impart additional taste or color• Single step that inactivates all enteric pathogens• Efficacy is not compromised by contaminants or particles in the water as for halogenation and filtratino• Does not improve taste, smell or appearance of poor quality water• Fuel sources may be scarce, expensive or unavailable• Does not prevent recontamination during storageFiltration• Simple to operate• Requires no holding time for treatment• Large choice of commercial products• Adds no unpleasant taste and often improves taste and appearance of water• Rationally combined with halogens for removal or destruction of all pathogenic waterborne microbes• Adds bulk and weight to baggage• Many are not reliable for removal of viruses• Channeling of water or high pressure can force microorganisms through the filter• Relatively expensive, compared to chemical treatment• Eventually clogs from suspended particulate matter and may require some maintenance or repair in the fieldHalogens• Inexpensive and widely available in liquid or tablet forms• Taste can be removed by several techniques• Flexible dosing• Equally easy to treat large and small volumes• Corrosive and stains clothing• Imparts taste and odor to water• Flexibility requires understanding of principles• Iodine is physiologically active, with potential adverse effects• Not readily effective against *Cryptosporidium* oocysts• Efficacy decreases with low water temperature and decreasing water clarityChlorine Dioxide• Low doses have no taste or color• Simple to use and available in liquid or tablet form• More potent than equivalent doses of chlorine• Effective against all waterborne pathogens• Volatile and sensitive to sunlight: do not expose tablets to air and use generated solutions rapidly• No persistent residual, so does not prevent recontamination during storageUltraviolet• Imparts no taste• Portable devices now available• Effective against all waterborne pathogens• Requires clear water• Does not improve taste or appearance of water• Relatively expensive• Requires batteries or power source• Difficult to know if devices are delivering required UV doses Field Techniques for Water Treatment {#cesec383} ------------------------------------ ### Heat {#cesec384} Common intestinal pathogens are readily inactivated by heat. Microorganisms are killed in a shorter time at higher temperatures, whereas temperatures as low as 140° F (60° C) are effective with a longer contact time. Pasteurization uses this principle to kill food-borne enteric pathogens and spoiling organisms at temperatures between 140° F (60° C) and 158° F (70° C), well below the boiling point of water (212° F; 100° C). Although attaining boiling temperature is not necessary for inactivation of common intestinal pathogens, it is the only easily recognizable endpoint without using a thermometer. Microorganisms begin to die as water is heated on a stove or fire from 150° F (65° C) to boiling. All organisms except bacterial spores, which are not usually waterborne enteric pathogens, are killed within seconds at boiling temperature. Therefore, any water brought to a boil should be adequately disinfected. CDC and the Environmental Protection Agency recommend boiling for 1 minute to allow for a margin of safety and so users are clear that the water is truly boiling. Because the boiling point decreases with increasing altitude, CDC advises boiling water for 3 minutes at altitudes greater than 6,562 feet (\>2000 m). If no other means of water treatment is available, a potential alternative to boiling is to use tap water that is too hot to touch, which is probably at a temperature between 131° F (55° C) and 140° F (60° C). This temperature may be adequate to kill pathogens if the water has been kept hot in the tank for some time. However, because one cannot know for certain that this temperature has been maintained for long enough to kill all waterborne pathogens, boiling is still advisable if possible. Travelers with access to electricity can bring a small electric heating coil or a lightweight beverage warmer to boil water. ### Filtration {#cesec385} Filter pore size is the primary determinant of a filter\'s effectiveness, but microorganisms also adhere to filter media by electrochemical reactions. Microfilters with "absolute" pore sizes of 0.1--0.4 mm are usually effective for removal of cysts and bacteria but may not adequately remove viruses, which are a major concern in water with high levels of fecal contamination ([Table 2-29](#cetable28){ref-type="table"} ). Environmental Protection Agency (EPA) designation of water "purifier" indicates that company-sponsored testing has substantiated claims for removing 10^4^ (9,999 of 10,000) viruses although EPA does not independently test the validity of these claims.Table 2-29Microorganism size and susceptibility to filtrationOrganismAverage Size (μm)Maximum recommended filter rating (μm Absolute)[1](#cetablefn77){ref-type="table-fn"}Viruses0.03Not specifiedEnteric bacteria (*E. coli*)0.5 × 3.0--8.00.2--0.4*Cryptosporidium* oocyst4--61*Giardia* cyst6.0--10.0 × 8.0--15.03.0--5.0[^85] Reverse osmosis filtration can both remove microbiologic contamination and desalinate water. The high price and slow output of small hand-pump reverse-osmosis units currently prohibit use by land-based travelers; however, they are important survival aids for ocean voyagers. If the water supply is suspected of being heavily contaminated with biologic wastes and additional assurance is needed, then a second step with chemical treatment of the water before filtration can kill viruses. Many filters contain a charcoal stage that will remove the taste of added chlorine or iodine. ### Chemical Disinfection {#cesec386} The most common chemical water disinfectants are chlorine and iodine (halogens). Worldwide, chemical disinfection with chlorine is the most commonly used method for improving and maintaining microbiologic quality of drinking water. Sodium hypochlorite, common household bleach, is the primary disinfectant promoted by CDC and the WHO Safe Water System for individual household use in the developing world. Primary factors that determine the rate and proportion of microorganisms killed are the concentration of halogen (measured in mg/L or parts per million) and the length of time microorganisms are exposed to the halogen (contact time, measured in minutes). Given adequate concentrations and contact times, both chlorine and iodine have similar activity and are effective against many bacteria. Due to many uncontrolled factors in the field, extending the contact time adds a margin of safety. Cloudy water contains material that will use added disinfectant so it will require higher concentrations or contact times. However, some common waterborne parasites such as *Cryptosporidium*, are poorly inactivated by halogen disinfection, even at practical extended contact times. Therefore, chemical disinfection should be supplemented with adequate filtration to remove these disease-causing microorganisms from drinking water. Both chlorine and iodine are available in liquid and tablet form ([Table 2-30](#cetable29){ref-type="table"} ). Iodine has physiologic activity (it is used by the thyroid), so WHO recommends limiting iodine water disinfection to a few weeks of emergency use. It is not recommended in persons with unstable thyroid disease, known iodine allergy, or pregnancy (because of the potential effect on the fetal thyroid).Table 2-30Iodine and chlorine formulations and dosesIodination Techniques added to 1 liter or quart of waterYield 4 mg/L Contact (Wait) Time 45 min at 30° C 180 min at 5° C[1](#cetablefn78){ref-type="table-fn"}Yield 8 mg/L Contact (Wait) Time 15 min at 30° C 60 min at 5° C[1](#cetablefn78){ref-type="table-fn"}Iodine tablets (tetraglycine hydroperiodide) (e.g., Potable Aqua, Globaline)½ tablet1 tablet2% iodine solution (tincture)0.2 mL5 gtts[2](#cetablefn79){ref-type="table-fn"}0.4 mL10 gtts[2](#cetablefn79){ref-type="table-fn"}Saturated solution: iodine crystals in water (e.g., Polar Pure)13.0 mL26.0 mL**Chlorination techniquesYield 5 mg/LYield 10 mg/L**Sodium hypochlorite0.1 mL0.2 mL  Household bleach 5%2 drops4 dropsSodium dichloroisocyanurate (e.g., AquaClear)1 tabletChlorine plus flocculating agent (e.g., Chlor-floc)1 tablet[^86][^87] The taste of halogens in water can be improved by several means:•Reduce concentration and increase contact time.•Use a filter that contains activated carbon after contact time.•Add a tiny pinch of ascorbic acid (vitamin C, available in powder or crystal form and an ingredient in most flavored drink mixes) after the required contact time to remove the taste of halogens. (This works by converting iodine to iodide or chlorine to chloride, which have no taste or color.) ### Iodine Resins {#cesec387} Iodine resins transfer iodine to microorganisms that come into contact with the resin, but leave little iodine dissolved in the water. The resins have been incorporated into many different filter designs available for field use. Most contain a 1-mm cyst filter, which should effectively remove protozoan cysts (it should say 1-mm or 1 micron "absolute"). Few models are sold in the United States because of inconsistent test results, but some models are still available for international use. ### Salt (Sodium Chloride) Electrolysis {#cesec388} Passing a current through a simple brine salt solution generates mixed oxidants, primarily chlorine, which can be used for disinfection of microbes. See the discussion on chlorine above. The process was recently designed in a pocket-sized instrument that uses salt, water and electrical current generated from camera batteries to produce a disinfectant solution that is added to water. ### Chlorine Dioxide {#cesec389} Chlorine dioxide (ClO~2~) is capable of inactivating most water-borne pathogens, including *Cryptosporidium* oocysts, at practical doses and contact times. There are several new chemical methods for generating chlorine dioxide in the field for small-quantity water treatment. ### Ultraviolet (UV) Light {#cesec390} UV light can be used as a pathogen reduction method against microorganisms. The technology requires effective pre-filtering due to its dependence on low water turbidity (cloudiness), the correct power delivery, and correct contact times to achieve maximum pathogen reduction. UV might be an effective method in pathogen reduction in backcountry water. However, there is a lack of independent testing data available on specific systems. ### Solar Irradiation and Heating {#cesec391} UV irradiation by sunlight in the UVA range can substantially improve the microbiologic quality of water. Recent work has confirmed the efficacy and optimal procedures of the solar disinfection (SODIS) technique. Transparent bottles (e.g., clear plastic beverage bottles), preferably lying on a dark surface, are exposed to sunlight for a minimum of 4 hours. UV and thermal inactivation are synergistic for solar disinfection of drinking water. Use of a simple reflector or solar cooker can achieve temperatures of 149° F (65° C), which will pasteurize the water after 4 hours. In emergency situations such as refugee camps and disaster areas, where strong sunshine is available, solar disinfection of drinking water can improve water quality. ### Silver and Other Products {#cesec392} Silver ion has bactericidal effects in low doses and some attractive features, including absence of color, taste, and odor. The use of silver as a drinking water disinfectant is popular in Europe, but it is not approved for this purpose in the United States because silver concentration in water is strongly affected by adsorption onto the surface of the container and there has been limited testing on viruses and cysts. Several other common products have known antibacterial effects in water and are marketed in commercial products for travelers, including hydrogen peroxide, citrus juice, and potassium permanganate. None have sufficient data to recommend them for water disinfection in the field. Granular activated carbon (GAC) removes organic and inorganic chemicals (including chemical disinfectants) through adsorption onto carbon particles, thereby improving odor and taste. GAC may trap but does not kill microorganisms. GAC is a common component of field filters. Coagulation--flocculation (CF) removes suspended particles that cause a cloudy appearance and bad taste and do not settle by gravity; this process removes many but not all microorganisms. Alum, or one of several other substances, is added to the water, stirred well, allowed to settle, then poured through a simple coffee filter or fine cloth to remove the sediment. CF is an ancient technique that is still used routinely in municipal water treatment in conjunction with other treatment methods, such as disinfection, filtration, UV radiation, and ozonation. ### The Preferred Technique {#cesec393} The optimal technique for an individual or group depends on personal preference, size of the group, water source, and the style of travel. Boiling is most effective but may not be practical in all situations. Unfortunately, alternative treatment may require a two-step process of 1) coagulation--flocculation and/or filtration and 2) halogenation. It is best to filter first and then add the halogen. On long-distance, oceangoing boats where water must be desalinated during the voyage, only reverse-osmosis membrane filters are adequate. When the water will be stored for a period of time, such as on a boat, motor home, or a home with rainwater collection, halogens should be used to prevent the water from becoming recontaminated. A tightly sealed container is best to decrease risk of contamination. A minimum residual of 3--4 mg/L of hypochlorite should be maintained in the stored water. For short-term home storage, narrow-mouth jars or containers with water spigots prevent contamination from repeated contact with hands or utensils. SUNBURN {#subchapter31} ======= Ansdell Vernon E. Description {#cesec394} ----------- •Travelers to the tropics and subtropics are at increased risk of overexposure to the sun. Important consequences include sunburn, premature aging of the skin, wrinkling, and skin cancer, including melanoma.•Sunlight consists of ultraviolet (UV) rays (i.e., UVA, UVB, and UVC).○UVA rays are present throughout the day and are the most important cause of premature aging of the skin. In addition, UVA rays are responsible for photosensitivity reactions and also contribute to skin cancer.○UVB rays are intense from 10 am to 4 pm and are most responsible for sunburn and skin cancer development.○UVC rays are filtered by the ozone layer and do not reach the earth\'s surface.•The benefits of UV radiation include vitamin D protection, which is important for calcium absorption. Occurrence and Risk For Travelers {#cesec395} --------------------------------- Increased exposure to UV radiation occurs nearer the equator, during summer months, at higher elevation and between 10 am and 4 pm. Reflection from the snow, sand, and water increases exposure, a particularly important consideration for beach activities, skiing, swimming, and sailing. ### Commonly Used Medications that May Cause Photosensitivity Reactions {#cesec396} #### Antimicrobials {#cesec397} Fluoroquinolones, sulfonamides, and tetracyclines (especially demeclocycline); less frequently, doxycycline, oxytetracycline, and tetracycline; rarely, minocycline. #### Antimalarials {#cesec398} Doxycycline. #### Others {#cesec399} Nonsteroidal anti-inflammatory drugs, thiazide diuretics, furosemide, amiodarone, sulfonylureas, acetazolamide (Diamox), phenothiazines. Clinical Presentation {#cesec400} --------------------- •Symptoms from sunburn appear 3--5 hours after overexposure, worsen over the next 24--36 hours, and resolve in 3--5 days.•Serious burns are painful, and the skin may be tender, swollen and blistered. There may be fever, headache, itching, and malaise. Skin peeling occurs 3--8 days after excessive sun exposure.•Overexposure to the sun over several years leads to premature aging of the skin, wrinkling, age spots, and an increased risk for skin cancer, including melanoma.•Overexposure to the sun can cause red, dry painful eyes. Repeated exposure to sunlight results in pterygium formation and important causes of blindness such as cataracts and macular degeneration. Preventive Measures for Travelers {#cesec401} --------------------------------- ### Sun Protection Factor (SPF) {#cesec402} SPF defines the extra protection against UVB rays that an individual will get by using a sunscreen. For example, if a person using SPF 15 sunscreen normally acquires a sunburn within 20 minutes without protection, the benefit will be 20 × 15 minutes extra protection with sunscreen (i.e., 300 minutes = 5 hours). SPF does not refer to protection against UVA rays. Products containing Mexoryl, Parsol 1789, titanium dioxide, zinc oxide, or avobenzone block UVA rays. ### UV Index {#cesec403} The UV index provides travelers with an indication of the risk of UV radiation. Information is often available on the Internet or in local newspapers. The UV index ranges from 1 (low) to 11 or higher (extremely high). ### Sun Avoidance {#cesec404} Staying indoors or seeking shade between 10 am and 4 pm is very important in limiting exposure to UV rays, particularly UVB rays. Be aware that sunburn and sun damage can occur even on cloudy days. ### Protective Clothing {#cesec405} •Wide-brimmed hats and long sleeves and pants provide important protection against UV rays.•Tightly woven clothing and darker fabrics provide additional protection.•High SPF sun-protective clothing is recommended for those at increased risk of sunburn or with a history of skin cancer. This type of clothing contains colorless compounds, fluorescent brighteners, or specifically treated resins that absorb UV rays and often provides an SPF of 30 or higher.•Sunglasses that provide 100% protection against UV radiation are strongly recommended. ### Sunscreens {#cesec406} Sunscreens protect the skin by absorbing or reflecting UV radiation. ***Physical Sunscreens*** contain large particulate substances such as titanium dioxide and zinc oxide, which act to reflect and scatter both visible and UV light. They are effective sunscreens but are less popular because of aesthetically unappealing characteristics such as opaqueness and tendency to stain clothing. They are recommended for those who burn easily or who take medications that may cause photosensitivity reactions. ***Chemical Sunscreens*** absorb rather than reflect UV radiation. A combination of agents is recommended to provide broad-spectrum protection against UVA and UVB rays. #### Key Points Regarding Sunscreens {#cesec407} •Choose a sunscreen with at least 15 SPF.•Select a water- and sweat-resistant product that provides protection against both UVA and UVB rays.•Look for a sunscreen with at least three different active ingredients to provide broad-spectrum UVA and UVB ray protection. These ingredients generally include PABA derivatives, salicylates (homosalate, octyl salicylate), or cinnamates (octyll methoxycinnamate and cinoxate) for UVB ray absorption; benzophenones (oxybenzone, dioxybenzone, sulisobenzone) for shorter-wavelength UVA ray protection; and avobenzone (Parsol1789), ecamsule (Mexoryl), titanium dioxide, or zinc oxide for the remaining UVA spectrum.•Apply 30 minutes before exposure to the sun.•At least 1 oz of sunscreen is needed for total body application (i.e., quarter of a 4-oz bottle).•Apply to all exposed areas, especially the ears, scalp, lips, back of the neck, tops of the feet, and backs of the hands.•Reapply after 1--2 hours and after sweating, swimming, or toweling (even on cloudy days).•Many sunscreens lose potency after 1--2 years.•Sunscreens should be applied to the skin before insect repellents.•Avoid products that contain sunscreens and insect repellents. (DEET-containing insect repellents may decrease the effectiveness of sunscreens and may increase absorption of DEET through the skin.) Treatment {#cesec408} --------- •Hydration and staying in a cool, shaded, or indoor environment•Topical and oral nonsteroidal anti-inflammatory drugs decrease erythema if used before or soon after exposure to UVB rays and may relieve symptoms such as headache, fever, and local pain. Topical steroids are of limited benefit, and systemic steroids appear to be ineffective.•Moisturizing creams, aloe vera, and diphenhydramine may help to relieve symptoms.•In severe cases, narcotic analgesics may be indicated to relieve pain. PROBLEMS WITH HEAT AND COLD {#subchapter32} =========================== Backer Howard D. Shlim David R. Background {#cesec409} ---------- Foreign travel involves heading into new environments, and climate is one of the most important factors to consider. Travelers may encounter temperature and weather extremes that are either much hotter or colder than they are used to, and either extreme can have health consequences. Travelers should try to determine the likely climate extremes that they will face during their journey and to prepare with proper clothing, knowledge, and equipment. This section gives a brief overview of the topic. Problems Associated with a Hot Climate {#cesec410} -------------------------------------- ### Risk for Travelers {#cesec411} Many of the most popular travel destinations are tropical or desert areas. Travelers who sit on the beach or by the pool and do only short walking tours incur minimal risk of heat illness. Those who do strenuous hiking or biking in the heat may have significant risk, especially travelers coming from cool or temperate climates who are not in good physical condition and unacclimatized to the heat. ### Clinical Presentations {#cesec412} #### Physiology of Heat Injuries {#cesec413} Tolerance to heat depends primarily on physiologic factors, unlike cold environments where adaptive behaviors are more important. The major means of heat dissipation are radiation at rest and evaporation of sweat during exercise, both of which become minimal with air temperatures above 95° F (35° C) and high humidity. The major organs involved in temperature regulation are the skin, where sweating and heat exchange take place, and the cardiovascular system, which must greatly increase blood flow to shunt heat from the core to the surface while meeting the metabolic demands of exercise. Cardiovascular status and conditioning are the major physiologic variables affecting the response to heat stress at all ages. Dehydration is the most important predisposing factor in heat illness; temperature and heart rate increase in direct proportion to the level of dehydration. Sweat is a hypotonic fluid containing sodium and chloride. Sweat rates commonly reach 1--2 L/hr, which may result in significant fluid and sodium loss. #### Minor Heat Disorders {#cesec414} **Heat cramps** are painful muscle contractions following exercise in heat. They begin an hour or more after stopping exercise, most often involving heavily used muscles in the calves, thighs, and abdomen. If rest and passive stretching of the muscle do not resolve cramps, an oral salt solution, as in rehydration solutions, will rapidly relieve symptoms. **Heat syncope** is sudden fainting in heat that occurs in unacclimatized people while standing or after 15--20 minutes of exercise. Consciousness rapidly returns to normal when the patient is supine. Rest, relief from heat, and oral fluids are sufficient treatments. **Heat edema** is mild swelling of the hands and feet, which is more frequent in women during the first few days of heat exposure. It resolves spontaneously and should not be treated with diuretics, which may delay acclimatization and cause dehydration. **Prickly heat** (e.g., miliaria, heat rash) manifests as small, red, pruritic lesions on the skin caused by obstruction of the sweat ducts. It is best prevented by wearing light, loose clothing and avoiding heavy, continuous sweating. #### Major Heat Disorders {#cesec415} ##### Heat Exhaustion {#cesec416} •Most people who experience acute collapse or other symptoms associated with exercise in the heat are suffering from heat exhaustion, simply defined as the inability to continue exertion in the heat.•The presumed cause of heat exhaustion is loss of fluid and electrolytes, but there are no objective markers to define the syndrome, which is a spectrum ranging from minor complaints to a vague boundary shared with heat stroke.•Transient mental changes, such as irritability, confusion, or irrational behavior, may be present, but neurologic signs, such as seizures or coma, would indicate heat stroke or hyponatremia.•Body temperature may be normal or elevated.•Most cases can be treated with supine rest in a cool place and oral water or fluids containing glucose and salt. Spontaneous cooling occurs, and patients recover within hours, preventing progression to more serious illness. An oral solution for treating minor heat disorders or for fluid and electrolyte replacement can be made by adding ¼ teaspoon or two l-gm salt tablets to l liter of water, plus 4--8 tsp of sugar if desired for taste.•Subacute heat exhaustion may develop over several days and is often misdiagnosed as "summer flu" because of findings of weakness, fatigue, headache, dizziness, anorexia, nausea, vomiting, and diarrhea. Treatment is as described for acute heat exhaustion. ##### Exercise-Induced Hyponatremia {#cesec417} •Some travelers are so concerned about preventing heat illness and dehydration that they adopt the attitude that "you can\'t drink too much." Sadly, this attitude can lead to tragic outcomes.•Hyponatremia due to excessive water intake occurs in both endurance athletes and recreational hikers, particularly if the person is replacing sodium loss through sweating with plain water.•In the field setting, altered mental status with normal body temperature and a history of large volumes of water intake are highly suggestive of hyponatremia. The vague and nonspecific symptoms are the same as those described for hyponatremia in other settings (e.g., anorexia, nausea, emesis, headache, muscle weakness, lethargy, confusion, and seizures).•Until clinically apparent alterations in mental status appear, heat exhaustion is difficult to distinguish from early hyponatremia.•A delay before onset of major symptoms or deterioration after cessation of exercise and heat exposure are unique aspects of hyponatremia.•Prevention includes sodium supplementation with prolonged exercise or heat exposure. For hikers and wilderness users, food is the most efficient vehicle for salt replacement. Trail snacks should include salty foods (e.g., trail mix, crackers, pretzels, jerky), and not just sweets. ##### Heat Stroke {#cesec418} •Heat stroke is an extreme medical emergency requiring aggressive cooling measures and hospitalization for support.•Heat stroke is the only form of heat illness in which the mechanisms for thermal homeostasis have failed. As a result of uncontrolled fever and circulatory collapse, organ damage can occur in the brain, liver, kidneys, and heart.•The onset of heat stroke may be acute (exertional heat stroke) or gradual (nonexertional heat stroke, also referred to as classic or epidemic).•A presumptive diagnosis of heatstroke is made when patients have hyperpyrexia and marked alteration of mental status.•Body temperatures in excess of 106° F (41° C) can be observed; even without a thermometer, these patients will feel hot to touch. If a thermometer is available, a rectal temperature is the safest and most reliable way to check the temperature in someone who may have heatstroke.•In the field, institute evaporative cooling by maximizing skin exposure, spraying tepid water on the skin, and maintaining air movement over the body by fans. If ice is available, apply cold packs to the neck, axillas, and groin and massage the skin with ice. Immersion in a nearby pool or natural body of water can initiate cooling.•Unless the recovery is very rapid, the person should be evacuated to a hospital. If that is not possible, encourage rehydration, if the person is able to take oral fluids, and monitor closely for several hours. ### Prevention of Heat Disorders {#cesec419} #### Heat Acclimatization {#cesec420} Heat acclimatization is a process of physiologic adaptation to a hot environment that occurs in both residents and visitors. The result of acclimatization is an increase in sweating, and decreased energy expenditure with lower rise in body temperature for a given workload. Only partial adaptation occurs by passive exposure to heat. Full acclimatization, especially cardiovascular response, requires 1--2 hours of exercise in the heat each day. Most acclimatization changes occur within 10 days, provided a suitable amount of exercise is taken each day in the heat. After this time, only increased physical fitness will result in further exercise tolerance. Decay of acclimatization occurs within days to weeks if there is no heat exposure. #### Physical Conditioning and Acclimatization {#cesec421} Higher levels of physical fitness improve exercise tolerance and capacity in heat, but not as much as acclimatization. If possible, travelers should acclimatize before leaving by exercising at least 1 hour daily in the heat. If this is not possible before departing, exercise in heat during the first week of travel should be limited in intensity and duration (30- to 90-minute periods) with rest in between. It is a good idea to conform to the local practice in most hot regions and avoid strenuous activity during the hottest part of the day. #### Clothing {#cesec422} Clothing should be lightweight, loose, and light-colored to allow maximum air circulation for evaporation yet give protection from the sun. A wide-brimmed hat markedly reduces radiant heat exposure. #### Fluid and Electrolyte Replacement {#cesec423} During exertion, fluid intake improves performance and decreases the likelihood of illness. Reliance on thirst alone is not sufficient to prevent significant dehydration. During mild to moderate exertion, electrolyte replacement offers no advantage over plain water. However, for those exercising many hours in heat, a weak solution similar to commercial electrolyte drinks is recommended. Salty snacks or light salting of mealtime food or fluids is the most efficient way to replace salt losses. Salt tablets, when swallowed whole, may cause gastrointestinal irritation and vomiting, but two tablets can be dissolved in one liter of water. Urine volume and color are a readily available means to monitor fluid needs. Problems Associated with a Cold Climate {#cesec424} --------------------------------------- ### Risk for Travelers {#cesec425} Travelers do not have to be in an arctic or high-altitude environment to encounter problems with the cold. Humidity, rain, and wind can produce hypothermia even with temperatures around 50° F (12° C--14° C). Reports of severe hypothermia in international travelers are rare. Many high-altitude destinations are not wilderness areas, and villages offer an escape from extreme weather. In Nepal, trekkers almost never experience hypothermia except in the rare instance in which they may get lost in a storm. Even in a temperate climate, the traveler in a small boat that overturns in very cold water can rapidly become hypothermic. ### Clinical Presentations {#cesec426} #### Hypothermia {#cesec427} Hypothermia can be defined, in general terms, as having a core body temperature of \<95° F (35° C). When persons are faced with an environment in which they cannot keep warm, they first feel chilled, then begin to shiver, and eventually stop shivering as their metabolic reserves are exhausted. At that point, body temperature continues to decrease, dependent upon the ambient temperatures. As the core temperature falls, neurologic functioning decreases until almost all hypothermic people with a core temperature of £86° F (30° C) are comatose. The record low core body temperature in an adult who survived is 56° F (13° C). Travelers headed to a cold climate should be encouraged to ask questions and research appropriate clothing and equipment. Travelers who will be recreating or working around cold water face a different sort of risk. Immersion hypothermia can render a person unable to swim or keep floating within 30--60 minutes. In these cases, a personal flotation device is critical, as is knowledge about self-rescue and righting a capsized boat. The other medical conditions associated with cold affect mainly the skin and the extremities. These can be divided into nonfreezing cold injuries and freezing injuries (frostbite). #### Nonfreezing Cold Injury {#cesec428} The nonfreezing cold injuries are---•**Trench foot** (immersion foot): This condition is caused by prolonged immersion of the feet in cold water (32° F--59° F, 0° C--15° C). The damage is mainly to nerves and blood vessels, and the result is pain that is aggravated by heat and a dependent position of the limb. Severe cases can take months to resolve. Unlike the treatment for frostbite, immersion foot should not be rapidly rewarmed, which can make the damage much worse.•**Pernio** (chilblains): Pernio are localized, inflammatory lesions that occur mainly on the hands of susceptible individuals. They can occur with exposure to only moderately cold weather. The bluish-red lesions are thought to be caused by prolonged, cold-induced vasoconstriction. As with trench foot, rapid rewarming should be avoided, as it makes the pain worse. Nifedipine may be an effective treatment.•**Cold urticaria**: This condition involves the formation of localized or general wheals and itching after exposure to cold. It is not the absolute temperature that induces this form of urticara, but the rate of change of temperature in the skin. #### Freezing Cold Injury {#cesec429} ##### Categories of Frostbite {#cesec430} •Frostbite is the term that is used to describe tissue damage from direct freezing of the skin.•Modern equipment and clothing have greatly decreased the risk of frostbite in most adventurous tourist destinations, and frostbite occurs mainly during an accident, severe unexpected weather, or as a result of poor planning.•Once frostbite injury has occurred, little can be done to reverse the changes. Therefore, taking great care to prevent frostbite is crucial.•Frostbite is usually graded like burns.○First-degree frostbite involves reddening of the skin without deeper damage. The prognosis for complete healing is virtually 100%.○Second-degree frostbite involves blister formation. Blisters filled with clear fluid have a better prognosis than blood-tinged blisters.○Third-degree frostbite represents full-thickness injury to the skin, and possibly the underlying tissues. No blister forms, the skin darkens over time and may turn black, and if the tissue is completely devascularized, amputation will be necessary. ##### Management of Frostbite {#cesec431} Frostbitten skin is numb and appears whitish or waxy. The generally accepted method for treating a frozen digit or limb is through rapid rewarming in water heated to 104° F--108° F (40° C--42° C). The frozen area should be completely immersed in the warm water. A thermometer is needed to maintain the water at the correct temperature. Rewarming can be associated with severe pain, and analgesics can be given if needed. Once the area is rewarmed, it must be safeguarded against freezing again. It is thought to be better to keep digits frozen a little longer and rapidly rewarm them, than to allow them to thaw out slowly or to thaw and refreeze. A cycle of freeze--thaw--refreeze is devastating to tissue and leads more directly to the need for amputation. Once the area has rewarmed, it can be examined. If blisters are present, it is important to note whether they extend to the end of the digit. Proximal blisters usually mean that the tissue distal to the blister has suffered full-thickness damage. Treatment consists of avoiding further mechanical trauma to the area and preventing infection. Reasonable field treatment consists of washing the area thoroughly with a disinfectant such as povidone--iodine, putting dressings between the toes or fingers to prevent maceration, using fluffs (expanded gauze sponges) for padding, and covering with a roller gauze bandage. These dressings can safely be left on for up to 3 days at a time. By leaving the dressings on longer, the traveler can preserve what may be limited supplies of bandages. Prophylactic antibiotics are not needed in most situations. Once the patient has reached a definitive medical setting, there should be no rush to do surgery. The usual time from injury to surgery is 4--5 weeks. By that time the dead tissue has begun to separate from viable tissue, and the surgeon can plan surgery that maximizes the remaining digits. FOOD POISONING FROM MARINE TOXINS {#subchapter33} ================================= Ansdell Vernon E. Description {#cesec432} ----------- •Seafood poisoning from marine toxins is an underrecognized hazard for travelers, particularly in the tropics and subtropics. Furthermore, the risk is increasing as a result of multiple factors such as global warming, coral reef damage, and spread of toxic algal blooms.•**Ciguatera fish poisoning** and shellfish poisoning are caused by potent toxins that originate in small marine organisms (dinoflagellates and diatoms).•**Scombroid** poisoning is caused by eating improperly chilled fish that contains large quantities of histamine. Ciguatera Fish Poisoning {#cesec433} ------------------------ Ciguatera fish poisoning occurs after eating reef fish contaminated with toxins such as ciguatoxin or maitotoxin. These potent toxins originate from small marine organisms (dinoflagellates) that grow on and around coral reefs. Dinoflagellates are ingested by herbivorous fish, and the toxins are concentrated as they pass up the food chain to large (usually \>6 pounds) carnivorous fish and finally to humans. Toxin in fish is concentrated in the liver, intestinal tract, roe, and head. *Gambierdiscus toxicus*, which produces ciguatoxin, tends to proliferate on dead coral reefs. The risk of ciguatera is likely to increase as more coral reefs die as a result of factors such as global warming, construction, and nutrient runoff. ### Risk for Travelers {#cesec434} •Over 50,000 cases of ciguatera poisoning occur every year.•The incidence in travelers to endemic areas has been estimated as high as 3/100.•Ciguatera is widespread in tropical and subtropical waters, usually between the latitudes of 35 degrees north and 35 degrees south; it is particularly common in the Pacific and Indian Oceans and the Caribbean Sea.•Fish that are most likely to cause ciguatera poisoning are carnivorous reef fish, including barracuda, grouper, moray eel, amberjack, sea bass, or sturgeon. Omnivorous and herbivorous fish such as parrot fish, surgeon fish, and red snapper can also be a risk. ### Clinical Presentation {#cesec435} •Typical ciguatera poisoning results in a gastrointestinal illness, followed by neurologic symptoms and, very rarely, cardiovascular collapse.•The first symptoms usually appear 1--3 hours after eating contaminated fish and include nausea, vomiting, diarrhea, and abdominal pain.•Neurologic symptoms appear 3--72 hours after the meal and include paresthesias, pain in the teeth or the sensation that the teeth are loose, itching, metallic taste, blurred vision, or even transient blindness. Temperature reversal (hot objects feel cold and cold objects feel hot) is very characteristic. Neurologic symptoms usually last a few days to several weeks.•Chronic neuropsychiatric symptoms resembling chronic fatigue syndrome may be very disabling, last several months, and include malaise, depression, headaches, myalgias, and fatigue. Cardiac manifestations include bradycardia, other arrythmias, and hypotension.•Overall mortality from ciguatera poisoning is about 0.1% but varies due to the toxin dose absorbed and availability of adequate medical care to deal with serious complications such as cardiovascular collapse or respiratory failure.•The diagnosis of ciguatera poisoning is based on the clinical signs and symptoms and a history of eating fish that are known to carry ciguatera toxin. Commercial kits are available to test for ciguatera in fish, but there is no test for ciguatera in humans. ### Preventive Measures for Travelers {#cesec436} •Avoid or limit consumption of the reef fish listed above, particularly when the individual fish weighs 6 pounds or more.•Never eat high-risk fish such as barracuda or moray eel.•Avoid the parts of the fish that concentrate ciguatera toxin, such as liver, intestines, roe, and head.•Remember that ciguatera toxins do not affect the texture, taste or smell of fish, and they are not destroyed by gastric acid, cooking, smoking, freezing, canning, salting, or pickling.•Commercial kits (if available) can be used to check if the fish is safe to eat. ### Treatment {#cesec437} •There is no specific antidote for ciguatoxin or maitotoxin.•Treatment is generally symptomatic and supportive.•Intravenous mannitol has been reported to reduce the severity and duration of neurologic symptoms, particularly if given early. Scombroid {#cesec438} --------- Scombroid, one of the commonest fish poisonings, occurs worldwide in both temperate and tropical waters. The illness occurs after eating improperly refrigerated or preserved fish containing high levels of histamine and often resembles a moderate to severe IgE-mediated allergic reaction. Fish that cause scombroid have naturally high levels of histidine in the flesh and include tuna, mackerel, mahimahi (dolphin fish), sardine, anchovy, herring, bluefish, amberjack, and marlin. Histidine is converted to histamine by bacterial overgrowth in fish that has been improperly stored (over 20° C) after capture. Histamine and other scombrotoxins are resistant to cooking, smoking, canning, or freezing. Scombroid fish poisoning occurs worldwide in both temperate and tropical waters. ### Clinical Presentation {#cesec439} •Symptoms of scombroid poisoning resemble an acute allergic reaction and usually appear 10--60 minutes after eating contaminated fish. They include flushing of the face and upper body (resembling sunburn), severe headache, palpitations, itching, blurred vision, abdominal cramps, and diarrhea.•Untreated, symptoms usually resolve within 12 hours. Rarely, there may be respiratory compromise, malignant arrythmias, and hypotension requiring hospitalization.•Diagnosis is usually clinical. A clustering of cases helps to exclude the possibility of fish allergy. ### Preventive Measures for Travelers {#cesec440} •Fish contaminated with histamine may have a peppery, sharp, salty, or bubbly taste, but may also look, smell, and taste normal.•The key to prevention is to make sure that the fish is promptly chilled (below 15° C--20° C) after capture.•Cooking, smoking, canning, or freezing will not destroy histamine in contaminated fish. ### Treatment {#cesec441} •Scombroid poisoning usually responds well to H1 antihistamines.•H2 antihistamines may also be of benefit. Shellfish Poisoning {#cesec442} ------------------- There are several forms of shellfish poisoning. All occur after ingesting filter-feeding bivalve mollusks, such as mussels, oysters, clams, scallops, and cockles containing potent toxins. The toxins originate in small marine organisms (dinoflagellates or diatoms) that are ingested and concentrated by shellfish. ### Risk for Travelers {#cesec443} Contaminated shellfish may be found in temperate and tropical waters, typically during or after dinoflagellate blooms or "red tides." ### Clinical Presentation {#cesec444} Poisoning results in gastrointestinal and neurologic illness of varying severity. Symptoms typically appear 30--60 minutes after ingesting toxic shellfish but can be delayed for several hours. Diagnosis is usually made clinically together with a history of recent shellfish ingestion. #### Paralytic Shellfish Poisoning {#cesec445} This is the most common and most severe form of shellfish poisoning. Symptoms usually appear 30--60 minutes after eating toxic shellfish and include numbness and tingling of the face, lips, tongue, arms, and legs. There may be headache, nausea, vomiting, and diarrhea. Severe cases are associated with ingestion of large doses of toxin and clinical features such as ataxia, dysphagia, mental status changes, flaccid paralysis, and respiratory failure. The case--fatality rate averages 6% and may be particularly high in children. #### Neurotoxic Shellfish Poisoning {#cesec446} Usually presents as gastroenteritis accompanied by minor neurologic symptoms, resembling mild ciguatera poisoning or mild paralytic shellfish poisoning. Inhalation of aerosolized toxin in the sea spray associated with a red tide may cause an acute respiratory illness, rhinorrhea, and bronchoconstriction. #### Diarrheic Shellfish Poisoning {#cesec447} This produces chills, nausea, vomiting, abdominal cramps, and diarrhea. No fatalities have been reported. #### Amnesic Shellfish Poisoning {#cesec448} This is a rare form of shellfish poisoning that produces a gastroenteritis that may be accompanied by headache, confusion, and permanent short-term memory loss. In severe cases, seizures, paralysis, and death may occur. ### Preventive Measures for Travelers {#cesec449} •Shellfish poisoning can be prevented by avoiding potentially contaminated bivalve molluscs. This is particularly important in areas during or shortly after "red tides."•Travelers to developing countries should avoid eating all shellfish, because they carry a high risk of viral and bacterial infections.•Marine shellfish toxins cannot be destroyed by cooking or freezing. ### Treatment {#cesec450} •Treatment is symptomatic and supportive.•Severe cases of paralytic shellfish poisoning may require mechanical ventilation. ANIMAL-ASSOCIATED HAZARDS {#subchapter34} ========================= Marano Nina Galland G. Gale Human Interaction with Animals: A Risk Factor for Injury {#cesec451} -------------------------------------------------------- Animals in general tend to avoid human beings, but they can attack if they perceive threat, are protecting their young or territory, or are injured or ill. Although attacks by wild animals are more dramatic, attacks by domestic animals are far more common. Animals cause injury through bites, kicks, or blunt trauma, or by the use of horns or claws. Further damage can occur if injuries become secondarily infected, as these infections may result in serious systemic disease. In addition, animals can transmit zoonotic infections such as rabies. A recent 10-year retrospective review of dog bites in Austria showed that 75% of the bites were preventable because the person intentionally interacted with the dog. Bite Wounds {#cesec452} ----------- ### Prevention {#cesec453} •Before departure, travelers should have a current tetanus vaccination or should have documentation of receiving a booster vaccination within the prior 5--10 years. An assessment of the traveler\'s need for pre-exposure rabies immunization should be made according to guidelines in [Table 2-17](#cetable17){ref-type="table"}.•During travel, travelers should never try to pet, handle, or feed unfamiliar animals, domestic or wild, particularly in areas of endemic rabies. Young children are more likely to be bitten by animals and sustain more severe injuries from animal bites. ### Management {#cesec454} •All wounds should receive prompt local treatment by thorough cleansing and debridement of the wound if necrotic tissue or dirt is present to prevent infection and illness, especially tetanus or rabies-prone wounds (see the [Rabies](#subchapter9){ref-type="sec"} and [Tetanus](#subchapter19){ref-type="sec"} sections earlier in this chapter).•Any animal bite should be evaluated by a health-care provider as soon as possible, after cleaning the wound. Travelers who might have been exposed to rabies should contact a reliable health practitioner for advice about rabies postexposure prophylaxis (see the [Rabies](#subchapter9){ref-type="sec"} section earlier in this chapter). Since rabies immune globulin or rabies vaccine may not be available in the destination country, travelers should have a strategy in place prior to travel as to how to respond to a possible rabies exposure. This strategy may require the traveler to fly to a different country to obtain the appropriate treatment.○Travelers who have purchased medical evacuation insurance should contact their insurance provider for guidance on seeking medical care.○U.S. citizens can contact the local U.S. Embassy or Consulate in the country they are visiting for assistance in locating a health-care professional at their destination. Consular personnel at U.S. Embassies and Consulates abroad and in the United States are available 24 hours a day, 7 days a week, to provide emergency assistance to U.S. citizens. To contact the U.S. Department of State\'s Overseas Citizens Services:□Dial: 888-407-4747 if calling from the U.S. or Canada□Dial: 202-501-4444 if calling from overseas•Travelers who received their most recent tetanus toxoid-containing vaccine \>5 years previously or who have not received at least three doses of tetanus toxoid-containing vaccines may require a dose of tetanus toxoid-containing vaccine (Tdap, Td, or DTaP), according to the guidelines in [Table 2-21](#cetable21){ref-type="table"}. A wide variety of animals and insects can cause illness and injury to travelers; a short synopsis of risks by species is provided below. Monkeys {#cesec455} ------- •Macaques, a type of monkey, pose a threat for rabies and herpes B virus. Macaques are native to Asia and Northern Africa. They are also housed in research facilities, zoos, wildlife or amusement parks, and are kept as pets in private homes throughout the world. Monkey bites occasionally occur in certain urban sites, such as temples in Nepal or India.•Herpes B virus is related to the herpes simplex viruses, which cause oral and genital ulcers. Herpes B virus was discovered in 1933, and since that time approximately 50 cases have been reported in humans, with an 80% mortality rate. Herpes B infection is rare in humans, and most documented cases have resulted from occupational exposures. No cases of herpes B infection have been reported in travelers or others exposed to monkeys in the wild. However, travelers to areas where free-ranging macaques exist should be aware of the potential risk. An infected monkey may appear completely healthy.•Documented routes of human infection include animal bites and scratches, exposure to infected tissue or body fluids from splashes, and, in one instance, human-to-human spread. Even minor scratches or bites should be considered potential exposures as, experimentally, herpes B virus has been isolated from surfaces for up to 2 weeks after it was applied (unpublished data, National Institutes of Health B Virus Reference Laboratory).•The incubation period for herpes B may be less than 1 week to a month or longer.•Neurologic symptoms develop as the virus infects the central nervous system and may lead to ascending paralysis and respiratory failure.•Increased public and physician awareness about the risks associated with an injury from a macaque, improved first aid postexposure, the availability of better diagnostic tests, and improved anti-viral therapeutics have decreased the mortality rate to 20% in treated individuals. As a result, from 1987 to 2004 there have been only five fatal infections. ### Prevention {#cesec456} Travelers should never attempt to feed, pet, or otherwise handle any monkeys. ### Management {#cesec457} •Travelers should seek first aid immediately after being bitten or scratched by a monkey. The wound should be thoroughly cleaned, and travelers should seek health care immediately.•If the history is strongly suggestive of exposure to herpes B through contact with monkeys, there are published guidelines for the prevention of herpes B infection after exposure and for the treatment of established infection. These guidelines have recommendations for serologic tests and postexposure prophylaxis. When potentially exposed travelers return home, they should follow up with their health-care providers for care. Additional information and photos of macaques can be found at the website for the National B Virus Resource Center at the Georgia State University Viral Immunology Center: [www2.gsu.edu/∼wwwvir/](http://www2.gsu.edu/~wwwvir/){#interref83}. Snakes {#cesec458} ------ •Poisonous snakes are hazards in many locations, although deaths from snakebites are rare. Snakebites usually occur in areas where dense human populations coexist with dense snake populations (e.g., Southeast Asia, sub-Saharan Africa, and tropical America).•Common sense is the best precaution. Most snakebites are the direct result of startling, handling, or harassing snakes. Therefore, all snakes should be left alone. Travelers should maintain awareness of their surroundings, especially at night and during warm weather when snakes tend to be more active. For extra precaution, when practical, travelers should wear heavy, ankle high or higher boots, and long pants when walking outdoors at night in areas possibly inhabited by venomous snakes. ### Management {#cesec459} •Travelers should be advised to seek immediate medical attention any time a bite wound breaks the skin, or when snake venom is ejected into their eyes or mucous membranes.•Immobilization of the affected limb and application of a pressure bandage that does not restrict blood flow are recommended first-aid measures while the victim is moved as quickly as possible to a medical facility.•Incision of the bite site and tourniquets that restrict blood flow to the affected limb are not recommended.•Specific therapy for snakebites is controversial and should be left to the judgment of local emergency medical personnel. Specific antivenins are available for some snakes in some areas, so trying to ascertain the species of snake that bit the victim may be critical. Insects {#cesec460} ------- Bites and stings from insects such as spiders and scorpions can be painful and can result in significant morbidity and mortality, particularly among infants and children. Many insects can transmit communicable diseases, even without the traveler\'s awareness of the bite. This is particularly true when camping or staying in rustic accommodations. ### Prevention {#cesec461} Exposure to insect bites and scorpion envenomations can be avoided by wearing long sleeves and pants while hiking, sleeping under mosquito nets, and shaking clothing and shoes before putting them on. ### Management {#cesec462} Travelers should be advised to seek medical attention if an insect bite or sting causes redness, swelling, bruising, or persistent pain. Those who have a history of severe allergic reactions to insect bites or stings should also ask their physician to evaluate them for the need to carry an epinephrine autoinjector (EpiPen) to use in case of recurrence (both in general and especially while traveling). Bats {#cesec463} ---- •Bats can be found almost anywhere in the world except the polar regions and extreme deserts. Bats are reservoir hosts for viruses that can cross species barriers to infect humans and other domestic and wild mammals. Viruses such as rabies virus can be transmitted directly from bats to people.•It is not possible to tell if a bat has rabies; however, any bat that is active by day, is found in a place where bats are not usually seen (for example, indoors or outdoors in areas in close proximity to humans), or is unable to fly is far more likely than others to be rabid.•Human exposure to bats can occur during adventure activities such as caving. Exposure can include bites, scratches, and mucosal or cutaneous exposure to bat saliva. Like any other wild animal, any bat, whether it is sick or healthy, will bite in self-defense if handled. ### Prevention {#cesec464} Bats should never be handled. Travelers should be discouraged from going into caves that have a large bat infestation. Depending on the country being visited, pre-exposure rabies vaccination may be recommended for persons engaged in outdoor activities such as caving and spelunking. ### Management {#cesec465} •If a bite occurs or if infectious material (such as saliva) from a bat gets into the eyes, nose, mouth, or a wound, the traveler should wash the affected area thoroughly and get medical advice immediately. Any suspected or documented bite or scratch from a bat should be grounds for seeking postexposure rabies immunoprophylaxis.•People usually know when they have been bitten by a bat. However, bats have tiny teeth and not all wounds may be apparent. There are situations in which travelers should seek medical advice even in the absence of an obvious bite wound, such as upon awakening and finding a bat in the room or seeing a bat in the room of a child. Marine Animals {#cesec466} -------------- •Venomous injuries from marine fish and invertebrates are increasing with the popularity of surfing, scuba diving, and snorkeling. The majority of species responsible for human injuries and envenomation reside in tropical coastal waters and include stingrays, jellyfish, stonefish, and scorpionfish.•Travelers should be advised to use protective footwear and maintain vigilance while engaging in recreational water activities. Traumatic injury, envenomation and wound infection are common sequelae. Identification of the species involved is helpful in determining the best course of treatment. Birds {#cesec467} ----- •When traveling in an area that is experiencing an outbreak of avian influenza ([www.cdc.gov/flu/avian/outbreaks/current.htm](http://www.cdc.gov/flu/avian/outbreaks/current.htm){#interref84}), travelers should avoid all contact with poultry (e.g., chickens, ducks, geese, pigeons, turkeys, and quail) or any wild birds, and avoid settings where H5N1-infected poultry may be present, such as commercial or backyard poultry farms and live poultry markets.•Travelers should not eat uncooked or undercooked poultry or poultry products, including dishes made with uncooked poultry blood. DEEP VEIN THROMBOSIS AND PULMONARY EMBOLISM {#subchapter35} =========================================== Barbeau Deborah Nicolls Background {#cesec468} ---------- Venous thromboembolism (VTE) consists of two related conditions: 1) deep vein thrombosis (DVT) and 2) pulmonary embolism (PE). DVT occurs when there is a partial or complete blockage of a deep vein by a blood clot, most commonly in the legs. The clot may break off and travel to the vessels in the lung, causing a life-threatening PE. VTE associated with air travel was first described in the early 1950s. Previous studies have shown a two- to four-fold increased risk of VTE following air travel. In 2001, the World Health Organization set up the WHO Research into Global Hazards of Travel (WRIGHT) Project, a large collaborative research study to confirm the association between VTE and air travel. The goals of this project are to determine the magnitude of the risk of VTE due to air travel, to determine the effect of other factors on the association, and to study the effect of preventive measures on risk. The results of Phase I of the project were published recently. Phase II will address the effect of preventive measures. Risk for Travelers {#cesec469} ------------------ Several factors have been associated with an increased risk for developing VTE ([Box 2-4](#f13){ref-type="fig"} ).Box 2-4Risk factors for venous thromboembolism (VTE)Rights were not granted to include this box in electronic media. Please refer to the printed publication.© 2009 American Heart Association2009Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company\'s public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Combined effects have been observed between these established risk factors and different forms of travel. A population-based case--control study of adults receiving treatment for their first VTE found that long-distance travel (Δ4 hours) doubled the risk of VTE. The effect was greatest in the first week after travel but remained elevated for 2 months. Travel by air increased the risk to the same extent as travel by bus, train, or car, suggesting that the increased risk of air travel is due primarily to prolonged immobility. Synergistic effects were noted with factor V Leiden mutations, women who used oral contraceptives, BMI \>30 kg/m^2^, and height \>1.9 m (approximately 6 ft 3 in). Some of these effects were greatest following air travel. Furthermore, people shorter than 1.6 m (approximately 5 ft 3 in) had an increased risk of VTE only after prolonged air travel. These findings suggest that additional factors related to air travel may be involved in the increased risk for VTE. Occurrence {#cesec470} ---------- Two recent retrospective cohort studies address the issue of air travel-associated VTE incidence. The first was a cohort of 2,499 healthy Dutch commercial pilots. The incidence of VTE in this group was 0.3 per 1,000 person-years. When the data were adjusted for age and sex, the rate was not different from that in the general Dutch population. There was no association between the number of hours flown. The second study was among 8,755 employees of several international organizations. The overall incidence of VTE following flights \>4 hours was 1.4 per 1,000 person-years. The absolute risk of VTE was 1 per 4,656 flights. The rates of VTE were higher in women, especially those using oral contraceptives. Incidence was also higher in employees with a BMI \>25 kg/m^2^ and those with height \<1.65 m (5 ft 5 in) or \>1.85 m (6 ft 1 in). The risk of VTE increased with flight duration and with the number of times the employee flew during an 8-week period; the risk of VTE tripled in employees who went on five or more long-haul (Δ4 hours) flights. Each extra flight increased the risk of VTE 1.4-fold. The risk of VTE was highest in the first 2 weeks after a long-haul flight and gradually decreased to baseline after 8 weeks. Both these studies were performed among populations that are younger (mean age 35--40 years) and healthier than the general population and are not, therefore, generalizable to a higher-risk population. Clinical Presentation {#cesec471} --------------------- Symptoms of DVT include swelling, redness, pain, or tenderness, and increased warmth over the skin. It may be difficult to distinguish from muscle strain, injury, or skin infection. Symptoms of PE range from mild and nonspecific to acute, resembling heart attack or stroke. Once a clot has traveled to the lungs, common symptoms of PE are chest pain and shortness of breath. Other symptoms include dizziness, fainting, anxiety, and malaise. PE can occur in the absence of overt signs of DVT. Diagnosis {#cesec472} --------- Specialized imaging tests (e.g., duplex venous ultrasound, venography, computed tomography (CT) scans, and magnetic resonance imaging) are needed to make a definitive diagnosis of DVT. Helical CT or ventilation--perfusion scans are commonly used to diagnose PE. Preventive Measures for Travelers {#cesec473} --------------------------------- Several randomized, controlled trials have been performed to assess the effect of prophylactic measures on VTE risk after air travel. All studies examined the risk of asymptomatic DVT in travelers making flights Δ7 hours. All travelers were encouraged to do regular exercises and to drink nonalcoholic beverages during the flight. DVT was diagnosed by venous ultrasound from 90 minutes to 48 hours after the flight. Interventions that were studied include compression stockings, aspirin, low-molecular weight heparin, and various natural extracts with anticoagulant properties. No significant effect was seen in any of the pharmacologic interventions. Compression stockings (10--20 mm Hg and 20--30 mm Hg) were shown to significantly reduce the risk of asymptomatic DVT; however, four travelers wearing compression stockings in one study developed superficial thrombophlebitis. Symptomatic DVT and PE were not observed in any of the travelers enrolled in the studies. All travelers should keep hydrated, wear loose-fitting clothing, and make efforts to walk and stretch at regular intervals during long-distance travel. Compression stockings may be beneficial to travelers with other risk factors for VTE. Currently no convincing data suggest that pharmacologic interventions reduce the risk of significant VTE during travel. The American College of Chest Physicians published the 8th edition of their Antithrombotic and Thrombolytic Therapy Evidence-Based Clinical Practice Guidelines in a June 2008 Supplement to Chest. Recommendations for long-distance travel associated VTE are the following:•For travelers who are taking flights \>8 hours, the following general measures are recommended: avoidance of constrictive clothing around the lower extremities or waist, maintenance of adequate hydration, and frequent calf muscle contraction (Grade 1C).•For long-distance travelers with additional risk factors for VTE, we recommend the general measures listed above. If active thromboprophylaxis is considered because of a perceived high risk of VTE, we suggest the use of properly fitted, below-knee graduated compression stockings (GCS), providing 15--30 mm Hg of pressure at the ankle (Grade 2C), or a single prophylactic dose of low-molecular-weight heparin (LMWH), injected prior to departure (Grade 2C).•For long-distance travelers, we recommend against the use of aspirin for VTE prevention (Grade 1B). INJURIES AND SAFETY {#subchapter36} =================== Sleet David A. Wallace L.J. David Shlim David R. Overview {#cesec474} -------- According to the World Health Organization, injuries are among the leading causes of death and disability in the world, and they are the leading cause of preventable death in travelers. Of the approximately five million people killed due to injuries in the world, approximately 1.2 million people died of road traffic incidents, 815,000 from suicide and 520,000 from homicides. In addition to the considerable number of deaths, millions more are wounded or suffer other nonfatal health consequences. Worldwide, among persons aged 5--44 years, injuries account for 6 of the 15 leading causes of death. In 2007, just over 64 million Americans traveled outside the United States. The vast majority of these trips occurred without any serious health problems, but fatal and serious injuries occur to Americans every year while traveling internationally. Among travelers abroad, injuries are one of the leading causes of death. Compared with injuries, infectious diseases, for example, only account for a small proportion (2%) of deaths to overseas travelers. The U.S. Department of State collects data on U.S. citizens who die in a foreign country from non-natural causes for the most recent 3-year period and makes these data available on the Department of State website. These deaths are categorized by location where the death occurred, date of death, and cause of death. These deaths should be considered a conservative estimate of the true number of U.S. citizens who die in foreign countries, as some deaths may not be reported to the Department of State. We analyzed these data and found that from 2003 to 2005 an estimated 2,276 U.S. citizens died from injuries and violence while in foreign countries (excluding deaths occurring in Iraq and Afghanistan). Road traffic crashes headed the list of causes (34%), followed by homicide (17%), and drowning (13%) ([Figure 2-3](#f3){ref-type="fig"} ). By comparison to U.S. injury fatalities in 2003, road traffic crashes accounted for 27%, homicide 11%, and drowning 2% of all injury deaths.Figure 2-3Leading causes of injury death of U.S. citizens in foreign countries, 2003--2005.(From U.S. Department of State. U.S. citizen deaths from non-natural causes. Washington D.C.: U.S. Department of State. Available from: <http://travel.state.gov/law/family_issues/death/death_600.html>.) Depending on travel destination, duration, and planned activities, other common injury and safety concerns include natural hazards and disasters, civil unrest, terrorism, hate crimes against Americans, falls, burns, poisoning, drug-related overdose, and suicide. If seriously injured, emergency care may not be available or acceptable by U.S. standards. Trauma centers which are capable of providing optimal trauma care are uncommon outside urban areas. Males, compared with females, are more likely to die from injury causes while traveling internationally. Acquaintance rape and sexual assault are among the important risks to women travelers. Travelers should be aware of the increased risk of certain injuries while traveling abroad, particularly in low-income countries, and be prepared to take preventive steps to avoid them. Road Traffic Injuries {#cesec475} --------------------- Road traffic injuries are the leading cause of injury-related deaths worldwide. An estimated 3,000 people are killed each day around the world in road traffic crashes involving cars, buses, motorcycles, bicycles, trucks, or pedestrians. Each year another 20 to 50 million are seriously injured. In response to this crisis, in 2008 the United Nations General Assembly, passed resolution 62/244 'Improving global road safety" to strength international cooperation to develop policies and practices to reduce crash risks around the world. According to U.S. Department of State data, road traffic crashes are also the leading cause of injury death to U.S. citizens while traveling internationally (see [Figure 2-3](#f3){ref-type="fig"}). An estimated 768 Americans were killed in road traffic crashes in the period from 2003 to 2005. Approximately 13% of these road traffic deaths involved motorcycles and 7% were pedestrians. A study from Bermuda reported that tourists sustain a much higher rate of motorbike injuries than the local population, with the highest rate in persons aged 50--59 years. Loss of vehicular control, unfamiliar equipment, and inexperience with motorized two-wheelers contributed to crashes and injuries, even when traveling at speeds less than 30 mph. Road traffic crashes are also a leading cause of nonfatal injury among U.S. citizens requiring emergency transport back to the United States. Road traffic crashes are common in foreign tourists for a number of reasons: lack of familiarity with the roads, driving on the opposite side of the road than in one\'s home country, poorly made or maintained vehicles, travel fatigue, poor road surfaces without shoulders, unprotected curves and cliffs, and poor visibility due to lack of adequate lighting, both on the road and on the vehicle. In many low-income areas of the world, unsafe roads and vehicles and an inadequate transportation infrastructure contribute to the traffic injury problem. A safety concern in low-income countries is the mixing of motor vehicles with vulnerable road users such as pedestrians, bicyclists, and motorbike users. It is common in low-income countries to have cars, buses, and large trucks all sharing the same road with pedestrians, motorbikes, bicycles, rickshaws and even animals. This mixing of road users all in the same travel lane increases the risk for crashes and injuries. Sometimes travelers have few options in getting to remote areas, but if there are choices, they should look for better-maintained vehicles, daytime travel, seatbelts and a trained and licensed or hired driver. ### Prevention of Road Traffic Injuries {#cesec476} Health advisors should counsel the traveler to:•Use safety belts and child safety seats whenever possible. Safety belts reduce the risk of death in a crash by 45%--60%, child safety seats by 54% and infant seats by 70%. When traveling, rent newer vehicles with safety belts and airbags and bring a child safety seat from home.•Rent larger vehicles if possible, because they provide more protection in a crash.•Try to ride only in taxis with functional safety belts and ride in the rear seat.•Wear helmets when riding motorcycles, motorbikes, and bicycles. If helmets are likely to be unavailable at the destination, they should be brought from home.•Avoid drinking alcohol and driving or biking. U.S. data show that an alcohol-impaired driver has a 17 times greater risk of being involved in a fatal crash.•Visit the websites of the Association for International Road Travel (ASIRT) ([www.asirt.org](http://www.asirt.org){#interref86}) and Make Roads Safe ([www.makeroadssafe.org](http://www.makeroadssafe.org){#interref87}), both NGOs, which have useful safety tips for international travelers, including road safety checklists and country specific driving risks.•Check the safety and security information from the U.S. Department of State ([www.travel.state.gov](http://www.travel.state.gov){#interref88}).•Consider hiring a driver familiar with the destination, the language and an expert in maneuvering through local traffic.•Avoid riding on overcrowded, overweight, top heavy busses, or minivans, or riding with any driver who has consumed alcohol.•Be aware of pedestrians and be aware as a pedestrian of the dangers. Walk with a friend, rather than alone, as this helps with safety. Water Injuries {#cesec477} -------------- Drowning accounts for 13% of deaths of Americans abroad. The risk factors have not been clearly defined, but are suspected to be related to unfamiliarity with local water currents and water conditions. Drowning was the leading cause of injury death to Americans visiting countries where exposure to water recreation was a major activity such as Fiji, Dutch Antilles, Aruba, and Costa Rica. Studies have found that young men are particularly at risk of head and spinal cord injuries from diving into shallow water, with alcohol a factor in some cases. In 2000, approximately 449,000 people drowned worldwide; the exact number of travelers who suffer from nonfatal drowning is not precisely known. Alcohol is also a suspected contributing factor to both drowning and boating mishaps. Scuba diving is a frequent pursuit of travelers in ocean destinations. Travelers should either be experienced divers, or dive with a reliable dive shop and instructors. They should be reminded not to dive on the same day they arrive by airplane. The fatality rate among all divers, worldwide, is thought to be 15 to 20 deaths per 100,000 divers per year. Other Unintentional Injuries {#cesec478} ---------------------------- From 2003 to 2005, other than drowning, airplane crashes, natural disasters, and other unintentional injuries accounted for over a third of all injury deaths to Americans in foreign countries (see [Figure 2-3](#f3){ref-type="fig"}). Fires can be a substantial risk in low-income countries where building codes are not present or enforced, where there\'s an absence of smoke alarms, where there is no emergency access to 9-1-1 services, and where the fire department focus is on putting out fires rather than on fire prevention or victim rescue. ### Preventing Other Unintentional Injuries {#cesec479} Health advisors should counsel the traveler on the following:•Travelers should consider purchasing special health and evacuation insurance if their destinations include countries where there may not be access to good medical care.•Because trauma care is poor in many countries, victims of injuries can die before ever reaching a hospital, and there may be no coordinated ambulance services. In remote areas, medical assistance, drugs and medicines may be unavailable and travel can take a long time to the nearest medical facility.•Where possible avoid using local unscheduled small aircraft. If available choose larger aircraft (greater than 30 seats) as they have undergone more strict and regular safety inspections. Larger aircraft also provide more protection in the event of a crash. From 2003 to 2005 an estimated 83 Americans were killed in airplane crashes in foreign countries (see [Figure 2-3](#f3){ref-type="fig"}). For country-specific airline crash events, see [www.airsafe.com](http://www.airsafe.com){#interref89}.•To prevent fire-related injuries, select accommodations on the 6th floor or below (fire ladders generally cannot reach above the 6th). If possible, stay in hotels with smoke alarms and preferably sprinkler systems. Be alert for improperly vented heating devices which may cause poisoning from carbon monoxide (CO), a colorless odorless gas and by-product of all fossil fuel combustions. Some travelers choose to carry a personal CO detector. Travelers should identify two escape routes from buildings and remember to escape a fire by crawling low under smoke and by covering one\'s mouth with a wet cloth. Violence-Related Injuries {#cesec480} ------------------------- Violence is a leading worldwide public health problem and a growing concern of travelers. In 2000, about 1.6 million persons lost their lives to violence and only one-fifth were casualties of armed conflicts. Rates of violent deaths in low- to middle-income countries are more than 3 times those in higher-income countries, although there are great variations within countries, depending on regional demographic differences. Homicide was the second leading cause of injury death among American travelers in foreign countries accounting for almost 400 deaths from 2003 to 2005 (see [Figure 2-3](#f3){ref-type="fig"}). For some low-income countries such as Honduras, Colombia, Guatemala, and Haiti homicide was the leading cause of injury death for Americans accounting for 43%--65% of all injury deaths. Terrorism-related deaths among Americans in foreign countries, while alarming, are still relatively rare events and accounted for only 2% of all injury deaths (see [Figure 2-3](#f3){ref-type="fig"}). The vast majority of terrorism deaths among Americans occurred in countries of the Middle East. According to data from the State Department, 2003--2005, 82% of the injury deaths among Americans in Saudi Arabia and 55% of injury deaths in Israel/West Bank/Gaza were from terrorism. Suicide is the fourth leading cause of injury death to U.S. citizens traveling abroad (see [Figure 2-3](#f3){ref-type="fig"}). Factors contributing to homicide and suicide may be different while traveling than at home. Unfamiliarity with a destination, not being vigilant to one\'s surroundings, and alcohol involvement may increase risk of assault and homicide. For longer-term travelers (e.g., missionaries and volunteers), social isolation and substance abuse, particularly in the face of living in areas of poverty and rigid gender roles, may increase the risk of depression and suicide. If a traveler is the victim of a crime overseas, the nearest U.S. embassy, consulate, or consular agency for assistance should be contacted at [www.travel.state.gov](http://www.travel.state.gov){#interref90}. ### Prevention of Violence {#cesec481} U.S. travelers are viewed by many criminals as wealthy, naïve targets, who are inexperienced and unfamiliar with the culture and inept at seeking assistance once victimized. Traveling in high poverty areas, civil unrest, alcohol or drug use, and traveling in unfamiliar environments at night increase the likelihood that a U.S. traveler will be the victim of planned or random violence. To avoid violence while traveling, limit travel at night, travel with a companion, and vary the routine travel habits. Travelers should wear locally available accessories that are more typical of a country-savvy expatriate community and avoid expensive or provocative clothing, or accessories. Accommodations on the ground floor of hotels or immediately next to the stairwell should be avoided. Criminals are less likely to victimize upper level floors. All doors and windows should be locked. Some carry a door intruder alarm, a smoke alarm, and a rubber door stop that can be used as a supplemental door lock. Persons unknown to the traveler should not be invited into one\'s accommodations as this can be misinterpreted or against local laws and customs. The U.S. Department of State website ([www.travel.state.gov](http://www.travel.state.gov){#interref91}) has useful information regarding safety and security. Summary {#cesec482} ------- Injuries and violence are as much a public health problem to travelers overseas as are infectious and chronic diseases---and they are in many ways more deadly. Injuries are still the most frequent cause of death abroad in developing countries. Effective prevention strategies are available, particularly for travelers who find themselves in new environments and who may be more likely to be unaware of risks or complacent in exotic surroundings. Despite greater understanding and increased research efforts in this field, data on the magnitude and severity of injuries is still incomplete or unreliable in many countries. Existing data indicate that injury and violence are among the most important causes of premature death and ill-health to U.S. travelers overseas. Travel health advisors and other health-care providers should alert the public to the known risks and especially about simple and effective preventive measures to implement during international travel. NATURAL DISASTERS AND ENVIRONMENTAL HAZARDS {#subchapter37} =========================================== Malilay Josephine Batts Dahna Ansari Armin Miller Charles W. Brown Clive M. Natural Disasters {#cesec483} ----------------- Travelers should be aware of the potential for natural phenomena such as hurricanes, tornadoes, or earthquakes. Natural disasters can contribute to the transmission of some diseases, especially since water supplies and sewage systems may be disrupted, sanitation and hygiene compromised by population displacement and overcrowding, and normal public health services interrupted. ### Disease Risks {#cesec484} •The risk for infectious diseases is minimal unless a disease is endemic in an area prior to the disaster event, since transmission cannot take place unless the causative agent is present.○Although typhoid can be endemic in developing countries, natural disasters have seldom led to epidemic levels of disease.○Floods have been known to prompt outbreaks of leptospirosis in areas where the organism is found in water sources (see the Leptospirosis section in Chapter 5).•When water and sewage systems have been disrupted, safe water and food supplies are of great importance in preventing enteric disease transmission. If contamination is suspected, water should be boiled and appropriately disinfected (see the [Water Disinfection for Travelers](#subchapter30){ref-type="sec"} section earlier in this chapter).•Travelers who are injured during a natural disaster should have a medical evaluation to determine what additional care may be required for wounds potentially contaminated with feces, soil, or saliva or that have been exposed to fresh or sea water that may contain parasites or bacteria.•Tetanus booster status should always be kept current. ### Injuries {#cesec485} •When arriving at a destination, travelers should be familiar with local risks for seismic, flood-related, landslide-related, tsunami-related, and other hazards, as well as warning systems, evacuation routes, and shelters in areas of high risk.•After natural disasters, deaths are rarely due to infectious diseases but most often to blunt trauma, crush-related injuries, or drowning. Travelers should thus be aware of the risks for injury before, during, and after a natural disaster.•In floods, people should avoid driving through swiftly moving water.•Travelers should exercise caution during clean-up, particularly when encountering downed power lines, water-affected electrical outlets, interrupted gas lines, and stray or frightened animals.•During natural disasters, technological malfunctions may release hazardous materials (e.g., release of toxic chemicals from a point source displaced by strong winds, seismic motion, or rapidly moving water). ### Environmental Risks {#cesec486} •Natural disasters often lead to wide-ranging air pollution in large cities. Uncontrolled forest fires have caused widespread pollution over vast expanses of the world.•Natural or manmade disasters resulting in massive structural collapse or dust clouds can cause the release of chemical or biologic contaminants (e.g., asbestos or the arthrospores that lead to coccidioidomycosis).•Health risks associated with these environmental occurrences have not been fully studied.•Travelers with chronic pulmonary disease may be more susceptible to adverse effects from these exposures. ### Event-Specific Information {#cesec487} Typically, following natural disasters of a magnitude that may impact travelers, current information about the disaster, as well as travel health information specific to those needing entry into such regions, is provided on the CDC Travelers\' Health website ([www.cdc.gov/travel](http://www.cdc.gov/travel){#interref100}). Recommendations may include specific immunizations or cautions regarding unique hazards in the affected area. Environmental Hazards {#cesec488} --------------------- ### Air {#cesec489} •Air pollution may be found in large cities throughout the world; its sources are often attributed to automobile exhaust and industrial emissions and may be aggravated by climate and geography.•The harmful effects of air pollution are difficult to avoid when visiting some cities; limiting strenuous activity and not smoking can help.•Any risk to healthy short-term travelers to such areas is probably small, but persons with pre-existing health conditions (e.g., asthma or chronic obstructive pulmonary disease) could be more susceptible.•Avoidance of dust clouds and areas of heavy dust or haze may be wise. ### Water {#cesec490} •Rivers, lakes, and the ocean may be contaminated with---○organic or inorganic chemical compounds (e.g., heavy metals or other toxins);○harmful algal blooms (i.e., cyanobacteria) that can be toxic both to fish and to people who eat the fish or who swim or bathe in the water; and○pathogens from human and animal waste that may cause disease in swimmers.•Such hazards may not be immediately apparent in a body of water.•Extensive water damage after major hurricanes and floods increases the likelihood of mold contamination in buildings. U.S. residents may visit flooded areas overseas as part of emergency, medical, or humanitarian missions. Mold is a greater hazard for persons with conditions such as impaired host defenses or mold allergies. To prevent exposure that could result in adverse health effects from disturbed mold, persons should---○Avoid areas where mold contamination is obvious.○Use personal protective equipment (PPE) (e.g. gloves, goggles, tight-fitting NIOSH-approved N-95 respirator). Travelers should take sufficient PPE with them, as these may be scarce in the countries visited.○Keep hands, skin, and clothing clean and free from mold-contaminated dust.○The CDC MMWR guidance, "Mold Prevention Strategies and Possible Health Effects in the Aftermath of Hurricanes and Major Floods," provides recommendations for dealing with mold in these settings. ### Radiation {#cesec491} •Natural background radiation levels can vary substantially from region to region, but these natural variations are not a health concern for either the traveler or resident population.•Travelers should be aware of regions known to have been contaminated with radioactive materials, such as the area surrounding the Chernobyl nuclear power station, 100 km (62 miles) northwest of Kiev, Ukraine. This unprecedented radiation emergency and subsequent contamination primarily affected regions in three republics---Ukraine, Belarus, and Russia---with the highest radioactive ground contamination within 30 km (19 miles) of Chernobyl.•In most countries, known areas of radioactive contamination are fenced or marked with signs. These areas should not be trespassed.•Any traveler seeking long-term (more than a few months) residence near a known or suspected contaminated area should consult with staff of the nearest U.S. Embassy and inquire about any applicable advisories in that area regarding drinking water quality or purchase of meat, fruit, and vegetables from local farmers.•Radiation emergencies are rare events. In case of such an emergency, however, travelers should:○Follow instructions provided by local emergency and public health authorities.○If such information is not forthcoming, U.S. travelers should immediately seek advice from the nearest U.S. embassy.•Natural disasters (such as floods) may also result in displacement of industrial or clinical radioactive sources. In all circumstances, travelers should exercise caution when they encounter unknown objects or equipment, especially if they bear the radioactive symbol. If a questionable object is encountered, appropriate authorities should be notified. SCUBA DIVING {#subchapter38} ============ Nord Daniel A. Scuba diving can present a variety of unique medical challenges for the traveling diver. Because diving injuries are generally rare, few health-care providers are trained in their diagnosis and treatment. Thus, the recreational diver must be able to recognize the signs of injury and ensure the availability of dive medicine help when needed. Fitness to Dive {#cesec492} --------------- Planning for dive-related travel should take into account any changes in health status, recent injuries, or surgery. In general, respiratory disorders, as well as any disorders affecting higher function and consciousness (e.g., diabetes mellitus or seizures), respiratory function (e.g., asthma), psychological problems (e.g., anxiety), and pregnancy raise special concerns about diving fitness. Diving Disorders {#cesec493} ---------------- ### Barotrauma {#cesec494} #### Ear and Sinus {#cesec495} Ear barotrauma is the most common injury in divers. On descent, failure to equalize pressure changes within the middle ear space creates a pressure gradient across the eardrum, which can cause bleeding or fluid accumulation in the middle ear, as well as stretching or rupture of the eardrum and the membranes covering the windows of the inner ear. Symptoms can include---•pain•tinnitus (ringing in the ears)•vertigo (dizziness or sensation of spinning)•sensation of fullness•effusion (fluid accumulation in the ear)•decreased hearing Paranasal sinuses, because of their relatively narrow connecting passageways, are uniquely susceptible to barotraumas, generally on descent. With small changes in pressure (depth), symptoms are usually mild and short lived, but can be exacerbated by continued diving. Larger pressure changes, especially with forceful attempts at equilibration (e.g., valsalva maneuver), can be more injurious. Additional risk factors for ear and sinus barotrauma include---•earplugs•medications•ear and/or sinus surgery•nasal deformity•disease A diver who may have sustained ear or sinus barotrauma should discontinue diving and seek medical attention. #### Pulmonary {#cesec496} It is critical for a scuba diver to exhale (or breathe normally) while ascending slowly. Overinflation of the lungs, which usually happens when a novice diver panics, can result as a scuba diver ascends toward the surface without exhaling. During ascent, compressed gas trapped in the lung increases in volume until the expansion exceeds the elastic limit of lung tissue, causing damage and allowing gas bubbles to escape into one or more of three possible locations, as follows:•Gas entering the pleural space can cause lung collapse or pneumothorax.•Gas entering the mediastinum (space around the heart, trachea and esophagus) causes mediastinal emphysema and frequently tracks under the skin (subcutaneous emphysema) or into the tissue around the larynx, sometimes precipitating a change in the voice characteristics.•Gas rupturing the alveolar walls can dissect into the pulmonary capillaries and pass via the pulmonary veins to the left side of the heart, where it is distributed according to relative blood flow, resulting in arterial gas embolism (AGE). While mediastinal or subcutaneous emphysema usually resolves spontaneously, pneumothorax generally requires specific treatment to remove the air and reinflate the lung. AGE is a medical emergency requiring appropriate intervention, which includes recompression treatment with hyperbaric oxygen. Lung overinflation injuries from scuba diving can range from dramatic and life threatening to mild symptoms of chest pain and dyspnea. Although pulmonary barotrauma is relatively uncommon in divers, prompt medical evaluation is necessary, and evidence for this condition should always be considered in the presence of respiratory or neurologic symptoms following a dive. ### Decompression Illness {#cesec497} Decompression illness (DCI) is an all-inclusive term that describes the dysbaric injuries, AGE, and decompression sickness (DCS). Because the two diseases are considered to result from separate causes, they are described here separately. However, from a clinical and practical standpoint, distinguishing between them in the field may be impossible---and unnecessary, since the initial treatment is the same for both. DCI can occur even in divers who have carefully followed the standard decompression tables and the principles of safe diving. ### Arterial Gas Embolism (AGE) {#cesec498} Gas entering the arterial blood through ruptured pulmonary vessels can distribute bubbles into the body tissues, including the heart and brain, where they disrupt circulation. AGE may cause minimal neurologic symptoms or dramatic symptoms that require immediate attention. These signs and symptoms include---•numbness•weakness•tingling•dizziness•visual blurring•chest pain•personality change•paralysis or seizures•loss of consciousness•death In general, any scuba diver who surfaces unconscious or loses consciousness within 10 minutes after surfacing should be assumed to have AGE. Intervention with basic life support is indicated, including the administration of 100% oxygen, followed by rapid evacuation to a hyperbaric oxygen treatment facility. ### Decompression Sickness {#cesec499} Breathing air under pressure causes excess inert gas (usually nitrogen) to dissolve in body tissues. The amount dissolved is proportional to and increases with depth and time. As the diver ascends to the surface, the excess dissolved gas must be cleared through respiration via the bloodstream. Depending on the amount dissolved and the rate of ascent, some gas can supersaturate tissues, where it separates from solution to form bubbles, interfering with blood flow and tissue oxygenation and causing signs and symptoms of decompression sickness. These symptoms include---•joint aches or pain•numbness and/or tingling•mottling or marbling of skin•coughing spasms or shortness of breath•itching•unusual fatigue•dizziness•weakness•personality changes•loss of bowel or bladder function•staggering, loss of coordination, and/or tremors•paralysis•collapse or unconsciousness Serious permanent injury may result from either AGE or DCS. Flying after Diving {#cesec500} ------------------- The risk of developing decompression sickness is increased when divers are exposed to increased altitude too soon following a dive. The cabin pressure of commercial aircraft may be the equivalent of 8,000 ft (2,438 m). Thus, divers should avoid flying or an altitude exposure \>2,000 ft (610 m) for---•a minimum of 12 hours after surfacing from a single no-decompression dive, or•after repetitive dives and/or multiple days of diving, wait a minimum of 18 hours before ascending to altitude, to reduce the risk of decompression sickness. These recommended preflight surface intervals do not guarantee avoidance of DCS. Longer surface intervals will further reduce DCS risk. Prevention of Diving Disorders {#cesec501} ------------------------------ Recreational divers should dive conservatively and well within the limits of their dive tables or computers. Risk factors for DCI are primarily dive depth and bottom time; however, factors such as rapid ascent, repetitive dives, strenuous exercise, dives \>60 feet, altitude exposure soon after a dive, and physiological variability also increase risk. Divers should be cautioned to stay well hydrated and rested, dive within the limits of their training, and follow established guidelines for dives unique to their travel destination. Diving is a skill that requires appropriate training and certification and should be done with a companion. Treatment of Diving Disorders {#cesec502} ----------------------------- Definitive treatment of DCI begins with early recognition of symptoms, followed by recompression with hyperbaric oxygen. A high concentration (100%) of supplemental oxygen is considered effective first aid in relieving the signs and symptoms of decompression illness and should be administered as soon as possible. Divers are often dehydrated, either because of incidental causes, immersion, or DCI itself, which can cause a capillary leak. Administration of isotonic glucose-free intravenous fluid is recommended in most cases. Oral rehydration fluids may also be helpful, provided they can be safely administered (e.g., if the diver is conscious). The definitive treatment of DCI is recompression and oxygen administration in a hyperbaric chamber. The Divers Alert Network (DAN) maintains a 24-hour emergency consultation and evacuation assistance at 919-684-8111 or 919-684-4326 (collect calls are accepted). DAN will provide assistance with management of the injured diver, help in deciding if recompression is needed, the location of the closest recompression facility, and assistance in arranging patient transport. DAN can also be contacted for routine nonemergency consultation by telephone at 919-684-2948, ext. 222, or by accessing the website [www.diversalertnetwork.org](http://www.diversalertnetwork.org){#interref103}. Travelers who plan to scuba dive may want to ascertain whether there are recompression facilities at their destination prior to embarking on their trip. MEDICAL TOURISM {#subchapter39} =============== Reed Christie M. Introduction {#cesec503} ------------ Travel for the purpose of obtaining health care abroad has received a great deal of attention in the popular media recently---even Wikipedia has recently devoted a section to the practice (<http://en.wikipedia.org/wiki/Medical_tourism>). However, it is not the only form of "medical tourism." The term has also been applied to travel by health-care professionals for the purpose of providing health care. The extent of either form of travel is not well characterized, but the overarching issues for both types of travelers, their primary health-care providers, and travel medicine providers are outlined below. Travel to Obtain Care {#cesec504} --------------------- Data from the annual U.S. Department of Commerce in-flight survey during 2003--2006 show an overall annual increase in the number of trips taken by U.S. residents for which at least one purpose was health care. In 2006, there were approximately half a million overseas trips in which health treatment was at least one purpose of travel. Common cited procedures include:•Dentistry•Reproductive procedures•Surgeries (cosmetic, joint replacement, and cardiac) Lower cost is often mentioned as the motivation for this type of medical tourism, and an entire industry has grown up around this phenomenon. One can search for a provider and research accreditation status of the facility online, opt for an online concierge service that will make all the arrangements or, more recently, find that health insurance coverage may include the option of "outsourced" health care. The dynamic nature of the field was described in a recent roundtable discussion in Merrell et al.,"*In recent years, standards have been rising in other parts of the world even faster than prices have surged in the U.S. Many physicians abroad trained in the U.S. and the Joint Commission International (JCI) applies strict standards to accreditation of offshore facilities. Those facilities use the same implants, supplies, and drugs as their U.S. counterparts. However, a heart bypass in Thailand costs \$11,000 compared to as much as \$130,000 in the U.S. Spinal fusion surgery in India at \$5,500 compares to over \$60,000 in the U.S.*" However, the quality of facilities, assistance services, and care is neither uniform nor regulated; thus, in most instances, responsibility for assessing suitability of an individual program or facility lies solely with the traveler. ### Guidelines for Travelers Seeking Care Abroad {#cesec505} Potential patients should consider that, whatever procedure is being contemplated, travelers undergoing medical treatment outside their accustomed environment are almost always at a disadvantage, particularly if there are complications. Concerns are---•Resolution of financial issues if costs escalate, such as in the case of complications.•Language and cultural differences may impede accurate interpretation of both verbal and nonverbal communication.•Religious and ethical differences may be encountered over issues such as heroic efforts to preserve life or limb or in care of the terminally ill.•Lack of familiarity with the local medical system, limited access to past medical history, unfamiliar drugs and medicines.•Legal recourse may be fairly limited, difficult to obtain, or nonexistent.•Follow-up care back in the United States may be more difficult to arrange and may be fraught with problems, should there be complications. Potential patients should consider the guiding principles developed by the American Medical Association for employers, insurance companies, and other entities that facilitate or offer incentives for care outside the United States, although in some circumstances it is unclear how realistic they may be (see [www.ama-assn.org/ama1/pub/upload/mm/31/medicaltourism.pdf](http://www.ama-assn.org/ama1/pub/upload/mm/31/medicaltourism.pdf){#interref105}). These principles stipulate that international care must be voluntary and provided by accredited institutions; financial incentives should not inappropriately limit or restrict patient options; there should be continuity of care, including coverage of costs upon return; patients should be informed of their rights and legal recourse before travel; patients should have access to licensing, outcome, and accrediting information when seeking care; medical record transfers should comply with Health Insurance Portability and Accountability Act (HIPAA) guidelines; and patients should be informed of potential risks of combining surgical procedures with long flights and vacation activities. The American Society for Plastic Surgery emphasizes plastic surgery is "real" surgery and outlines the issues every patient undergoing surgery should consider, whether at home or abroad, on their website at [www.plasticsurgery.org/patients_consumers/patient_safety/Medical-Tourism.cfm](http://www.plasticsurgery.org/patients_consumers/patient_safety/Medical-Tourism.cfm){#interref106}. Several clusters of mycobacterial wound infections in travelers returning from cosmetic procedures abroad have been published. Similarly, the American Dental Association provides informational documents, including: "Traveler\'s Guide to Safe Dental Care" through the Global Dental Safety Organization for Safety and Asepsis Procedures at [www.osap.org](http://www.osap.org){#interref107} and "Dental Care Away from Home" at [www.ada.org/public/manage/care/index.asp](http://www.ada.org/public/manage/care/index.asp){#interref108}. Individuals researching accreditation status should note that, although facilities may be part of a chain, they are surveyed and accredited individually. They should also check the duration of the accreditation and validate that the information is current by consulting the public portion of the appropriate accrediting agency website (see [references](#cebib39){ref-type="sec"} below). ### Pre-Travel Advice for the Medical Tourist {#cesec506} As discussed in the Planning for Healthy Travel section in Chapter 1, patients who do elect to travel should consult a travel health-care practitioner for advice tailored to individual health needs, preferably at least 4--6 weeks in advance of travel. This is particularly true for patients considering invasive procedures, who should consult as soon as travel is considered to allow for assessment of hepatitis B vaccination status (see the [Hepatitis B](#subchapter4){ref-type="sec"} section earlier in this chapter). Hepatitis B and C viruses and HIV are examples of blood-borne infections that can be transmitted via contaminated equipment, from infected health-care providers during invasive procedures, via transfusion of blood or blood products, or through transplantation of tissue or organs that have not been properly screened. Prevalence rates of these viruses vary considerably around the world and are generally higher in developing parts of the world than in the United States. U.S. policies address hepatitis B vaccination status of health-care workers, but these policies are not uniform worldwide and there are no currently licensed vaccines for hepatitis C and HIV. Blood transfusion programs in the United States and other developed areas rely on voluntary, nonremunerated donors; screen the donated blood for a variety of potentially blood-borne pathogens; and are closely regulated. Standards in other parts of the world vary. Based on data from 2000--2001, the latest available on the WHO Global Database on Blood Safety ([www.who.int/bloodsafety/global_database/en/](http://www.who.int/bloodsafety/global_database/en/){#interref109}), 70 countries did not test all donated blood for the three major blood-borne viruses, HIV and hepatitis B and C. ### Organ Transplantation {#cesec507} Organ transplantation in the United States is also a voluntary, closely monitored process coordinated by the United Network for Organ Sharing ([www.optn.org](http://www.optn.org){#interref110}). The need for transplantable organs, however, far exceeds the available supply worldwide. Travel to a country with less rigorous methods of distribution for the purpose of obtaining a transplant has been termed "transplant tourism" or "organ trafficking." Recently, there have been reports in the media of investigations and arrests associated with "rings" that use unscrupulous methods to obtain organs. In 2004, the World Health Assembly Resolution 57.18 encouraged member countries to protect vulnerable populations. Some countries have begun experimenting with controlled programs to relieve the shortage, support the health of the donor, and remove incentives for clandestine operations. A revised set of eleven WHO Guiding Principles on Human Cell, Tissue and Organ Transplantation will be presented to the World Health Assembly in 2009 ([www.who.int/transplantation/](http://www.who.int/transplantation/){#interref111}). Travel for the Purpose of Delivering Health Care {#cesec508} ------------------------------------------------ There are many structured opportunities for health-care professionals, students, or trainees to participate in established programs in developing areas of the world that are mutually beneficial to both the local population and the traveler. Travel by health-care workers in their professional capacity should be governed by the principle of *Primum non nocere*, or "first, do no harm." The traveling health-care worker should have sufficient experience or be at a stage in training to be able to contribute labor, knowledge, and skills to the host community. Benefits to the traveling health-care worker include exposure to patients with tropical diseases and conditions that are not commonly seen or are at a more advanced stage than in the country of residence; local diagnostic skills which are often less dependent on technology; and new cultures and new ways of thinking, in addition to any personal gratification. Many medical schools and universities have established reciprocal relationships with institutions in developing areas in which there is an exchange of students and faculty. A variety of organizations match volunteers with local needs for skills-building or to address specific problems. Doctors Without Borders/Médecins Sans Frontières (MSF), which received the Nobel Peace Prize in 1999 for humanitarian efforts around the world, requires a minimum 6-month commitment from physicians and a shorter commitment for surgeons. Interventional programs such as dentistry or surgery simultaneously provide reparative or reconstructive services to the population and train local staff to perform the procedures and provide follow-up care, often donating excess supplies. Other ongoing volunteer relationships exist between faith-based or service organizations and local communities. The involvement of the local health establishment is key to determining needs and maximizing benefit to the local population, as well as educating the visitors on local customs and medical issues and providing translation, if needed, to adequately assess the patients, obtain consent, and advise on postprocedure care. These forms of international capacity-building should be differentiated from---•medicine that is practiced on local populations ad hoc by independent travelers to areas that seem to have no system of health care,•the development of adventure holidays sold to groups of doctors specifically for the purposes of research or providing health care in the absence of prior consultation, and•students or trainees who travel to "gain practical experience" beyond their training with minimal supervision or absence of structured learning, or practitioners performing outside the area of their expertise. The acts performed in a life-threatening emergency are justified, but if a local health-care system exists there should still be follow-up with the nearest local provider. Health-care professionals contemplating an international clinical experience should also consult the Humanitarian Aid Workers section in Chapter 8 for a discussion of emotional and physical fitness to participate, preparation, and after-care issues. The Primary Health-Care Provider {#cesec509} -------------------------------- Primary health-care providers play a crucial role in several aspects of medical tourism. For un- or under-insured patients who cannot afford their prescribed course of treatment, the primary care provider may be asked to provide counsel regarding international treatment options, assist with vetting available options, optimize patient status prior to travel, or coordinate care on return. Each provider will need to assess individually his or her ability to address travel health issues or refer to a travel medicine provider. Clinicians who care for immigrant populations should also be aware that the majority of health-seeking travelers in 2004 were current U.S. citizens born outside the United States, followed by non-U.S. citizens. Health-care needs, such as dentistry, are often included in visits home, due to familiarity with care in the country of origin, the high cost of health care in the United States, and lack of insurance coverage in these populations. There are also recent reports that patients on transplant waiting lists may also travel abroad for the procedure and return to the developed country of residence for continued care, often requiring immediate hospitalization and intense initial management with little documentation. Options for dialysis care are also increasing in developing areas; thus patients requiring this level of care may return home for visits and obtain local care. Acute hepatitis B infections have been diagnosed in patients returning to developed countries from both scenarios. Clinicians providing care to immigrant populations should consider routinely inquiring about future or recent travel home to visit friends and relatives, whether health care will be sought or occurred during travel and advise accordingly (see the VFR section in Chapter 8). Travel Medicine Providers {#cesec510} ------------------------- Patients who plan to seek medical care abroad may not divulge this activity during the consultation. The desire for anonymity may be a reason for seeking procedures, such as cosmetic surgery or sex-change operations, abroad. As previously mentioned, cost is often an issue, and patients may be uncomfortable self-disclosing. Clinicians may find that routine discussion of hepatitis B vaccination with all patients in the context of risk due to tattoo, sex, emergency medical care, and invasive procedures offers an environment for patients to initiate further discussion. Health-care providers may also find that the medical industry and associated resources that are rapidly expanding in the developing world related to medical tourism intersect directly with the medical care options for patients with pre-existing illness who travel, emergency care for travelers, and health-care options for expatriates (see the [Obtaining Health Care Abroad for the Ill Traveler](#subchapter44){ref-type="sec"} section later in this chapter). Additional Resources for Medical Tourism and Accredidation {#cesec511} ---------------------------------------------------------- •Joint Commission International (jointcommissioninternational.org/)•Trent International Accreditation Scheme (trentaccreditationscheme.org/)•Australian Council for Healthcare Standards International ([www.achs.org.au/ACHSI/](http://www.achs.org.au/ACHSI/){#interref112})•Canadian Council on Health Services ([www.cchsa.ca/](http://www.cchsa.ca/){#interref113})•International Society of Plastic Surgery also certifies international surgeons who meet U.S. standards ([www.isaps.org](http://www.isaps.org){#interref114}) *PERSPECTIVES:* COUNTERFEIT DRUGS {#subchapter40} ================================= Green Michael D. GENERAL INFORMATION {#cesec512} ------------------- Counterfeit and substandard drugs are an international problem contributing to morbidity, mortality, toxicity, and drug resistance. A counterfeit medicine is a compound that is not made by an authorized manufacturer but is presented to the consumer as if it were. Overall, global estimates of drug counterfeiting are somewhat ambiguous, depending on geographic region, but proportions range from 1% of sales in developed countries to \>10% in developing countries. In specific regions in Africa, Asia, and Latin America, chances of purchasing a counterfeit drug may be higher than 30%. Although the availability of fake drugs is a worldwide occurrence, developing countries lacking adequate resources to effectively monitor and maintain good drug quality are most susceptible. These conditions allow for the proliferation of counterfeit as well as substandard medicines. Since counterfeit drugs are not made by the legitimate manufacturer and are produced under unlawful circumstances, contaminants or lack of proper ingredients may result in serious harm to one\'s health. For example, the active pharmaceutical ingredient (API) may be completely lacking, present in small quantities, or substituted by another less-effective compound. In addition, the wrong inactive ingredients (excipients) can contribute to poor drug dissolution and bioavailability. As a result, a patient may not respond to treatment, or they may exhibit adverse reactions to unknown substituted ingredients. Prior to international departure, travel clinics should alert travelers of the dangers of counterfeit and substandard drugs and provide suggestions on how to avoid them. Listed are main points of which to be aware. HOW TO AVOID COUNTERFEIT DRUGS WHEN TRAVELING {#cesec513} --------------------------------------------- The best way to avoid counterfeit drugs is to reduce the need to purchase medications abroad. Anticipated amounts of medications for chronic conditions such as hypertension, sinusitis, arthritis, hay fever, etc., medications for gastroenteritis (travelers\' diarrhea), and prophylactic medications for infectious diseases such as malaria (depending on the destinations) should all be purchased at home prior to traveling. ### To do before you leave: {#cesec514} ¢**Make sure you have all your vaccinations before embarking**. Immunizations provide the best protection against many serious diseases.¢**Purchase in advance, in your home country, all the medicines you will need for the entire trip**. Prescriptions from your doctor usually cannot be filled overseas, and over-the-counter medicines may not be available in many foreign countries. Checked baggage can get lost; therefore pack as much as possible in a carry-on bag. Bring along extra in case of travel delays.¢**Make sure your medicines are in their original containers**. If the drug is a prescription, make sure your name and dosage requirements are on the container.¢**Bring your "Patient Prescription Information" sheet**. This sheet provides information on common generic and brand names, usage, side effects, precautions, and drug interactions. ### What to do if you run out and require additional medications: {#cesec515} ¢**Purchase medicines from a legitimate pharmacy**. In some places, it is difficult to know if a pharmacy has a genuine license. Your chances of receiving a counterfeit drug are less if you avoid buying from open markets, street vendors, or suspicious-looking pharmacies. Request a receipt when making the purchase. The U.S. Embassy may be able to assist you in finding a legitimate pharmacy in the area.¢**Do not buy medicines that are significantly cheaper than the typical price**. Although generics are usually less expensive, many counterfeited brand names are sold at prices significantly below the normal price for that particular brand.¢**Make sure the medicines you purchase are in their original packages or containers**. Many times medicines are sold to the pharmacy in bulk and the pharmacist will dispense the required amount of medicine into another container. If you receive medicines as loose tablets or capsules supplied in a plastic bag or envelope, ask the pharmacist to see the container from which it was originally dispensed. Record the brand, batch number, and expiration date. Sometimes a wary consumer will prompt the seller into making sure he or she supplies you with quality medicine.¢**Be familiar with your medications**. The size, shape, color, and taste of counterfeit medicines may be different from the authentic. Discoloration, splits, cracks, spots, and stickiness of the tablets or capsules are indications of a possible counterfeit. Keep examples of authentic medications available for comparison if you purchase the same brand.¢**Be familiar with the packaging**. Different color inks, poor-quality print or packaging material, and misspelled words are clues to counterfeit material. Also, keep an example of packaging for comparison. Observe the expiration date to make sure the medicine has not expired and the package contains the drug insert. USEFUL WEBSITES ON COUNTERFEITS {#cesec516} ------------------------------- ¢General Information:○CDC:□[wwwn.cdc.gov/travel/contentCounterfeitDrugs.aspx](http://wwwn.cdc.gov/travel/contentCounterfeitDrugs.aspx){#interref117}□[www.cdc.gov/malaria/travel/counterfeit_drugs.htm](http://www.cdc.gov/malaria/travel/counterfeit_drugs.htm){#interref118}○World Health Organization: who.int/mediacentre/factsheets/fs275/en¢U.S. Food and Drug Administration: [www.fda.gov/counterfeit/](http://www.fda.gov/counterfeit/){#interref119}¢U.S. Pharmacopeia: [www.usp.org/worldwide/dqi/drugQuality.html](http://www.usp.org/worldwide/dqi/drugQuality.html){#interref120}¢Warnings and alerts: [www.safemedicines.org/in_the_news/drug_alerts.php](http://www.safemedicines.org/in_the_news/drug_alerts.php){#interref121}¢What can you pack in your luggage (for travelers with disabilities and medical conditions):○Transportation Security Administration: [www.tsa.gov/travelers/airtravel/specialneeds/editorial_1059.shtm](http://www.tsa.gov/travelers/airtravel/specialneeds/editorial_1059.shtm){#interref122}¢What can you bring back:○U.S. Customs and Border Protection: [www.cbp.gov/xp/cgov/travel/clearing/restricted/medication_drugs.xml](http://www.cbp.gov/xp/cgov/travel/clearing/restricted/medication_drugs.xml){#interref123}¢Reporting counterfeit cases: [www.who.int/medicines/services/counterfeit/report/en/](http://www.who.int/medicines/services/counterfeit/report/en/){#interref124} DRUG--VACCINE AND DRUG--DRUG INTERACTIONS {#subchapter41} ========================================= Barnett Elizabeth D. The pre-travel travel medicine visit potentially exposes people to a number of different vaccines, prophylactic medications, and therapeutic drugs. In addition, the traveler may already be taking one or more medications on a regular basis. Travel medicine practitioners need to think about the possible interactions between all these products. Although a comprehensive list of interactions is beyond the scope of this section, some of the more significant interactions of commonly used vaccines and medications are discussed here. Interactions between Travel Vaccines and Drugs {#cesec517} ---------------------------------------------- ### Oral Typhoid Vaccine {#cesec518} There is a concern that antibiotics or anti-malarials with antibiotic activity should not be taken at the same time as the oral typhoid vaccine (a live-bacteria vaccine) as they may be active against the vaccine strain and prevent an adequate immune response to the vaccine. These issues should be addressed separately. Sulfonamides and antibiotics taken orally should not be taken at the same time as oral typhoid vaccine. Parenteral typhoid vaccine is a more appropriate choice for individuals taking antibiotics. The current mefloquine (Lariam) product insert recommends vaccinations with attenuated live bacteria be completed at least 3 days before the first dose of Lariam. However, one study in humans failed to show any decrease in immunogenicity when the vaccine was given to those on mefloquine prophylaxis. Mefloquine can be given concurrently with the oral typhoid vaccine. Although the antibody response to oral typhoid vaccine was reduced by higher dose proguanil in one study examining the effects of concomitant use of oral typhoid vaccine and chloroquine, mefloquine, and proguanil, a second study using approved prophylaxis doses of atovaquone--proguanil showed no decrease in immunogenicity. Atovaquone--proguanil at prophylaxis doses can be given concurrently with the oral typhoid vaccine. However, since the oral typhoid vaccine should be completed 14 days prior to traveling, there should be no opportunity for concomitant administration with atovaquone/proguanil in most travelers. This same study also showed no decrease in immunogenicity with chloroquine. Chloroquine can be given concurrently with the oral typhoid vaccine. Doxycycline is an antibiotic with both antibacterial and antimalarial activity. Thus it should not be given concurrently with oral typhoid vaccine. However, since the oral typhoid vaccine should be completed 14 days prior to traveling, there should be no opportunity for interaction with doxycycline in most travelers. ### Rabies Vaccine {#cesec519} Concomitant use of chloroquine may reduce antibody response to intradermal rabies vaccine administered for pre-exposure prophylaxis. The intramuscular route should be used for persons taking chloroquine concurrently (the intradermal route is currently not approved for use in the United States); ideally, the rabies pre-exposure prophylaxis series should be completed before beginning chloroquine. Corticosteroids and other immunosuppressive agents may interfere with response to rabies immunization; when these are used concurrently with rabies vaccine for postexposure prophylaxis, testing should be done to ensure adequate antibody response. Interactions between Antimalarials and Other Drugs {#cesec520} -------------------------------------------------- ### Mefloquine {#cesec521} Mefloquine may interact with several categories of drugs, including other antimalarials, drugs that alter cardiac conduction, and anticonvulsants. Although the antimalarial halofantrine is not available in the United States, potentially fatal prolongation of the QTc interval of the electrocardiogram may occur if halofantrine is given after mefloquine. Halofantrine should not be given with or after mefloquine. If halofantrine is given for treatment of malaria, mefloquine for prophylaxis should not be resumed until at least 12 hours after the last halofantrine dose. However, no conclusive data are available with regard to coadministration of mefloquine and other drugs that may theoretically have an impact on cardiac conduction. These include anti-arrhythmic or beta-blocking agents, calcium-channel blockers, antihistamines, H1-blocking agents, tricyclic antidepressant, or phenothiazines. Use of these drugs along with mefloquine should be avoided, if possible. Mefloquine used with the anticonvulsants valproic acid, carbamazepine, phenobarbital, or phenytoin may lower anticonvulsant plasma levels, thus lowering seizure threshold. Monitoring anticonvulsant levels would be appropriate in persons for whom mefloquine must be used concomitantly with these drugs. ### Chloroquine {#cesec522} Chloroquine absorption may be reduced by antacids or kaolin; at least 4 hours should elapse between doses of these medications. Concomitant use of cimetidine and chloroquine should be avoided, as cimetidine can inhibit the metabolism of chloroquine and may increase drug levels. Chloroquine inhibits bioavailability of ampicillin; 2 hours should elapse between doses. ### Atovaquone--Proguanil {#cesec523} Tetracycline, rifampin, and rifabutin may reduce plasma concentrations of atovaquone and should not be used concurrently with atovaquone--proguanil. Metaclopramide may reduce bioavailability of atovaquone; unless no other antiemetics are available, this antiemetic should not be used for treatment of the vomiting that may accompany use of atovaquone at treatment doses. Atovaquone--proguanil should not be used with other proguanil-containing medications. Patients on anticoagulants may need to reduce their anticoagulant doses or more closely monitor their prothrombin time while taking atovaquone proguanil. ### Doxycycline {#cesec524} Doxycycline use theoretically may lead to decreased efficacy of oral contraceptives, although it has been difficult to quantify this effect in a way that is useful for counseling travelers. Changes in hormone levels in women taking oral contraceptives concurrently with doxycycline have not been demonstrated. Phenytoin, carbamazepine, and barbiturates may decrease the half-life of doxycycline. Patients on anticoagulants may need to reduce their anticoagulant doses while taking doxycycline because of its ability to depress plasma prothrombin activity. Absorption of tetracyclines may be impaired by bismuth subsalicyclate, iron-containing preparations, and antacids containing calcium, magnesium, or aluminum; these preparations should not be taken within 1--3 hours of doxycycline. Doxycycline absorption is not markedly affected by food or milk taken concurrently. Doxycyline may interfere with the bactericidal activity of penicillin, and these drugs should not be taken concurrently. Interactions with Antidiarrheal Drugs {#cesec525} ------------------------------------- ### Fluoroquinolones {#cesec526} Increase in the international normalized ratio (INR) has been reported when levofloxacin and warfarin are used concurrently. Concurrent administration of ciprofloxacin and magnesium or aluminum hydroxide containing antacids may reduce bioavailability of ciprofloxacin significantly. Ciprofloxacin decreases clearance of theophylline and caffeine; theophylline levels should be monitored when ciprofloxacin is used concurrently. Ciprofloxacin should not be used with tazanidine. ### Azithromycin {#cesec527} Close monitoring for side effects of azithromycin is recommended when azithromycin is used with nelfinavir. Increased anticoagulant effects have been noted when azithromycin is used with warfarin; monitoring of prothrombin time is recommended for such individuals. ### Rifaximin {#cesec528} No clinically significant drug interactions have been reported to date with rifaximin. Although the drug induces cytochrome P450 3A4 (CYP3A4), studies of concurrent administration of rifaximin with midazolam and with a single dose of the oral contraceptive ethinyl etradiol and norestironate did not show changes in the pharmacokinetics of these drugs. Interactions with Drugs Used for Travel to High Altitude {#cesec529} -------------------------------------------------------- ### Acetazolamide {#cesec530} Acetazolamide produces alkaline urine that can increase the rate of excretion of barbiturates and salicylates and may potentiate salicylate toxicity. Decreased excretion of dextroamphetamine, anticholinergics, mecamylamine, ephedrine, mexiletine, or quinide may also occur. Hypokalemia caused by corticosteroids may be potentiated by concurrent use of acetazolamide. ### Dexamethasone {#cesec531} Dexamethasone interacts with multiple classes of drugs. Use of this drug for treatment of altitude illness may, however, be lifesaving. Interactions may occur with the following drugs and drug classes: macrolide antibiotics, anticholinesterases, anticoagulants, hypoglycemic agents, isoniazid, digitalis preparations, oral contraceptives, and phenytoin. *PERSPECTIVES:* PPD TESTING OF TRAVELERS {#subchapter42} ======================================== LoBue Philip Screening travelers for asymptomatic tuberculosis (TB) infections should only be carried out among travelers who will be at significant risk of acquiring TB (see the TB section in Chapter 5). Screening with a tuberculin skin test (TST) in a very low-risk population may result in a false-positive test, leading to unnecessary further screening or unnecessary therapeutic treatment. Using even highly sensitive and specific tests in very low-prevalence populations will produce more false positives than true positives. Therefore, the TST should be considered only for travelers anticipating an extended stay over a period of years in a country with a high risk of TB or for those who could be expected to come in contact routinely with hospital, prison, or homeless shelter populations. The general recommendation is that persons at low risk for TB, which includes the vast majority of travelers, do not need to be screened before or after travel. For travelers who anticipate a long stay or contact with a high-risk population, careful pre-travel screening should be carried out. The two-step TST is recommended in this population, for the following reasons:¢The use of two-step testing can reduce the number of positive TSTs that would otherwise be misclassified as recent skin test conversions during future periodic screenings.¢Certain persons who were infected with *Mycobacterium tuberculosis* years earlier exhibit waning delayed-type hypersensitivity to tuberculin. When they are skin tested years after infection, they might have a false-negative TST result (even though they are truly infected). However, this first skin test years after the infection might stimulate the ability to react to subsequent tests, resulting in a "booster" reaction. When the test is repeated, the reaction might be misinterpreted as a new infection (recent conversion) rather than a boosted reaction.¢For two-step testing, persons whose baseline TSTs yield a negative result are retested 1--3 weeks after the initial test. If the second test result is negative, they are considered not infected. If the second test result is positive, they are classified as having had previous TB infection.¢Two-step testing should be considered for the baseline testing of persons who report no history of a recent TST and who will receive repeated TSTs as part of an ongoing monitoring of whether they have been exposed to TB.¢If the two-step TST result is negative, the traveler should have a repeat TST 8--10 weeks after returning from their trip, or as part of a periodic screening examination for those who remain at high risk. Two-step testing is particularly important for travelers who will have potential prolonged TB exposure; it is particularly important among those going to areas where drug resistance is very high. Two-step testing prior to travel will detect boosting and potentially prevent "false conversions"-positive TST results that appear to be indicative of infection acquired during travel, but which are really the result of previous TB infection. This is particularly important if the traveler is going to a country where XDR TB is rampant. It would be critical to know whether the person\'s skin test had actually been positive before the travel. Persons having repeat TSTs must be tested with the same commercial antigen, as switching antigens can also lead to false TST conversions. An alternative to two-step TST is a single FDA-approved interferon-gamma release assay (IGRA), such as the QuantiFERON TB test (Gold or Gold In-Tube versions). IGRAs are about equally specific as TST in non-BCG-vaccinated populations and much more specific in BCG-vaccinated populations. For a traveler whose time before departure is short, a single-step TST would be an acceptable alternative if there were insufficient time for the two-step TST and the IGRA were not available. In general it is best not to mix tests. There is about 15% discordance between TST and IGRA, usually with the TST positive and the IGRA negative. There are multiple reasons for the discordance, and in any individual it is often difficult to be confident about the reason for discordance. However, if the health-care provider decides to mix tests, it is better to go from TST to IGRA than the other way around, because the likelihood of having a discordant result, with the TST negative and the IGRA positive, is much lower. Such discordant results may become unavoidable as more medical establishments switch from TSTs to IGRAs. The use of TST among those visiting friends and relatives in TB-endemic areas should take into account the high rate of TST positivity in this population. In a study among 53,000 adults in Tennessee, the prevalence of a positive TST among the foreign born was 10 times that of the U.S. born (34.2% vs. 3.2%). Confirming TST status prior to travel would prevent the conclusion that a positive TST after travel was due to recent conversion. TRAVEL HEALTH KITS {#subchapter43} ================== Whatley Amanda D. Barbeau Deborah Nicolls The purpose of packing a travel health jit is to ensure travelers have supplies they need to---•manage pre-existing medical conditions and treat any exacerbations of these conditions,•prevent illness related to traveling, and•take care of minor health problems as they occur. Traveling with Medications {#cesec532} -------------------------- When medications are necessary for travel, it is important to remember the following:•**Original containers:** All medications should be carried in their original containers with clear labels, so the contents are easily identified. Although many travelers like placing medications into small containers or packing them in the daily-dose containers, officials at ports of entry may require proper identification of medications.•**Prescriptions:** Travelers should carry copies of all prescriptions, including their generic names.•**Physician notes:** For controlled substances and injectable medications, travelers are advised to carry a note from the prescribing physician on letterhead stationery.•**Restricted medications:** Travelers should be aware that certain medications are not permitted in certain countries. If there is a question about these restrictions, particularly with controlled substances, travelers are recommended to contact the embassy or consulate of the destination country.•**Availability:** A travel health kit is useful only when it is available. It should be carried with the traveler at all times (e.g., in a carry-on bag). Due to airline security rules, sharp objects and some liquids and gels must remain in checked luggage. Pre-Existing Medical Condition Supplies {#cesec533} --------------------------------------- Travelers with pre-existing medical conditions are advised to carry enough medication for the duration of their trip and an extra supply, in case the trip is extended for any reason. If additional supplies or medications are needed for the management of exacerbations of existing medical conditions, these should be carried as well. The health-care provider managing a traveler\'s pre-existing medical conditions should be consulted for the best plan of action (see the section Traveling with Chronic Medical Illnesses in Chapter 8). Persons with pre-existing conditions, such as diabetes or allergies to envenomations or medications, should consider wearing an alert bracelet and making sure this information is on a card in their wallet and with their other travel documents. General Travel Health Kit Supplies {#cesec534} ---------------------------------- A variety of health kits is available commercially and may even be purchased over the Internet (see below); however, similar kits can be assembled at home, often at lower cost. The specific contents of the health kit are based on destination, duration of travel, type of travel, and the traveler\'s pre-existing medical conditions. Although this is not a comprehensive list, basic items that should be considered are listed below. See Chapters 7 and 8 for additional suggestions that may be useful in planning the contents of a kit for travelers with specific needs. ### Medications {#cesec535} •Destination-related, if applicable:○Antimalarial medications○Medication to prevent or treat high-altitude illness•Pain or fever (one or more of the following, or an alternative):○Acetaminophen○Aspirin○Ibuprofen•Stomach upset or diarrhea:○Over-the-counter antidiarrheal medication (such as loperamide or bismuth subsalicylate)○Antibiotic for self-treatment of moderate to severe diarrhea○Oral rehydration solution packets○Mild laxative○Antacid•Items to treat throat and respiratory symptoms:○Antihistamine○Decongestant, alone or in combination with antihistamine○Cough suppressant/expectorant○Throat lozenges•Anti-motion sickness medication.•Epinephrine auto-injector (such as an EpiPen), especially if history of severe allergic reaction. Smaller-dose packages are available for children.•Any medications, prescription or over the counter, taken on a regular basis at home. ### Basic First Aid {#cesec536} •Disposable gloves (at least two pairs)•Adhesive bandages, multiple sizes•Gauze•Adhesive tape•Elastic bandage wrap for sprains and strains•Antiseptic•Cotton swabs•Tweezers[\*](#fn3){ref-type="fn"}•Scissors[\*](#fn3){ref-type="fn"}•Antifungal and antibacterial ointments or creams•1% hydrocortisone cream•Anti-itch gel or cream for insect bites and stings•Aloe gel for sunburns•Moleskin or molefoam for blisters•Digital thermometer•Saline eye drops•First-aid quick reference card ### Other Important Items {#cesec537} •Insect repellent•Sunscreen (SPF 15 or greater)•Antibacterial hand wipes or an alcohol-based hand sanitizer containing at least 60% alcohol•Useful items in certain circumstances:○Extra pair of contacts or prescription glasses, or both, for people who wear corrective lenses○Mild sedative (such as zolpidem), other sleep aid, or anti-anxiety medication○Latex condoms○Water purification tablets○Commercial suture/syringe kits to be used by a local health-care provider. (These items will also require a letter from the prescribing physician on letterhead stationery.) ### Contact Card {#cesec538} It is also important for travelers to locate and record important contact information, in case it is needed during their trip. Often this information is needed quickly; having a contact card with the following items will help save time in these urgent situations. Items to include on a contact card should be the address and phone numbers of the following:•Family member or close contact still in the United States•Health-care provider at home•Area hospitals or clinics•U.S. Embassy or Consulate in the destination country or countries See the next section in this chapter, [Obtaining Health Care Abroad for the Ill Traveler](#subchapter44){ref-type="sec"}, for information about how to locate local health care and embassy/consulate contacts. Commercial Medical Kits {#cesec539} ----------------------- Commercial medical kits are available for a wide range of circumstances, from basic first aid to advanced emergency life support. Many pharmacy, grocery, retail, and outdoor sporting goods stores sell their own basic first-aid kits. Travelers who choose to purchase a health kit rather than assemble their own should be certain to review the contents of the kit carefully to ensure that it has everything needed; additional items may be necessary. For more adventurous travelers, a number of companies produce advanced medical kits and will even customize kits based on specific travel needs. In addition, specialty kits are available for managing diabetes, dealing with dental emergencies, and handling aquatic environments. Below is a list of websites supplying a wide range of medical kits. There are many suppliers, and this list is not meant to be all-inclusive.•American Red Cross: [www.redcrossstore.org](http://www.redcrossstore.org){#interref126}•Adventure Medical Kits: [www.adventuremedicalkits.com](http://www.adventuremedicalkits.com){#interref127}•Chinook Medical Gear: [www.chinookmed.com](http://www.chinookmed.com){#interref128}•Travel Medicine, Inc.: [www.travmed.com](http://www.travmed.com){#interref129}•Wilderness Medicine Outfitters: [www.wildernessmedicine.com](http://www.wildernessmedicine.com){#interref130} OBTAINING HEALTH CARE ABROAD FOR THE ILL TRAVELER {#subchapter44} ================================================= Sommers Theresa Brunette Gary W. An important aspect of preparing for a trip abroad is to consider the possibility of becoming sick or injured during travel. The following resources and information will be useful to travelers, should they require medical assistance abroad. Traveling While Ill {#cesec540} ------------------- Health-care providers should advise their patients about the possible need to avoid traveling if they become ill during their trip. Those with certain health conditions may need to postpone their travel arrangements, including air and public ground transportation. In general, travelers who are ill with a communicable disease that is spread easily to other people should discuss the need for rescheduling travel with their provider. Travelers should be aware that some airlines check for visibly sick passengers in the waiting area and during boarding. If a waiting passenger looks visibly ill, the airline may prohibit that person from getting on the airplane. Locating a Health-Care Provider {#cesec541} ------------------------------- Several resources are available to American citizens who require medical attention during their travels. The following resources can assist travelers in finding adequate care:•The U.S. Department of State:○A U.S. consular officer can assist in locating appropriate medical services, as well as in notifying friends, family, or employer of an emergency.○For more information, see <http://travel.state.gov/travel/tips/brochures/brochures_1215.html>.•The International Society of Travel Medicine (ISTM):○ISTM maintains a directory of health-care professionals with expertise in travel medicine in almost 50 countries worldwide.○To access the directory, see [www.istm.org](http://www.istm.org){#interref132}.•The American Society of Tropical Medicine and Hygiene (ASTMH):○ASTMH maintains a worldwide directory of providers specializing in tropical medicine, medical parasitology, and travelers\' health.○To access the directory, see [www.astmh.org](http://www.astmh.org){#interref133}.•International Association for Medical Assistance to Travelers (IAMAT):○IAMAT maintains an international network of physicians, hospitals, and clinics who have agreed to treat IAMAT members in need of medical care while abroad.○Membership is free, although a donation to support IAMAT efforts is suggested. Members receive a directory of participating physicians and medical centers and have access to a variety of travel-related informational brochures.○For more information, see [www.iamat.org](http://www.iamat.org){#interref134}.•Travel Health Online:○This resource maintains a list of travel medicine providers worldwide. Information is obtained from a variety of sources, so the quality of services and the expertise of the providers cannot be guaranteed.○For more information, see <https://www.tripprep.com>. Travelers may also get information about local health care from embassies and consulates of other countries, hotel doctors, credit card companies, and multinational corporations, which may offer health-care services for their employees. In addition, travelers who obtain evacuation insurance before travel will have access to a 24-hour hotline for help in any medical emergency. Accreditation of International Health-Care Facilities {#cesec542} ----------------------------------------------------- The quality of health care from foreign medical centers can be variable, particularly in developing countries. To ensure a higher quality of care abroad, Joint Commission International attempts to continuously improve the safety and quality of care in the international community through the provision of education and consultation services and international accreditation. A list of accredited international health-care facilities is available at the Joint Commission International website ([www.jointcommissioninternational.org](http://www.jointcommissioninternational.org){#interref136}). Drugs/Pharmaceuticals Abroad {#cesec543} ---------------------------- The quality of drugs and medical products abroad cannot be guaranteed, as they may not meet U.S. standards or could be counterfeit (see [Perspectives: Counterfeit Drugs](#subchapter40){ref-type="sec"} earlier in this chapter). Travelers are advised to---•Bring with them all the drugs and medicines that they think they will need, including pain relievers, antidiarrheal medication, and, if applicable, antimalarials.•Exercise caution when buying medications (especially those that do not require a prescription). In many developing countries, virtually any drug can be purchased without prescription.•Travelers who may require an injection(s) abroad should bring their own injection supplies (see the [Travel Health Kits](#subchapter43){ref-type="sec"} section earlier in this chapter).•Travelers who do not have their own injection supplies yet require an injection should ask if the equipment is disposable and insist that a new needle and syringe be used. Emergency Care Abroad {#cesec544} --------------------- The quality and availability of proper emergency medical care abroad may be variable and, in situations requiring a blood transfusion, the safety of blood products often cannot be guaranteed.•Not all countries have accurate, reliable, and systematic screening of blood donations for infectious agents, which increases the risk of transfusion-related transmission of disease.•The 2001--2002 WHO Global Database on Blood Safety report supports this view:○40 countries reported they did not test all donated blood for HIV, hepatitis B and C viruses, and syphilis.○39 countries reported that, due to unavailable testing supplies, blood was released for clinical use without testing for transfusion-transmissible infections.•Due to this increased risk, travelers in developing countries should only receive a blood transfusion in life-and-death situations for which there may be no other options.•When a situation requires blood transfusion, travelers should make every effort to ensure that the blood has been screened for transmissible diseases, including HIV.•All travelers should consider being immunized against hepatitis B virus before their trip, especially---○Those who travel frequently to developing countries○Travelers whose itinerary indicate spending a prolonged period of time in developing countries○Travelers whose activities put them at higher risk for serious injury (e.g., adventure travel).•There are no medical indications for travelers to take blood with them from their home countries.•The limited storage period of blood and the need for special equipment negate the feasibility of independent blood banking for individual travelers or small groups. The international shipment of blood for transfusion is practical only when handled by agreement between two responsible organizations, such as national blood transfusion services. This mechanism is not useful for the emergency needs of individual travelers and should not be attempted by private travelers or organizations not operating recognized blood programs. TRAVEL INSURANCE AND EVACUATION INSURANCE {#subchapter45} ========================================= Sommers Theresa Brunette Gary W. It is important for travelers to consider the financial consequences of a severe illness or injury abroad. A growing number of people do not have health insurance at home. Those who do need to check their policies to determine if their care abroad will be covered and what limitations may apply. Those who have adequate health insurance may not be covered for medical evacuation from a resource-poor area to a hospital where definitive care can be obtained. Even if they have a policy that would reimburse evacuation costs, the health insurance company may not have the resources to help organize the evacuation. Evacuation-only policies are available to fill this gap. Evacuation by air ambulance can cost \$50,000 to \$100,000 and must be paid in advance by those who do not have insurance. Paying for Health Services Abroad {#cesec545} --------------------------------- Travelers who receive medical care in other countries will usually be required to pay in cash or with a credit card at the point of service, even if they have insurance coverage abroad. This could result in a large out-of-pocket expenditure of perhaps thousands of dollars for medical care.•Travelers with health insurance coverage should be sure to obtain copies of all bills and receipts from overseas medical care.•The U.S. consular office can assist travelers who are U.S. citizens with transferring funds from the United States.•In extreme circumstances, the U.S. consular office may be able to approve small government loans until private funds are available.•Medical evacuation insurance may only cover the cost to the nearest destination where definitive care can be obtained. Some policies will cover eventual repatriation to one\'s home country. The traveler should be sure to understand what coverage is purchased. Health Insurance Abroad {#cesec546} ----------------------- •Some health insurance carriers in the United States may provide coverage for emergencies that occur while traveling.•The first step for travelers is to examine their present coverage and planned itinerary. Determine exactly which medical services will be covered abroad and what supplemental insurance you will need. Things to look for include---○Exclusions for treatment of exacerbations of pre-existing medical conditions○The company\'s policy for "out-of-network" services○Coverage for complications of pregnancy○Exclusions for high-risk activities such as skydiving, scuba diving, and mountain climbing○Exclusions regarding psychiatric emergencies or injuries related to terrorist attacks or acts of war○Whether pre-authorization is needed for treatment, hospital admission, or other services○Whether a second opinion is required before obtaining emergency treatment•Medicare and Medicaid will not cover services outside the United States, except in very limited circumstances. Travel Health and Medical Evacuation Insurance {#cesec547} ---------------------------------------------- Travelers need to evaluate their existing health insurance policies to see whether they already have adequate coverage. Short-term supplemental policies that cover health-care costs on a trip can be purchased. Evacuation coverage can be sold separately or in conjunction with overseas health insurance. Evacuation companies often have better resources and experience in some parts of the world than others. Travelers may want to check with them about their resources in a given area before making a purchase. In general, travelers should purchase a policy that provides the following:•Arrangements with hospitals to guarantee payments directly. Travelers may want to check on this possibility for the planned itinerary.•Assistance via a 24-hour physician-backed support center. This is critical if the traveler is going to pay for evacuation insurance.•Emergency medical transport, including repatriation. Medical evacuation can be costly, ranging from a few thousand dollars to over \$100,000. While travel health and medical evacuation insurance is a consideration for all travelers, it is particularly important for travelers who---•Will be outside the United States for an extended period of time.•Have underlying illnesses. These travelers should make certain that complications of the underlying condition will be covered by the chosen policy.•Participate in activities involving greater risk for injury. Finding a Travel Health and Medical Evacuation Insurance Provider {#cesec548} ----------------------------------------------------------------- The following list, while not all-inclusive, gives a sample of resources for travelers seeking to purchase travel health and medical evacuation insurance:•U.S. Department of State ([www.travel.state.gov](http://www.travel.state.gov){#interref139})•International SOS ([www.internationalsos.com](http://www.internationalsos.com){#interref140})•MEDEX ([www.medexassist.com](http://www.medexassist.com){#interref141})•International Association for Medical Assistance to Travelers ([www.iamat.org](http://www.iamat.org){#interref142}) Special Considerations for Travelers with Underlying Medical Conditions {#cesec549} ----------------------------------------------------------------------- Travelers with underlying medical conditions may want to take extra precautions in preparing for travel.•Travelers should choose a medical assistance company that allows customers to store their medical history before departure, so it can be accessed from anywhere in the world, if needed.•Travelers should carry a letter from their physician listing underlying medical conditions and all current medications (including their generic names).•If possible, travelers may want to carry with them the name of their medical condition and medications written in the local language(s) of the areas they plan to visit. Special Considerations for Medicare/Medicaid Beneficiaries {#cesec550} ---------------------------------------------------------- •The Social Security Medicare program does not provide coverage for medical costs outside the United States, except under very limited circumstances.•Medicare beneficiaries can purchase supplemental travel health insurance to cover medical expenses outside of the United States.•Some Medigap plans available to people enrolled in the original Medicare plan provide limited coverage for emergency care abroad.•As with all travelers, Medicare beneficiaries should examine their present coverage carefully to know exactly what will be covered abroad and supplement with additional travel health insurance as appropriate. MENTAL HEALTH AND TRAVEL {#subchapter46} ======================== Balaban Victor Description {#cesec551} ----------- Travel is undertaken for a number of reasons, such as adventure, pleasure, business, or personal growth. While most travelers complete their journeys with a manageable amount of stress, foreign travel can produce a wide range of psychiatric, behavioral, and neurologic issues in travelers. Any journey can produce challenges, but longer journeys to more remote and strange environments can increase the psychological stresses for travelers. Risk Factors {#cesec552} ------------ Risk factors that have been identified for developing psychiatric and neurologic problems during and after travel include---•Pre-existing psychiatric issues: Stress can trigger or exacerbate psychiatric reactions in travelers with pre-existing psychiatric or behavioral conditions.•Side effects of mefloquine or other drugs:○People with underlying psychiatric disorders should not receive the antimalarial medication mefloquine. The neuropsychiatric side effects associated with mefloquine may become pronounced in these patients.○The neuropsychiatric side effects associated with mefloquine may also be compounded when administered concurrently with the antiretroviral medication efavirenz, which also carries the risk of neurologic toxicity.○Elderly travelers and travelers with memory or cognitive deficits may be more prone to develop delirium in flight, particularly when combined with dehydration, alcohol, or the use of sleep aids such as zolpidem.○The use of recreational drugs has also been found to be a trigger for psychiatric symptoms in travelers.•Stressful events during travel, such as loneliness, a feeling of loss of control, financial difficulties, or a traumatic event such as a serious illness or viewing disturbing sights, can have behavioral and psychosocial consequences for travelers. Occurrence and Risk for Travelers {#cesec553} --------------------------------- Data are limited on the prevalence of travel-related psychiatric and neurologic disorders:•A study of Israeli long-term travelers to Southeast Asia found that 11.3% reported psychiatric or neurologic symptoms during travel. The most common symptoms were sleep disturbances, fatigue, and dizziness. The majority of symptoms were short-lived and transient, but 2.5% of travelers reported severe psychiatric or neurologic symptoms, and 1.2% had symptoms lasting more than 2 months.•A study of urgent repatriation of British diplomats found that 41% of evacuations for nonphysical causes were due to depression.•Adventure travelers in extreme settings, such as polar expeditions, have been found to undergo psychiatric changes, including disturbed sleep, impaired cognitive ability, negative affect, and interpersonal tension and conflict; approximately 5% have been found to meet the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) or the International Classification of Diseases (ICD) criteria for psychiatric disorders (including substance-related and sleep disorders). Pre-Travel Mental Health Evaluation {#cesec554} ----------------------------------- Pre-travel screening should assess risk factors that might indicate a need for a traveler to be referred to a mental health professional for evaluation, especially prior to travel that is likely to be stressful. Factors that should be assessed include---•pre-existing psychiatric diagnoses, such as depression or anxiety disorders•history of psychosis in the traveler or a close family member•history of suicide attempts•evidence of depressed mood at assessment•exposure to prior traumas (e.g., disasters, severe injury, abuse, assault, etc.), particularly prior to travel that could involve re-exposure to traumatic events or situations•recent major life stressors or emotional strain•use of medications that may have psychiatric or neurologic side effects•pre-travel anxieties and phobias that are severe enough to interfere with a patients\' ability to function or to prepare for and enjoy their travel. Long-term travelers, aid workers, military personnel and other travelers likely to be exposed to stressful situations should be advised that the stresses and challenges they may face, particularly if combined with long hours of work, lack of sleep, or fatigue, can contribute to stress and anxiety. Long-term travelers should be encouraged to---•learn how to recognize signs of stress, exhaustion, depression, and anxiety in themselves;•take care of themselves physically by eating and exercising regularly; and•use their full allotment of time off or annual leave, particularly if they recognize signs of stress or exhaustion in themselves. During Travel {#cesec555} ------------- Severe mental illness occurring abroad can be extremely stressful for travelers, their families, and those who try to care for them. Acute psychosis leading to disruptive behavior can land a traveler in jail in a developing country. Inpatient psychiatric facilities may be either nonexistent or completely inadequate for a foreigner. It can be very difficult to repatriate a psychotic person until the symptoms have been brought under control with medication. Someone will most often have to accompany the person home. Many evacuation insurance plans specifically exclude psychiatric illness from their coverage. Post-Travel Mental Health Evaluation {#cesec556} ------------------------------------ Returning travelers may have experienced physical illnesses, personal difficulties, or traumas that could result in psychiatric reactions. Travel-related injuries and diseases that affect quality of life can also have profound and long-term psychiatric impacts. Even in the absence of trauma, some returning long-term travelers report experiencing "reverse culture shock" after their return, characterized by feelings of disorientation, unfamiliarity, and loss of confidence. Approximately 36% of aid workers report depression shortly after returning home, and as many as 60% of returned aid workers have reported feeling predominantly negative emotions on returning home, even though many reported that their time overseas was positive and fulfilling. Post-travel evaluations should assess---•Behavioral and psychiatric symptoms, including:○Experiences during or soon after travel, which have been painful, hard to reconcile or which still cause distress, anxiety, or avoidance.○Persistent sleep disturbance or unusual fatigue.○Excessive use of alcohol or drugs.○Behavioral or interpersonal difficulties in home, school, work, in friendships or relationships.•Somatic symptoms that can also be indications of distress, including:○Unexplained somatic symptoms, such as headaches, backaches, or abdominal pain; and somatic disorders such as fibromyalgia, chronic fatigue syndrome, temporomandibular disorder, and irritable bowel syndrome.○Rashes, itching, and skin diseases, such as psoriasis, atopic dermatitis, and urticaria, which can be exacerbated by stress. Clinicians should be aware that some travelers may be reluctant to acknowledge psychiatric symptoms or distress. For example, many cultures have stigmas associated with experiencing or disclosing behaviors associated with mental illness, as well as different culturally appropriate ways of expressing grief, pain, and loss. In addition, some travelers may fear being penalized or stigmatized at work if they have psychiatric diagnoses noted on their medical records. Regardless of the type or duration of travel, and whether or not travelers appear to meet criteria for a psychiatric diagnosis, returned travelers who are having difficulties functioning or who appear to be unduly depressed or distressed should be encouraged to seek appropriate treatment or counseling. Estimates of prevalence of antibody to hepatitis A virus (anti-HAV), a marker of previous HAV infection, are based on limited data and might not reflect current prevalence. In addition, anti-HAV prevalence might vary within countries by subpopulation and locality. As used on this map, the terms "high," "medium," and "low" endemicity reflect available evidence of how widespread HAV infection is within each country, rather than precise quantitative assessments. Measles vaccine alone is recommended for infants vaccinated before 12 months of age if it is available, otherwise MMR should be administered. Infants vaccinated before 12 months of age must be revaccinated on or after the first birthday with two doses of measles-containing vaccine separated by at least 28 days. MMRV is not licensed for children \<12 months of age. MMRV vaccine is licensed for children 12 months to 12 years of age and may be used in place of MMR vaccine if vaccination for measles, mumps, rubella and varicella are needed. ***Note:*** This recommendation refers to EPA-registered repellent products containing the active ingredient oil of lemon eucalyptus (or PMD). "Pure" oil of lemon eucalyptus (e.g., essential oil) is not the same product and has not received similar, validated testing for safety and efficacy, is not registered with EPA as an insect repellent, and is not covered by this recommendation. ***Note:*** Pack these items in checked baggage, since they may be considered sharp objects and confiscated by airport or airline security if packed in carry-on bags. [^1]: Booster dosing may be appropriate for certain populations, such as hemodialysis patients. [^2]: Adapted from General Recommendations on Immunization, MMWR, 2006. This table is not intended for determining the correct indications and dosage for the use of IG preparations. Unvaccinated people may not be fully protected against measles during the entire recommended interval, and additional doses of immune globulin (IG) or measles vaccine may be indicated after measles exposure. Concentrations of measles antibody in an IG preparation can vary by manufacturer\'s lot. For example, fourfold or greater variation in the amount of measles antibody titers has been demonstrated in different IG preparations. Rates of antibody clearance after receipt of an IG preparation can also vary. Recommended intervals are extrapolated from an estimated half-life of 30 days for passively acquired antibody and an observed interference with the immune response to measles vaccine for 5 months after a dose of 80 mg IgG/kg. [^3]: IG, immune globulin; IM, intramuscular; IV, intravenous. [^4]: Assumes a serum IgG concentration of 16 mg/mL. [^5]: Contains only antibody to respiratory syncytial virus. [^6]: DtaP, diphtheria and tetanus toxoids and acellular pertussis vaccine, pediatric (6 weeks through 6 years); MMR, measles, mumps and rubella; TIV, trivalent (inactivated) influenza vaccine; LAIV, live, attenuated (intranasal) influenza vaccine; Td, tetanus and reduced diphtheria toxoids, Tdap, tetanus toxoid, reduced diphtheria toxoid, and reduced acellular pertussis vaccine. [^7]: Combination vaccines are available. Use of licensed combination vaccines is generally preferred over separate injections of their equivalent component vaccines (CDC. Combination vaccines for childhood immunization: recommendations of the Advisory Committee on Immunization Practices (ACIP), the American Academy of Pediatrics (AAP), and the American Academy of Family Physicians (AAFP). MMWR Recomm Rep. 1999;48(RR-5):5). When administering combination vaccines, the minimum age for administration is the oldest age for any of the individual components; the minimum interval between doses is equal to the greatest interval of any of the individual components. [^8]: Combination vaccines containing the HepB component are available (HepB-Hib, DTaP-HepB-IPV, HepA-HepB). These vaccines should not be administered to infants younger than 6 weeks of age because of the other components (i.e., Hib, DTaP, IPV). HepA-HepB is not licensed for persons \<18 years of age in the United States. [^9]: HepB-3 should be administered at least 8 weeks after Hep B-2 and at least 16 weeks after Hep B-1; it should not be administered before age 24 weeks. [^10]: Calendar months. [^11]: The minimum recommended interval between DTaP-3 and DTaP-4 is 6 months. However, DTaP-4 need not be repeated if administered at least 4 months after DTaP-3. Adapted from [Table 1](#cetable1){ref-type="table"}, CDC. General recommendations on immunization. Recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Recomm Rep. 2006; 55(RR-15):1--48. [^12]: For Hib and PCV, children receiving the first dose of vaccine at ≥7 months of age require fewer doses to complete the series (see the current childhood and adolescent immunization schedule at [www.cdc.gov/vaccines/](http://www.cdc.gov/vaccines/){#interref2}). [^13]: If PRP-OMP (Pedvax-Hib, Merck Vaccine Division) was administered at 2 and 4 months of age, a dose at 6 months of age is not indicated. Adapted from [Table 1](#cetable1){ref-type="table"}, CDC. General recommendations on immunization. Recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Recomm Rep. 2006; 55(RR-15):1--48. [^14]: Combination MMR-varicella can be used for children 12 months through 12 years of age. Also see footnote [9](#cetablefn14){ref-type="table-fn"}. [^15]: The minimum interval from VAR-1 to VAR-2 for persons beginning the series at ≥13 years of age is 4 weeks. [^16]: Two doses of influenza vaccine are recommended only for children \<9 years of age who are receiving the vaccine for the first time and for certain incompletely vaccinated children. See reference [@bib5]. [^17]: Some experts recommend that a second dose of MPSV be given 3 years after the first dose for persons at increased risk for meningococcal disease. [^18]: A second dose of meningococcal vaccine is recommended for persons previously vaccinated with MPSV who remain at high risk for meningococcal disease. MCV is preferred when revaccinating persons 2--55 years of age (CDC. Prevention and control of meningococcal disease. Recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Recomm Rep. 2005;54(RR07);1--21.) Adapted from [Table 1](#cetable1){ref-type="table"}, CDC. General recommendations on immunization. Recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Recomm Rep. 2006; 55(RR-15):1--48. [^19]: Only one dose of Tdap is recommended. Subsequent doses should be given as Td. If vaccination to prevent tetanus and/or diphtheria disease is required for children 7--9 years of age, Td should be given (minimum age for Td is 7 years). For one brand of Tdap, the minimum age is 11 years. The preferred interval between Tdap and a previous dose of Td is 5 years, but Tdap may be administered earlier if pertussis immunity is needed. For management of a tetanus-prone wound, the minimum interval after a previous dose of any tetanus-containing vaccine is 5 years. [^20]: A second dose of PPV is recommended for persons at highest risk for serious pneumococcal infection and those who are likely to have a rapid decline in pneumococcal antibody concentration. (CDC. Prevention of pneumococcal disease. Recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Recomm Rep. 1997; 46(RR-8):1--24.) [^21]: HPV is approved only for females 9--26 years of age. HPV-3 should be administered at least 12 weeks after HPV-2 and at least 24 weeks after HPV-1. [^22]: The first dose of RV must be administered by 14 weeks and 6 days of age. The vaccine series should not be started at 15 weeks of age or older. The final dose in the series should be administered by age 8 months 0 days. If Rotarix rotavirus vaccine is administered at 2 and 4 months of age, a dose at 6 months of age is not indicated. [^23]: Herpes zoster vaccine is approved as a single dose for persons 60 years of age and older. [^24]: Oral typhoid vaccine is recommended to be administered 1 hour before a meal with a cold or lukewarm drink (temperature not to exceed body temperature) (i.e., 98.6° F (37° C)) on alternate days, for a total of 4 doses. [^25]: Yellow fever vaccine may be administered to children younger than 9 months of age in certain situations. (CDC. Yellow Fever Vaccine Recommendations of the Advisory Committee on Immunization Practices (ACIP), 2002. MMWR Recomm Rep. 2002;51(RR-17):6--7.) [^26]: There is no minimum age for pre-exposure immunization for rabies. (CDC. Human rabies prevention---United States, 2008: recommendations of the Advisory Committee on Immunization Practices. MMWR Recomm Rep. 2008; 57(RR-3):1--28.) [^27]: Hepatitis A vaccine, inactivated, GlaxoSmithKline. [^28]: EL.U., enzyme-linked immunosorbent assay (ELISA) units of inactivated hepatitis A virus. [^29]: Hepatitis A vaccine, inactivated, Merck & Co., Inc. [^30]: U., units of hepatitis A virus antigen. [^31]: Combined hepatitis A and hepatitis B vaccine, GlaxoSmithKline. [^32]: EL.U., enzyme-linked immunosorbent assay (ELISA) units of inactivated hepatitis A virus/micrograms hepatitis B surface antigen. [^33]: IG should be administered by intramuscular injection into either the deltoid or gluteal muscle. For children \<12 months of age, IG can be administered in the anterolateral thigh muscle. [^34]: Repeat every 5 months if continued exposure to hepatitis A virus occurs. [^35]: Combined hepatitis B--*Haemophilus influenzae* type b conjugate vaccine. This vaccine cannot be administered before age 6 weeks or after age 71 months. [^36]: Combined hepatitis B--diphtheria, tetanus, pertussis-inactivated poliovirus vaccine. This vaccine cannot be administered at birth, before age 6 weeks, or after age 7 years. [^37]: Combined hepatitis A and hepatitis B vaccine. This vaccine is recommended for persons ≥18 years who are at increased risk for both hepatitis A virus and hepatitis B virus infections. [^38]: Recombinant hepatitis B surface antigen dose. [^39]: Not applicable. [^40]: Adult formulation administered on a 2-dose schedule. [^41]: Higher doses might be more immunogenic, but no specific recommendations have been made. [^42]: Dialysis formulation administered on a 3-dose schedule at 0, 1, and 6 months. [^43]: Two 1.0-mL doses administered at one site, on a 4-dose schedule at 0, 1, 2, and 6 months. [^44]: Administer with cool liquid no warmer than 98.6° F (37° C). [^45]: HDCV, human diploid cell vaccine; PCEC, purified chick embryo cell. Patients who are immunosuppressed by disease or medications should postpone pre-exposure vaccinations and consider avoiding activities for which rabies pre-exposure prophylaxis is indicated. When this course is not possible, immunosuppressed persons who are at risk for rabies should have their antibody titers checked after vaccination. [^46]: RIG, rabies immune globulin; HDCV, human diploid cell (rabies) vaccine; PCEC, purified chick embryo cell. [^47]: All postexposure prophylaxis should begin with immediate, thorough cleansing of all wounds with soap and water. [^48]: Pre-exposure immunization with HDCV or PCEC, prior postexposure prophylaxis with HDCV or PCEC, or persons previously immunized with any other type of rabies vaccine and a documented history of positive antibody response to the prior vaccination. [^49]: RIG should not be administered. [^50]: Important details in the text. [^51]: For children \<7 years old, DTaP or DTP (DT, if pertussis vaccine contraindicated) preferred to tetanus toxoid alone. For children ≥7 years of age, Td preferred to tetanus toxoid alone. For adolescents and adults to age 64, tetanus toxoid as Tdap is preferred, if the patient has not previously been vaccinated with Tdap. [^52]: Yes, if more than 10 years since last dose. [^53]: Yes, if more than 5 years since last dose. More frequent boosters are not needed and can accentuate side effects. [^54]: Hepatitis B surface antigen. [^55]: IgG antibody to hepatitis B core antigen. [^56]: Immunoglobulin M. [^57]: Antibody to HBsAg. [^58]: Negative test result. [^59]: Positive test result. [^60]: To ensure that an HBsAg-positive test result is not a false-positive, samples with reactive HBsAg results should be tested with a licensed neutralizing confirmatory test if recommended in the manufacturer\'s package insert. [^61]: Persons positive only for anti-HBc are unlikely to be infectious except under unusual circumstances in which they are the source for direct percutaneous exposure of susceptible recipients to large quantities of virus (e.g., blood transfusion or organ transplant). [^62]: Milli-international units per milliliter. [^63]: The side effects of Ty21a are rare and mainly consist of abdominal discomfort, nausea, vomiting, and rash or urticaria. [^64]: Countries/areas where "a risk of yellow fever transmission is present," as defined by the World Health Organization, are countries or areas where "yellow fever has been reported currently or in the past, plus vectors and animal reservoirs currently exist" (see [www.who.int/ith/countries/2008_country_list.pdf](http://www.who.int/ith/countries/2008_country_list.pdf){#interref3}). [^65]: These countries are not holoendemic (i.e., only a portion of the country has risk of yellow fever transmission). Please see [Maps 2-3](#f6){ref-type="fig"} and [2-4](#f7){ref-type="fig"} and yellow fever vaccine recommendations ([Table 2-14](#cetable14){ref-type="table"}) for details. [^66]: Country requirements for yellow fever vaccination are subject to change at any time; therefore, CDC encourages travelers to check with the destination country\'s embassy or consulate before departure. [^67]: Yellow fever vaccine entry requirements are established by countries to prevent the importation and transmission of yellow fever virus, and are allowed under the International Health Regulations (IHR). Travelers must comply with these to enter the country, unless they have been issued a medical waiver. Certain countries require vaccination from travelers arriving from all countries, while some countries require vaccination only for travelers coming from "a country with risk of yellow fever transmission" (see [Table 2-12](#cetable12){ref-type="table"}). Country requirements are subject to change at any time; therefore, CDC encourages travelers to check with the destination country\'s embassy or consulate prior to departure. [^68]: As a part of an ongoing project, CDC, WHO, and other partners have made every effort to harmonize the listed yellow fever vaccine requirements and recommendations wherever possible. These efforts will continue, and will likely result in further changes to the printed version of this table. Please check the online version of the Yellow Book ([www.cdc.gov/yellowbook](http://www.cdc.gov/yellowbook){#interref5}) for the latest information on country requirements and vaccine recommendations. [^69]: The information in the section on yellow fever vaccine recommendations is advice given by CDC to prevent yellow fever infections among travelers. Note: CDC recommendations and country requirements may not be the same. [^70]: Recommendations are subject to change at any time if disease conditions change; therefore, CDC encourages travelers to check for relevant travel notices on the website ([www.cdc.gov/travel](http://www.cdc.gov/travel){#interref6}) prior to departure. [^71]: Please note, the U.S. Advisory Committee on Immunization Practices (ACIP) recommends avoiding vaccination of infants \<9 months of age. [^72]: Data are based on published reports and personal correspondence. Risk assessments should be performed cautiously because risk can vary within areas and from year to year, and surveillance data regarding human cases and JE virus transmission are incomplete. [^73]: In some endemic areas, human cases among residents are limited because of vaccination or natural immunity. However, because JE virus is maintained in an enzootic cycle between animals and mosquitoes, susceptible visitors to these areas still may be at risk for infection. [^74]: Bat rabies may exist in some areas that are reportedly free of rabies in other animals. [^75]: Bat lyssaviruses are known to exist in these areas that are reportedly free of rabies in other animals. [^76]: Most of Pacific Oceania is reportedly rabies-free. [^77]: Judgment of relative risk and extra monitoring of vaccination status of laboratory workers is the responsibility of the laboratory supervisor (see U.S. Department of Health and Human Services\' Biosafety in Microbiological and Biomedical Laboratories, 1999) [^78]: Pre-exposure booster immunization consists of one dose of human diploid cell (rabies) vaccine (HDCV) or purified chick embryo cell (PCEC) vaccine, 1.0-mL dose, intramuscular (IM) (deltoid area). Minimum acceptable antibody level is complete virus neutralization at a 1:5 serum dilution by the rapid fluorescent focus inhibition test. A booster dose should be administered if titer falls below this level. [^79]: 1 Glucose-6-phosphate dehydrogenase. All persons who take primaquine should have a documented normal G6PD level before starting the medication. [^80]: The information presented herein was accurate at the time of publication; however, factors that can change rapidly and from year to year, such as local weather conditions, mosquito vector density, and prevalence of infection, can markedly affect local malaria transmission patterns. Updated information may be found on the CDC Travelers\' Health website at [www.cdc.gov/travel](http://www.cdc.gov/travel){#interref64}. [^81]: Refers to *P. falciparum* malaria. [^82]: Estimates of malaria species are based on best available data from multiple sources. [^83]: Several medications are available for chemoprophylaxis. When deciding which drug to use, consider specific itinerary, length of trip, cost of drug, previous adverse reactions to antimalarials, drug allergies, and current medical history. All travelers should seek medical attention in the event of fever during or after return from travel to areas with malaria. [^84]: Primaquine can cause hemolytic anemia in persons with G6PD deficiency. Patients must be tested and documented to have a normal level of G6PD activity prior to starting primaquine. [^85]: NSF 53 rating on a filter certifies for cyst/oocyst removal. [^86]: Very cold water requires prolonged contact time with iodine or chlorine to kill *Giardia* cysts. These contact times have been extended from the usual recommendations in cold water to account for this and for the uncertainty of residual concentration. [^87]: Drops per minute.

ANNOUNCEMENT {#s1} ============ Achromobacter species are Gram-negative bacilli commonly found in soil and water but also are associated with human clinical samples. Relevant species include Achromobacter piechaudii, Achromobacter spanius, and Achromobacter marplatensis ([@B1], [@B2]). Here, we report the complete genome sequence of a hydrocarbon-degrading strain, Achromobacter sp. B7, isolated during bioremediation trials of a diesel-spiked soil from Laguna Verde, Valparaiso Region, Chile. Bioremediation involved the addition of a hydrocarbon-degrading bacterial enrichment ([@B3]). Achromobacter sp. B7 is able to grow in M9 minimal medium on hexane, octane, hexadecane, naphthalene, biphenyl, and diesel at 30°C ([@B4]). Genomic DNA was isolated from a fresh culture biomass of Achromobacter sp. B7 grown on tryptic soy agar (TSA) and lysed in EDTA-saline buffer (0.15 M NaCl, 0.01 M EDTA, pH 8.0) with 10 mg ml^−1^ lysozyme for 2 h at 37°C. Genomic DNA was isolated with the Promega Wizard kit and a modified Marmur procedure ([@B5]). DNA was sequenced on an Illumina HiSeq 4000 system (GATC Biotech, Germany), generating 10,102,770 pair-end reads of 150 bp each and yielding a total of 1,525.5 Mb. Standard genomic library preparation was performed with an optimized protocol, and standard Illumina adapter sequences were used. Sequence reads were trimmed, using Sickle (version 1.33) ([@B6]), with a Phred quality score threshold of 30. A subset of 6,666,666 high-quality paired-end reads with a total of 904,208,607 bp was obtained. Additionally, DNA was sequenced with an Oxford Nanopore Technologies (ONT) MinION instrument. A library was prepared with the ONT rapid barcoding kit (SQK-RBK004). Albacore version 2.1.10 was used for base calling, which yielded a total of 2,260 Mb distributed in 217,643 reads. The subsets of Illumina paired-end reads (904 Mb) and base-called Nanopore reads (2,260 Mb) were used to perform a hybrid assembly, using hybridSPAdes from SPAdes version 3.11.1 ([@B7], [@B8]). The hybrid assembly resulted in a final closed and complete chromosome sequence of 6,236,552 bp with a G+C content of 64.8%. The B7 genome was annotated with the NCBI Prokaryotic Genome Annotation Pipeline (PGAP) version 4.6 ([@B9]), which identified 5,578 coding sequences and 57 tRNAs. After 16S rRNA gene sequence analyses, A. spanius DSM 23806 was the most closely related species (100% identity). Analysis by average nucleotide identity based on BLAST (ANIb) ([@B10]), using JSpeciesWS version 3.0.20 ([@B11]), with A. piechaudii NBRC 102461^T^, A. spanius DSM 23806 ^T^, and A. marplatensis sp. B2^T^, resulted in ANIb values of 88.0%, 87.4%, and 85.4%, respectively. Therefore, strain B7 most likely represents a novel species of Achromobacter. Genome analysis, with the Comprehensive Antibiotic Resistance Database 3.0.0 (CARD) ([@B12]), identified four antibiotic resistance genes in strain B7 (DVB37_25330, DVB37_12020, DVB37_10250, and DVB37_17655), possessing 82%, 42%, 43%, and 43% identity, respectively, with a resistance-nodulation-cell division (RND) efflux pump for fluoroquinolone and tetracycline. Gentisate (DVB37_25350) and homoprotocatechuate (DVB37_16865) dioxygenases possessed 39% and 63% identity with Streptomyces sp. strain WA46 and Escherichiacoli C enzymes, respectively. Catechol-1,2-dioxygenase (DVB37_15220) and the protocatechuate-3,4-dioxygenase alpha subunit (DVB37_19170) possessed 42% and 79% identity with Achromobacter sp. strain ADP1 and Achromobacter lwoffii K24 enzymes. The genome sequence of Achromobacter sp. B7 represents essential data for genomic and metabolic characterization of environmental Achromobacter strains. Data availability. {#s1.1} ------------------ This whole-genome shotgun project has been deposited at DDBJ/ENA/GenBank under the accession number [CP032084](https://www.ncbi.nlm.nih.gov/nuccore/CP032084). The version described in this paper is the first version, CP032084.1. The accession number for the publicly available raw data at NCBI is [PRJNA481776](https://www.ncbi.nlm.nih.gov/bioproject/PRJNA481776). V.M., L.H., R.E.D., and M.S. acknowledge the FONDECYT 1151174 and CONICYT PIA Anillo GAMBIO ACT 172128 projects. D.J.-L. and E.R.B.M. acknowledge support by the Joint Program Initiative-Anti-Microbial Resistance (JPI-AMR) (Vetenskapsrådet Project 2016-06504). F.S.-S., D.J.-L., and E.R.B.M. were supported by the Culture Collection University of Gothenburg (CCUG), Department of Clinical Microbiology, Sahlgrenska University Hospital, Gothenburg, Sweden. [^1]: **Citation** Méndez V, Hernández L, Salvà-Serra F, Jaén-Luchoro D, Durán RE, Barra B, Piñeiro-Iglesias B, Moore ERB, Seeger M. 2018. Complete genome sequence of the hydrocarbon-degrading strain *Achromobacter* sp. B7, isolated during petroleum hydrocarbon bioremediation in the Valparaiso Region, Chile. Microbiol Resour Announc 7:e01326-18. <https://doi.org/10.1128/MRA.01326-18>.

Introduction ============ Clinical trial participants in sub-Saharan Africa often have limited understanding of the study information provided during the informed consent process. Low literacy and the difficulty of presenting information in local languages with no standard written form are contributory factors.[@R1]^,^[@R2] Nevertheless, international ethics guidelines[@R3]^,^[@R4] require informed consent to be obtained in a manner that can be understood by individuals volunteering for clinical studies. Moreover, the Declaration of Helsinki states that special attention should be given to the specific information needs of participants and to the methods used to deliver that information.[@R4] Consequently, study information must be provided in a medium and language understood by potential participants. However, informed consent documents are usually written in an official national language, often a common international language. In countries such as the Gambia, where local languages have no standard written form, translating documents into the local language and back-translating into the national language (English, in this case), to check consistency is both impractical and inaccurate.[@R2] Comprehension of consent information is essential for protecting study participants' rights and for complying with the principles of good clinical practice. In sub-Saharan Africa, an increasing number of clinical trials are being conducted in populations that are vulnerable to exploitation because of poverty, illiteracy, social exclusion or poor access to health care.[@R5]^,^[@R6] In particular, illiterate participants may not understand research concepts, which could undermine their ability to give truly informed consent.[@R5] Comprehension could be improved using multimedia consent tools that have been effective for communicating crucial research information in developed countries.[@R7]^--^[@R9] Moreover, empirical studies indicate that such tools provide an alternative means of presenting study information to vulnerable groups.[@R10]^,^[@R11] The effectiveness of multimedia consent tools among clinical trial participants with low English-language literacy in Africa has not been determined. Previously in the Gambia, we developed and validated both a multimedia tool for providing study information to clinical trial participants[@R12] and a computerized, audio questionnaire for assessing their comprehension of informed consent.[@R13] Here we report on the acceptability, ease of use and effectiveness of the multimedia tool among participants in a malaria treatment trial in the Gambia. Methods ======= We conducted a randomized controlled trial in Basse and Jahaly Provinces in the Upper River Region and Central River Region, respectively, of the Gambia from 15 August 2013 to 12 March 2014. The study was nested within a parent study -- the PRINOGAM trial.[@R14] The aim of this trial was to determine the optimal dosage of primaquine required to clear gametocytes and block disease transmission in asymptomatic malaria carriers. Participants in the PRINOGAM trial were aged 1 year or older and were seen on the day of inclusion (day 0, the baseline) and on days 3, 7, 14, 21, 28, 35 and 42. In the study areas, most residents were subsistence farmers and the adult literacy rate was about 50%.[@R15] To be included in our study, individuals had to be eligible for the PRINOGAM trial, to be aged 18 years or older, to speak and understand one of the three major Gambian languages (i.e. Mandinka, Fula or Wolof) and to have no obvious communication, visual or cognitive impairments. Since a systematic review showed that, on average, 47% of participants in African studies understood basic research concepts,[@R1] we estimated that a study with a 90% power to detect a 20% difference between intervention and control arms at the 5% significance level for a two-sided test would require 137 participants in each arm. On assuming a 10% attrition rate, we estimated that approximately 150 participants would be required in each group. In our study, participants were randomly assigned to the intervention or control arm on day 0 of the PRINOGAM trial, at the time of treatment randomization. An independent statistician used the RANDI2 web-based application (available at: <http://dschrimpf.github.io/randi3/>) to generate a randomization list for each trial site and participants were allocated to the intervention or control arm at a 1:1 ratio with a block size of four. In addition, participants were stratified by age group and sex. Intervention ------------ The multimedia informed consent tool has been reported in detail elsewhere.[@R12] Briefly, it contained information from the PRINOGAM consent document under the headings: (i) introduction; (ii) reason for the study; (iii) what glucose-6-phosphate dehydrogenase deficiency is; (iv) how to take part; (v) what happens if you take part; (vi) what blood tests are performed; (vii) what the side-effects and possible risks of taking part are; (viii) what the potential benefits are; (ix) how taking part is kept confidential; (x) who carried out an ethical review of the study; and (xi) who to contact if you have questions. The information in each section was presented in a context-specific visual form by members of the clinical trial team acting out various scenes after training. Video recordings were made by a multimedia expert and voice-overs were added separately in the three main Gambian languages. Adverse events of the study drugs that could not be adequately presented in the acted scenes, such as headache, diarrhoea or the passage of dark-coloured urine, were illustrated using animations.[@R12] In addition, the multimedia tool was tailored to the cultural and linguistic diversity of the Gambian population. A digital versatile disc (DVD) incorporating the three language versions was produced and uploaded onto laptop computers. For the intervention arm of our study, a trained field assistant selected the language preferred by each participant and played the DVD using a laptop computer in a quiet room. Individuals consented to participating in the trial by either signing or thumb-printing the consent form in the presence of an impartial witness. In the control arm, clinical trial information was presented using the current standard practice accepted by the national ethics committee in the Gambia (chair of the Gambia's National Ethics Committee, personal communication, 12 October 2010). In the absence of acceptable written versions of the local languages, the study's principal investigator trained field staff, who were native speakers of the major local languages, on the correct interpretation of the English version of the participants' information sheet. Subsequently, the study information was presented verbally to prospective participants during discussions on informed consent. Again, consent was given by either signing or thumb-printing the consent form.[@R2]^,^[@R16] Assessment ---------- We have previously shown that the computerized, audio, informed consent comprehension questionnaire is a reliable and valid tool for assessing comprehension of informed consent among Gambian trial participants.[@R13] It comprises 26 questions in the three major Gambian languages: nine open-ended, seven closed-ended and 10 multiple-choice questions. The questionnaire was administered using laptop computers by trained interviewers who entered participants' responses to each question. Responses were automatically recorded in the questionnaire computer database. Comprehension was assessed on the basis of recall and understanding.[@R17] Recall relates to the participant's ability to correctly answer closed-ended and multiple-choice questions. Understanding is defined as the participant's ability to correctly interpret or respond to open-ended questions. Our primary study outcome was comprehension of consent information as indicated by the participant's questionnaire score, expressed as a percentage, on day 0 at study inclusion. Secondary outcomes were comprehension on days 7, 14, 21 and 28. In addition, 119 randomly selected participants took part in focus group discussions on day 35, to further explore understanding of the PRINOGAM trial and to evaluate the acceptability and ease of use of the multimedia tool. Ten focus group discussion sessions were held in Basse Province; only six were held in Jahaly Province because there were fewer participants. Seven or eight participants were invited to each session and participants were segregated by sex so they could express their views more easily. The sessions were facilitated using a specially designed focus group discussion guide (available from corresponding author). Audio recordings of the sessions were transcribed into English by three translators and the accuracy of the translations was confirmed by independent translators fluent in the local languages and English. The transcribed texts were analysed using NVivo version 10.0 (QSR International Pty. Ltd, Doncaster, Australia) and the main themes that emerged were coded line by line to elucidate their meanings. The themes were subsequently sorted and collated into categories and subcategories and themes from the two sites were compared, integrated and refined. Finally, selected quotations from participants were used to illustrate differences in understanding of consent information between the intervention and control arms. Statistical analysis -------------------- Because data on comprehension of consent information were not normally distributed, we compared the median and interquartile range of participants' comprehension scores at each visit in the two study arms. Associations between participants' characteristics and baseline comprehension scores were assessed using the Mann--Whitney *U* test if the characteristic was classified using two categories and the Kruskal--Wallis test if more than two categories were used. We used forward stepwise variable selection in a multivariate logistic regression analysis to identify factors that were associated with comprehension on day 0. For this purpose, we reclassified the comprehension scores as a binary variable equal to 2 (below median) or 1 (median or above). Since participants were recruited at two different sites, we investigated the effect of clustering on comprehension using a mixed-effects model. Survival analysis was used to extrapolate the decline in comprehension scores beyond the end of the study follow-up. A *P*-value less than 0.05 was regarded as statistically significant. All statistical calculations were performed using Stata version 12.1 (StataCorp. LP, College Station, United States of America). Approval was obtained from the ethics committee of the London School of Hygiene & Tropical Medicine in the United Kingdom of Great Britain and Northern Ireland and the Gambian Government--Medical Research Council Joint Ethics Committee. The trial was registered with the Pan African Clinical Trials Registry (PACTR 201402000775274). Results ======= Of the 347 participants enrolled in the PRINOGAM trial, 26 refused to take part (7.5%) in the study of the multimedia informed consent tool. Most of those who refused said they did not have time to wait because of domestic demands. In addition, 10 participants (2.9%) insisted on using the multimedia tool without going through randomization, most likely because they had heard about the tool through friends or family already enrolled in the study. After excluding these 36 participants, 311 were enrolled in the study and included in final analysis ([Fig. 1](#F1){ref-type="fig"}). [Table 1](#T1){ref-type="table"} (available at: <http://www.who.int/bulletin/volumes/93/5/14-146159>) shows there was no significant difference in demographic characteristics at baseline between participants in the two study arms. The playing time of the multimedia DVD was 19.4 minutes and the standard consent process took 30 to 35 minutes depending on the communication skills and experience of the research assistant providing the information. On average, question-and-answer sessions after the consent interviews took 32 minutes. {#F1} ###### Sociodemographic characteristics of study participants in randomized controlled trial of a multimedia consent tool, the Gambia, 2013--2014 Characteristic No. (%) *P* ------------------------------------------- ------------ ------------ -------- **Age, years** 0.247 18--25 23 (14.8) 35 (22.4) NA 26--33 50 (32.3) 44 (28.2) NA 34--41 40 (25.8) 35 (22.4) NA 42--49 28 (18.1) 34 (21.8) NA \> 49 14 (9.0) 8 (5.1) NA **Sex** 0.692 Female 96 (61.9) 100 (64.1) NA Male 59 (38.1) 56 (35.9) NA **Place of residence**  0.443 Basse province 102 (65.8) 109 (69.9) NA Jahaly province 53 (34.2) 47 (30.1) NA **Ethnicity** 0.666 Mandinka 75 (48.4) 81 (51.9) NA Fula 66 (42.6) 62 (39.7) NA Wolof 8 (5.2) 5 (3.2) NA Sarahule 5 (3.2) 7 (4.5) NA Manjago 1 (0.7) 1 (0.6) NA **Education^b^** 0.097 Formal education 41 (26.5) 29 (18.6) NA No formal education 114 (73.5) 127 (81.4) NA **Religious affiliation** 0.995 Islam 153 (98.7) 154 (98.7) NA Christianity 2 (1.3) 2 (1.3) NA **Previous clinical trial participation** 0.071 Yes 14 (9.0) 28 (18.0) NA No 140 (90.3) 127 (81.4) NA Don't know 1 (0.7) 1 (0.6) NA NA: not applicable. ^a^ In the intervention arm, study participants were informed about the clinical trial using a multimedia informed consent tool; in the control arm, information was provided using the current standard method for informed consent. ^b^ For the purposes of this study, the term "formal education" was defined as education based on an English-language curriculum that involved the completion of primary school, with or without 3 years of junior secondary school. The median consent comprehension score of participants who used the multimedia tool was higher at all times points than those who received information using current standard practice. For example, at day 0 ([Fig. 2](#F2){ref-type="fig"}), the median comprehension score in the intervention arm was 64% compared with 40% in the control arm (*P* = 0.042). The corresponding comparisons for days 7, 14, 21 and 28 are shown in [Fig. 3](#F3){ref-type="fig"}, [Fig. 4](#F4){ref-type="fig"}, [Fig. 5](#F5){ref-type="fig"} and [Fig. 6](#F6){ref-type="fig"}, respectively. [Table 2](#T2){ref-type="table"} shows that comprehension of informed consent at baseline in the two study arms combined was significantly greater in male participants, those who resided in Basse province and those who had received formal education based on an English-language curriculum. In addition, multivariate logistic regression analysis found that, after controlling for other covariates, poorer comprehension at baseline was significantly and independently associated with female sex (odds ratio, OR: 0.29; 95% confidence interval, CI: 0.12--0.70) and residing in Jahaly province (OR: 0.33; 95% CI: 0.13--0.82; [Table 3](#T3){ref-type="table"}). However, on applying the mixed-effects model, place of residence was no longer significantly associated with comprehension at baseline (OR: 0.85; 95% CI: 0.45--1.60; details available from authors). Survival analysis showed that the risk that a participant's comprehension score would, during follow-up, drop to below 50% of that at day 0 was lower in the intervention arm than the control arm (hazard ratio 0.22, 95% CI: 0.16--0.31; details available from authors). Extrapolating beyond the end of follow-up indicated that the estimated median time for the comprehension score to drop below 50% was 67.0 days in the intervention arm compared with 40.6 days in the control arm (*P* \< 0.0001). A summary of the economic and financial costs of developing and administering the multimedia informed consent tool is available from corresponding author. {#F2} {#F3} {#F4} {#F5} {#F6} ###### Influence of sociodemographic characteristics on comprehension of informed consent at baseline, the Gambia, 2013--2014 Characteristic Informed consent comprehension questionnaire score, median (IQR) *P^a^* ------------------------------------------- ------------------------------------------------------------------ ------------------- ----------- **Age, years** 0.54^b^ 18--40 68.0 (63.0--73.0) 40.5 (33.5--46.5) NA ≥ 41 65.0 (61.0--72.0) 43.0 (35.5--53.0) NA **Sex** 0.032^b^ Male 68.0 (65.0--73.0) 45.0 (38.0--51.0) NA Female 67.0 (61.0--72.0) 39.0 (33.0--46.0) NA **Place of residence** 0.021^b^ Basse province 67.5 (63.0--73.0) 44.0 (39.0--51.0) NA Jahaly province 67.0 (61.0--74.0) 33.0 (30.0--38.0) NA **Education^c^** 0.0049^b^ Formal education 66.5 (61.0--72.0) 40.0 (33.0--48.0) NA No formal education 70.0 (65.0--74.0) 45.0 (40.0--47.0) NA **Language of assessment** 0.92^d^ Mandinka 67.0 (64.0--73.0) 41.0 (33.0--47.0) NA Fula 69.0 (61.0--74.0) 38.0 (30.0--45.0) NA Wolof ^e^ 67.5 (62.0--73.0) 42.0 (35.0--50.0) NA **Previous clinical trial participation** 0.21^d^ Yes 67.0 (63.0--73.0) 41.0 (34.0--48.0) NA No 69.0 (65.0--72.0) 40.0 (33.0--47.0) NA Don't know 48.0 (48.0--48.0) 43.0 (43.0--43.0) NA IQR: interquartile range; NA: not applicable. ^a^ The *P*-value is for the significance of the influence of the sociodemographic characteristic on the participants' comprehension of informed consent at baseline in the two arms combined. ^b^ The *P*-value for the significance of the association between the sociodemographic characteristic and baseline comprehension scores was assessed using the Mann--Whitney *U* test when there were two categories for the characteristic. ^c^ For the purposes of this study, the term "formal education" was defined as education based on an English-language curriculum that involved the completion of primary school, with or without 3 years of junior secondary school. ^d^ The *P*-value for the significance of the association between the sociodemographic characteristic and baseline comprehension scores was assessed using the Kruskal--Wallis test when there were more than two categories for the characteristic. ^e^ Included participants of Sarahule and Manjago ethnicities. ###### Influence of sociodemographic characteristics on comprehension of informed consent at baseline, by multivariate logistic regression analysis, the Gambia, 2013--2014 Characteristic Likelihood of better comprehension of informed consent,^a^ OR (95% CI) --------------------------------------------------------------------- ------------------------------------------------------------------------ Age group (18--40 years versus \> 41 years) 1.41 (0.62--3.21) Female versus male 0.29 (0.12--0.70) Resident of Jahaly province vs Basse province 0.33 (0.13--0.82) Formal education versus no formal education^b^ 0.67 (0.23--1.93) Assessment language: (Mandinka vs Wolof and Fula) 0.56 (0.29--1.08) Previous trial participation versus no previous trial participation 1.07 (0.42--2.73) CI: confidence interval; OR: odds ratio. ^a^ Better comprehension of informed consent was defined as an informed consent comprehension questionnaire score above the median for the intervention or control group, as appropriate. ^b^ For the purposes of this study, the term "formal education" was defined as education based on an English-language curriculum that involved the completion of primary school, with or without 3 years of junior secondary school. Focus group discussions ----------------------- The 119 participants in the focus groups were aged between 23 and 47 years, 60 were in the intervention arm and 59 were in the control arm, 79 were female and 40 were male, and 56 resided in Basse Province and 63 resided in Jahaly Province. In Basse, 69.6% were female compared with 63.4% in Jahaly. ### Informed consent There was general consensus that signing or thumb-printing the consent form implied a commitment to participate in the research. One participant from Basse said, "When you put your hand in that paper, then you have promised to be part of the study." However, there were divergent opinions on the right to withdraw. Whereas most participants in the control arm felt strongly that it was morally wrong to stop participating before the end of the study, the majority in the intervention arm stated that participants were free to leave at any time. One participant in the intervention arm said, "What we always think is that our doctors will be angry if we leave before the end of the study, but I now know after watching the 'film' that we have freedom to leave at any time, without telling them the reason for this... " Participants were unequivocal that incentives were needed to motivate them to join and stay in the trial. Whereas the majority considered the benefits of participation to be free medical care, a minority also wanted fertilizers and sponsorship of their children's education. One said, "We appreciate all the good things you have done to care for us and our children, but the real help that we expect and will never be forgotten is to give us fertilizers for our crops and train our children to be like you\...." Most participants in the control arm could not describe the risks involved, whereas participants in the intervention arm could often name the adverse events associated with the study medications, such as headache, abdominal pain, vomiting and diarrhoea. A male participant in the intervention arm described the passage of dark-coloured urine associated with haemolysis caused by primaquine in individuals with glucose-6-phosphate dehydrogenase deficiency, as urinating "*wonjo*". "*Wonjo*" is a popular, local, dark-red coloured, hibiscus drink. Most participants in the intervention arm described the randomization procedure graphically and some participants in the control arm said randomization was done to ensure all participants had an equal chance of participating. ### Multimedia informed consent tool Overall, 70% (42/60) of focus group participants from the intervention arm thought the visual and verbal information presented through the DVD was clear and easy to understand. However, a few expressed reservations. One said, "Although I like the (computer) pictures and sounds, I prefer face-to-face talking. I can easily ask questions that are not clear to me and this will make me understand better." Another said, "The Fula man (interpreter) on the computer (video) repeated the same information over and over, and this made everything boring to me." Discussion ========== Our study's findings confirm that use of a multimedia informed consent tool results in significantly better understanding of clinical trial information than the current standard method for obtaining consent. Participants using the multimedia tool achieved significantly higher consent comprehension scores at all study assessments. Moreover, results indicated that those who used the multimedia tool retained the study information significantly longer. Although education, place of residence and sex were associated with participants' comprehension scores at baseline, multivariate analysis found that the associations were significant only for place of residence and sex. This contrasts with findings of previous studies, which reported that educational level was an independent predictor of comprehension.[@R18]^,^[@R19] The difference may have arisen because the large majority of our study participants had no formal education, which further strengthens the case for using multimedia tools to provide information to participants with low levels of literacy. Moreover, the multimedia tool was well received: in fact, during recruitment some participants insisted on being allocated to the intervention arm without undergoing formal randomization. Also, in focus group discussions, participants said they liked the way the study information was presented visually and verbally through the multimedia tool. Our study adds to the emerging body of evidence that multimedia tools can increase trial participants' comprehension of informed consent in Africa, particularly in areas where low literacy is common.[@R10]^,^[@R11] In our study, research concepts that are known to be difficult to understand were clearly illustrated using video recordings and animations and clearly explained by sound tracks in three local languages. Furthermore, we nested our study within a malaria treatment trial to ensure that our findings were relevant to trials carried out in real-life settings, whereas previous studies conducted outside Africa adopted a simulated study design.[@R8]^,^[@R9] The first limitation of our study was that, since our centre has been conducting research projects in the Gambia for more than 60 years, the local population was familiar with research projects. The effectiveness of the multimedia tool may be different in areas where the population is less familiar with research. Second, there was some clustering of participants: around two-thirds were recruited in Basse province, where the prevalence of asymptomatic malaria infection was higher than in Jahaly province. However, sociodemographic and epidemiological characteristics were similar at the two sites. Moreover, the mixed-effects model showed that place of residence had no significant effect on comprehension, which suggests that the effect of clustering was not significant. In conclusion, our multimedia tool improved trial participants' comprehension and retention of information about informed consent in an area of the Gambia with low levels of literacy. Such tools can help address the fundamental ethical challenge of obtaining informed consent from individuals in these settings. We thank Joseph Okebe, Frank Sanya-Isijola and Edgard Dabira of the PRINOGAM trial, Jenny Mueller and Vivat Thomas-Njie of the Medical Research Council Unit's Clinical Trials Support Office, Odile Leroy and Nicola Viebig of the European Vaccine Initiative in Germany and the clinical trial teams and participants in Basse and Jahaly. The study was supported by a grant (IP.2008.31100.001) from the European and Developing Countries Clinical Trials Partnership. Nuala McGrath was supported by a Wellcome Trust Fellowship (grant WT083495MA) and Neal Alexander received support from the United Kingdom Medical Research Council and Department for International Development (MR/K012126/1). None declared.

###### Strengths and limitations of this study - Systematic synthesis of data on disease burden and existing coverage of interventions relevant to diarrhoea, malaria and HIV prevention. - Three alternative ways to prioritise countries for integrated prevention campaign (IPC) implementation in a visually accessible format. - Facilitation of more objective decision-making regarding areas for IPC scale-up. - Limitations in the availability of published data. Introduction {#s1} ============ The Millennium Development Goals (MDGs) provide specific targets for global improvements in access to healthcare by 2015.[@R1] However, despite the availability of simple, low-cost interventions for many diseases, the capacity of healthcare systems to deliver these interventions is often limited, and many countries are unlikely to meet these targets.[@R2] [@R3] In response, the United Nations General Assembly passed a resolution in 2010 identifying the integration of services, the increased use of common delivery platforms, and the scaling up of proven interventions as critical strategies to accelerate progress towards the MDGs.[@R4] Community-based interventions targeting multiple diseases have the potential to rapidly and equitably increase intervention uptake, often reaching greater numbers of underserved populations than interventions delivered in health facilities.[@R5] In 2008, an integrated prevention campaign (IPC) in Western Province, Kenya delivered insecticide-treated bed nets (ITNs), a point-of-use water filter, HIV testing, condoms and health messages to more than 80% of local adults in 7 days.[@R6] Participants who tested HIV-positive received on-site CD4 cell count, cotrimoxazole and referral for HIV care and treatment. The IPC was estimated to avert 16 deaths and 440 disability-adjusted life years (DALYs), and save more than US\$16 000/1000 participants.[@R7] Global scale-up of IPCs may represent a practical and cost-effective method to deliver multiple health interventions to populations at highest risk. However, to ensure that available funds are well used, stakeholders must identify areas where IPC implementation can have the greatest impact. To promote more objective decision-making than traditional processes, which are often opaque and based on subjective assessment,[@R8] stakeholders need systematic methods to identify areas of greatest opportunity. Using the example of the Kenya diarrhoea, malaria and HIV IPC, we developed a data-driven tool to assist stakeholders in synthesising country-specific data to determine the potential impact of IPC implementation. We have focused on one proven IPC in particular to help explore the utility of this type of approach; however, this type of tool could be readily adapted and used for a multitude of other diseases and potential interventions. Methods {#s2} ======= Overview {#s2a} -------- We developed three versions of an 'opportunity index' to identify countries with the greatest potential for IPC impact by adapting a previously developed method. Our goal is to provide a country-level, easy-to-read summary of the factors related to the potential success of an IPC focused on diarrhoea, malaria and HIV prevention. We collected data on relevant indicators from global databases and used a colour-coding system to represent each country\'s opportunity level based on each indicator. Finally, we ranked countries based on three composite measures: absolute burden across the three diseases (in DALYs per capita); burden rank across the three diseases in relation to other countries and disease burden plus 'intervention opportunity,' that is, the current lack of coverage for IPC-related interventions. DALY burden {#s2b} ----------- To quantify the overall disease burden attributed to diarrhoea, malaria and HIV, we used DALYs---a summary measure combining the number of Years of Life Lost due to premature mortality with Years of Life with Disability.[@R9] We calculated the total DALYs due to the three diseases for 214 World Bank-defined economies.[@R10] Total DALYs per disease were calculated as the product of annual cases (see online data supplement 1 for details) and the DALYs associated with each case. Using a discount rate of 3%, we estimated the DALYs due to a case of diarrhoea and a case of malaria using the following formula: where CFR is case death rate, DALY~d~ is the DALYs due to a death from the disease and DALY~m~ is DALYs due to morbidity from the disease. We calculated country-specific case death rates for malaria and diarrhoea (see online data supplement 1), and estimated the DALYs due to each malaria and diarrhoea death at 28 (author derivation).[@R11] Using published estimates of disability weights[@R12] and average duration of disease,[@R13] [@R14] we calculated an estimate of the DALYs due to each non-fatal episode of malaria and diarrhoea at 0.0037 and 0.0013, respectively. For HIV, we estimated 10 DALYs per case, assuming 18 years on antiretroviral treatment (ART), life expectancy at age 35 (average age of initiation of ART)[@R15] of 34 years in Kenya,[@R16] and 75% access to ART. This assumption is based on projected increases in ART access, and we examine uncertainty in this estimate in a sensitivity analysis in a separate IPC cost-effectiveness analysis paper. We obtained a combined total DALY burden in each country by summing the total DALYs across the three diseases. Country inclusion {#s2c} ----------------- To facilitate identification of those countries in which an IPC would be most beneficial, we limited the prioritisation analysis to low-income and middle-income countries as defined by the World Bank,[@R10] and countries with a total DALY burden for the three diseases in the highest tertile of the sample (≥87 000 DALYs). Country indicators {#s2d} ------------------ We identified 10 disease burden and intervention coverage indicators to help characterise countries based on their level of opportunity for IPC implementation ([table 1](#BMJOPEN2013004308TB1){ref-type="table"}; see online data supplement 1 for additional indicators). ###### Opportunity index indicators and definitions Category Indicator Definition Source ------------------------------------ ---------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ ----------------------------------------------------------------------------------------------------------------------------------- DALYs per capita DALYs per capita DALYs per person for diarrhoea, malaria and HIV. Calculated as the total DALY burden divided by the population[@R17] Author derivations; The World Bank Disease burden: diarrhoea Diarrhoea burden Percentage of childhood (\<5 years) deaths due to diarrhoea[@R18] Black *et al*[@R18] DALYs Total DALYs from diarrheal disease in the population. Calculated as number of annual cases of diarrhoea[@R19] times the number of DALYs due to a case of diarrheal disease (author calculation). Assumes an average diarrhoea episode duration of 4.43 days[@R14] and a disability weight for diarrhoea of 0.105[@R12] Author derivations based on data from Fischer Walker *et al*[@R19]; Lamberti *et al*[@R14]; Mathers *et al*[@R12] Disease burden: malaria Malaria burden Percentage of childhood (\<5 years) deaths due to malaria[@R18] Black *et al*[@R18] DALYs Total DALYs from malaria in the population. Calculated as number of annual cases of malaria[@R20] [@R21] times the number of DALYs due to a case of malaria (author calculation). Assumes an average malaria episode duration of 7 days[@R13] and a disability weight for malaria of 0.191[@R12] Author derivations based on data from Cibulskis *et al*[@R20]; Snow *et al*[@R13]; and Mathers *et al*[@R12] Disease burden: HIV HIV burden Prevalence in 15--49 years olds, 2009[@R22] Gapminder.org; Ethiopia and DRC: 2012 Country Progress Reports for UNAIDS; Afghanistan, Iraq and Yemen: UNAIDS 2010 Global Report DALYs Total DALYs from HIV/AIDS. Calculated as number of new HIV infections[@R26] times the number of DALYs due to a case of HIV (author calculation). Assumes 18 years on ART, life expectancy at 35 (average age of initiation of ART) of 34, and 75% access to ART UNAIDS AIDSInfo database Coverage of existing interventions Improved drinking water coverage (diarrhoea) Percentage of the population in 2010 using an improved drinking water source[@R27] WHO: Global Health Observatory Data Repository ITN coverage (malaria) Percentage of households in 2010 owning at least 1 insecticide-treated bed net. Countries with \>100% reported have been corrected to 100, and are designated with an asterisk[@R28] WHO Global Malaria Programme; World Malaria Report[@R21] Pregnant women tested, coverage (HIV) Percentage of pregnant women tested for HIV based on facility registers for ANC, L&D and postpartum care (2010). Denominator is \# pregnant women giving birth in the last 12 months[@R29] WHO: Global Health Observatory Data Repository ART, antiretroviral treatment; DALYs, disability-adjusted life years; ITN, insecticide-treated bed net. *Disease burden*: We calculated a DALYs per capita metric as the total DALY burden divided by the country\'s population.[@R17] For diarrhoea and malaria, we also collected data on the percentage of deaths under 5 due to diarrhoea and malaria, respectively, since the majority of cases, and particularly fatal cases, are in this demographic.[@R18] For HIV, we collected data on prevalence in the adult (15−49 years) population.[@R22]^--^[@R25] *Intervention coverage*: We compiled data on the existing coverage of relevant interventions that could affect burden in the three IPC diseases. For diarrhoea and malaria, respectively, this included the percentage of the population using an improved drinking water source [@R27] and the percentage of households owning at least one ITN.[@R28] For HIV, we collected data on the percentage of pregnant women tested for HIV in the past 12 months.[@R29] The latter was used as a proxy for HIV counselling and testing coverage since reliable data on population-level coverage is unavailable for all countries. Each of the indicators were colour coded into opportunity tertiles based on their ranking relative to other countries in the sample, with red, yellow and blue indicating high, medium and low opportunity, respectively. Visually, a country with more indicators coded in red suggests higher overall opportunity for the IPC compared with other countries in the sample. Opportunity ranking {#s2e} ------------------- To quantify each country\'s level of opportunity we created three opportunity metric ranking systems using the 10 indicators. First, we ranked countries based on *DALYs per capita* to identify the countries with the greatest disease burden attributed to diarrhoea, malaria and HIV. Since this combines DALYs for the three diseases, countries were ordered irrespective of whether the DALY burden was concentrated in one disease or spread across all three. Second, to identify countries where the relative burden was high for all three diseases, we ranked countries based on disease burden relative to other countries for each disease. To calculate this *composite ranked disease burden score* we assigned numerical values to three burden indicators: the percentages of childhood deaths due to diarrhoea and malaria, and HIV prevalence among adults. The values were on a scale of 1--3, where 1=low burden, 2=medium burden and 3=high burden in relation to the other countries in the sample. We calculated a country\'s ranked burden score by adding together numerical values for each burden indicator, and organised countries based on this new variable. For countries with the same numeric rank, those with a higher DALY per capita value were listed first. Finally, we developed a score combining disease burden and intervention opportunity. Intervention opportunity reflects existing intervention coverage and the potential gains from implementing an IPC. We created three intervention score variables representing the relative coverage score for each intervention, again using three levels: 1=high existing coverage (low opportunity), 2=medium existing coverage and 3=low existing coverage (high opportunity). Countries with missing data were assigned the medium coverage score. We then calculated a summary intervention opportunity score for each country. A *combined burden and intervention opportunity score (CBIO)* was constructed to explore the combined effect of relative disease burden and intervention opportunity. We weighted the individual disease burden score by a factor of two, and added the intervention opportunity score. Disease burden was assigned a greater weight than intervention coverage for two reasons: to filter out countries that appear to be high opportunity due to low existing intervention coverage, but which also have low disease burden and thus a lower need for an IPC; and the coverage data represent similar, but not identical, interventions as the IPC (ie, HIV testing among pregnant women is only a subset of the general adult population targeted by the IPC). We then ranked the countries based on the CBIO. Results {#s3} ======= Country sample {#s3a} -------------- In the 214 World Bank economies assessed for inclusion, we estimated the total annual DALY burden attributed to diarrhoea, malaria and HIV at nearly 135 million. The total DALY burden in each country ranged from 14 (Republic of Korea) to more than 33 million (India). Based on our country inclusion criteria of low-income and middle-income countries with a combined DALY burden in the top tertile, 70 countries were included in the final sample for analysis ([figure 1](#BMJOPEN2013004308F1){ref-type="fig"}). {#BMJOPEN2013004308F1} Forty-two of the 70 countries meeting our inclusion criteria were in Africa, with the majority of the rest from Asia and South and Central America. Collectively, the 70 countries in the sample accounted for 98% of the total DALYs attributed to diarrhoea, malaria and HIV in the world. Ninety per cent of the total global DALYs were concentrated in only 32 countries, and nearly three-quarters of the global DALYs were concentrated in just 16 countries. Opportunity indices {#s3b} ------------------- *Absolute DALY burden*: [Table 2](#BMJOPEN2013004308TB2){ref-type="table"} lists the highest opportunity (top tertile) countries based on DALYs per capita for diarrhoea, malaria and HIV. Swaziland was ranked highest based on burden across the three disease areas (0.15 DALYs per capita; dominated by HIV). All of the 23 countries ranking in the top tertile were in sub-Saharan Africa (see online supplementary appendix figure 1, data supplement 1). While the overall DALYs per capita for these countries was high, in several cases the DALY burden was concentrated in just one or two of the diseases. Although India had the highest total DALY burden for diarrhoea, malaria and HIV in our country sample, once the size of its population was factored in it did not appear in the top tertile of countries based on DALYs per capita (see online data supplement 2 for the complete opportunity indices including all 70 countries in our sample). ###### Highest opportunity countries based on DALYs per capita -------------------------------  ------------------------------- ITN coverage: values marked '100^\#^' were reported as \>100% by countries and corrected to 100 in this analysis. DALYs, disability-adjusted life years; ITN, insecticide-treated bed net. *Disease burden rank*: [Table 3](#BMJOPEN2013004308TB3){ref-type="table"} shows the opportunity index of the top 23 countries based on the composite ranked burden score. Countries at the top of the list have the highest relative burden in all three diseases. Five countries (Guinea-Bissau, Nigeria, Chad, Central African Republic and Cameroon) had high opportunity (ie, score of 3) in all three disease burden indicators, and 15 had high opportunity in at least two of the three, with medium opportunity (score of 2) in the third. Compared to the opportunity index based on DALYs per capita, five countries (Swaziland, Lesotho, South Africa, Guinea and Angola) no longer rank in the top tier, due to lower relative burden (ie, score of 1 or 2) in two of the diseases. These five countries were replaced by Tanzania, Togo, Rwanda, the Republic of Congo and Kenya---countries with medium opportunity based on DALYs per capita, but with higher relative opportunity when considering the burden of the three diseases equally. ###### Highest opportunity countries based on composite ranked burden score -------------------------------  ------------------------------- DALYs, disability-adjusted life years. *Combined disease burden and intervention opportunity*: After including existing intervention coverage levels, 21 countries that ranked as high opportunity based on the composite ranked burden score alone remained on the list of top 23 countries, although with changes to the relative order ([table 4](#BMJOPEN2013004308TB4){ref-type="table"}). Mozambique and Sierra Leone were replaced by Angola and Ethiopia, countries with the same composite ranked burden score but lower existing levels of intervention coverage. ###### Highest opportunity countries based on CBIO score -------------------------------  ------------------------------- CBIO, combined burden and intervention opportunity; ITN, insecticide-treated bed net. See online supplementary appendix figure 2 (data supplement 1) for maps providing a visual representation of where the greatest opportunity for IPC implementation exists. The three complete opportunity indices including data on all 70 countries in our country sample are available in online data supplement 2. Discussion {#s4} ========== This tool illustrates the application of an index comparing country-specific data on disease burden and intervention coverage to facilitate prioritisation for IPC scale-up. While the data presented here apply specifically to a diarrhoea, malaria and HIV IPC, the same methodology could be applied to prioritise other diseases or interventions. We estimated the total global burden due to diarrhoea, malaria and HIV at nearly 135 million DALYs per year, indicating a tremendous opportunity to impact global disease targets via interventions such as IPCs.[@R30] To determine countries in which implementation of a diarrhoea, malaria and HIV IPC would yield the most value, we used a visually accessible, systematic approach to summarise the opportunity for implementation in 70 high-burden countries. Based on each of the opportunity metrics we used, all of the countries with the highest opportunity for implementation are in sub-Saharan Africa. Although the overall rank order changes, 16 countries rank among the top 23 highest opportunity countries for all three opportunity metrics. The lists ranked by DALYs per capita and by composite ranked burden score vary somewhat in order and composition, since the former is an absolute ranking of *total* burden across the three diseases. Consequently, it is possible for one disease to dominate the DALY burden for a given country (as is the case with Swaziland, with low diarrhoea and malaria burden, but high HIV burden). Conversely, the indices ranked by the two composite scores are relative, so that the burden in each disease is weighted equally, maximising the countries where the burden in all three diseases is the highest. Our analysis indicates that five countries are classified as high opportunity based on DALYs per capita but fall down the list of opportunity when ranking based on the composite ranked burden score due to lower relative burden in at least one of the diseases. Stakeholders considering IPC scale-up may consider a number of factors when making decisions about where, and in how many countries, to implement: disease priority, the extent of funding resources, existing coverage of relevant interventions, etc. This analysis provides three ways to prioritise countries for IPC implementation: based on a high absolute burden, based on a high relative burden for all three diseases and by maximising countries where burden is high and existing coverage of IPC-relevant interventions is low (see online data supplement 1 for strategies for further prioritisation). In addition to factors impacting opportunity, feasibility factors, such as current levels of government expenditure on health, the presence of conflict, and access to routine health services are also important considerations. Community-based campaigns can enhance access to interventions among underserved and marginalised populations,[@R5] [@R31] [@R32] and IPCs may represent an efficient way to promote equitable coverage of important preventative interventions. Human resource capacity is another critical consideration; in countries with workforce shortages, IPCs may require mobilisation of existing healthcare workers for an extended period of time. In a separate analysis, we characterised our sample of 70 countries based on four feasibility metrics (see online data supplement 1). Given the variety of possible stakeholders in an IPC, feasibility determination and the specific measures for consideration will differ based on the implementing body. Once key feasibility metrics are pinpointed, this type of feasibility index could be applied to high-opportunity countries, and a revised list could be created to summarise the countries that fulfil both high opportunity and high feasibility criteria. Regardless of the approach used, stakeholders must be cognizant of feasibility considerations that could influence the potential success or failure of a campaign. There are several strengths to the approach presented here. The index method synthesises a large volume of data from disparate sources into a single table, enabling side-by-side comparisons of several indicators between countries. The system of colour-coding indicators into low, medium and high opportunity facilitates quick visual assessment of the overall opportunity within a country and the relative opportunity between countries. Finally, summary metrics synthesise data from various indicators, allowing quantitative ranking of countries based on priority areas, and facilitating more objective decision-making about where to implement an IPC. We acknowledge important limitations to our analysis. First, many factors could potentially influence the level of opportunity a given country has for IPC implementation. In our indices, we only included factors, such as disease burden and existing intervention coverage, which clearly have a large effect on the potential impact of an IPC intervention. Second, our list of opportunity indicators was limited by the availability of published data. Given our interest in examining indicators on a cross-country basis, we required standardised metrics reported by all countries, which may have resulted in the selection of less than ideal metrics for some variables. For example, we included data on the coverage of HIV testing in antenatal care settings, a widely and routinely collected indicator, whereas a more suitable assessment of existing levels of HIV testing would be based on coverage in the general population. However, such data was unavailable for many countries. We also assessed opportunity at the country level due to limited availability of regional data for all variables. There may be areas within a country with opportunities that depart radically from the overall country assessments to which our analysis is confined. Consequently, the rankings presented here could overlook the true opportunity for IPC implementation in particularly high-burden and low-coverage subregions of countries. In countries where such variation in burden, access and coverage are known to exist, collection of regional data and application of this type of index would help to identify regions for targeted campaign introduction, channelling resources to areas in greatest need. Finally, the choice of weighting when calculating the CBIO scores was subjective. However, if we were to weight disease burden by a factor of three instead of two, the overall composition of the top 23 countries ranked by CBIO score would be the same, although the relative order would change somewhat. Even when weighting disease burden and intervention opportunity equally, 16 countries would remain on the list of top 23. Conclusion {#s4a} ---------- Prior assessments have shown that IPCs can rapidly increase the uptake of communicable disease interventions, representing a promising strategy to accelerate progress in meeting MDGs. The index presented here provides a data-driven tool by which to prioritise countries for implementation of an IPC for diarrhoea, malaria and HIV. Application of this opportunity index, in conjunction with other stakeholder-specific assessments (eg, funding, feasibility, etc), may facilitate more objective decision-making regarding areas where IPC scale-up would yield the most value and lead to a more efficient use of resources. Supplementary Material ====================== ###### Author\'s manuscript ###### Reviewer comments The authors would like to thank Mikkel Vestergaard Frandsen and Navneet Garg for their thoughtful input into this project. The opportunity index method used in this paper was adapted from a similar index created by the Strategic Analysis, Research, and Training Program (START) at the University of Washington. The authors thank the authors of the original opportunity index project, Matheson AI, Manhart LE, Pavlinac PB, Means AR, Akullian A, Levine GA, Jacobson J, Shutes E, Walson JL, for their work on developing the original method. **Contributors:** AJ helped design the study, conducted the analyses and drafted and revised the paper. AM provided data for the study, helped with the analyses and revised the draft paper. JGK and EM helped guide design and implementation of the study, and edited the paper. AR and SV critiqued the analysis and revised the draft paper. JW conceived and guided design and implementation of the study, and edited the paper. **Funding:** This work was supported by Vestergaard Frandsen. **Competing interests:** JW, AJ, JGK and EM were contracted by Vestergaard Frandsen (VF) to conduct this analysis. AR was formerly employed by VF as an HIV/AIDS Advisor. JW and JGK have had additional consultancies with VF outside of this analysis. VF supported author SV\'s travel and expenses for the 2012 International AIDS Society conference. **Provenance and peer review:** Not commissioned; externally peer reviewed. **Data sharing statement:** The three complete opportunity indices for all 70 countries in our sample are available in online data supplement 2. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

{#sp1 .230} {#sp2 .231}

INTRODUCTION {#S1} ============ Acetylation of lysine residues in histone tails activates gene transcription by several mechanisms, including promotion of an open chromatin structure as well as attracting positively acting transcription factors to the genome^[@R1]--[@R5]^. Histone deacetylases (HDACs) catalyze the reversal of this process, promoting a closed state of chromatin and thus contributing to repression of gene expression^[@R1],[@R6]--[@R8]^. HDACs are widely expressed in metazoans, and are categorized as classical HDACs 1--11 as well as NAD-dependent HDACs known as sirtuins^[@R9]^. HDACs regulate many different biological processes including embryo development^[@R10]^, cell cycle regulation^[@R11]--[@R13]^, cell proliferation^[@R12]^, cancer progression^[@R14],[@R15]^, lipid metabolism^[@R16],[@R17]^, circadian rhythm^[@R16]^, macrophage alternative activation^[@R18]^, etc. HDAC3 is a Class I HDAC that has been suggested to interact with many protein partners. Unlike other HDACs, HDAC3 is present in specific complexes containing nuclear receptor corepressors NCOR1^[@R19]--[@R22]^ and NCOR2 (also known as [S]{.ul}ilencing [M]{.ul}ediator of [R]{.ul}etinoid and [T]{.ul}hyroid receptors, or SMRT)^[@R19],[@R21]--[@R24]^, and regulates many biological processes in a variety of tissues, as demonstrated by shRNA^[@R17],[@R25]^ and knockout (KO)^[@R26]--[@R28]^. For example, deletion of HDAC3 alters lipid metabolism in liver and heart^[@R17],[@R25],\ [@R10],\ [@R41]^, the polarization state of macrophages^[@R18]^, and the cell cycle progression of fibroblasts^[@R28]^. HDAC3 interacts with NCOR1 and SMRT via a conserved repression domain containing a SANT ([S]{.ul}WI3, [A]{.ul}DA2, [N]{.ul}CoR, [T]{.ul}FIIB) motif which is similar to that found in other transcriptional regulators^[@R29],[@R30]^. This SANT motif together with a unique N-terminal helical extension^[@R31]^ is required to activate the catalytic function of the HDAC3 enzyme^[@R20],[@R32]^, and is thus referred to as the Deacetylase Activating Domain (DAD). NCOR1 and SMRT activate HDAC3 both *in vitro*^[@R24],[@R33]^ and *in vivo*^[@R34],[@R35]^ via the DAD. The structure of HDAC3 bound to the SMRT DAD was recently solved, revealing the details of the atomic interactions as well as the involvement of an inositol tetraphosphate (IP4) molecule in stabilizing the complex^[@R36]^. Of particular note, a tyrosine residue at position 470 (Y470) of SMRT, corresponding to Y478 of NCOR1 and previously shown to be required for interaction with HDAC3^[@R31],[@R33]^, makes a critical contact with the IP4^[@R36]^. The mutation of Y478 of NCOR1 to alanine (Y478A) abrogates repression by thyroid hormone receptor *in vivo*^[@R34],[@R35],[@R37]^, and alters circadian and metabolic physiology in living mice^[@R34]^. While HDAC3 has been reported to contribute to many biological processes, the relation of these functions to NCOR1 and SMRT has been unclear. Indeed, although NCOR1 and SMRT are competent to activate the latent deacetylase activity of HDAC3, and other SANT domains tested to date have lacked this property^[@R33]^, the question of whether interaction with NCOR1 or SMRT is required for HDAC3 activity *in vivo* remained open. To address this important question, we have generated mice homozygous for the Y470A mutation in SMRT (S-DADm) and crossed these with mice homozygous for the Y478A mutation in NCOR1 (N-DADm) to generate mice lacking functional DADs in both NCOR1 and SMRT (NS-DADm). Using this mouse model, we report here for the first time that the enzyme activity of endogenous HDAC3 indeed requires the DAD of either the NCOR1 or SMRT. RESULTS {#S2} ======= NS-DADm mice are born and live to adulthood {#S3} ------------------------------------------- To determine whether NCOR1 and SMRT are required for HDAC3 enzyme activity *in vivo*, we generated C57BL/6 mice bearing mutations in the DAD of both alleles of *Ncor1* and *Ncor2.* In each case, the mutant protein contains a single tyrosine to alanine substitution in the DAD (Y478A in NCOR1, Y470A in SMRT) that prevents interaction with and activation of HDAC3^[@R33]^. The *Ncor1* DAD mutant C57BL/6 mice (N-DADm) have been described previously^[@R34]^, and the *Ncor2* DAD mutant mice (S-DADm) were generated in the C57BL/6 strain background using a similar strategy ([Supplementary Figure 1](#SD1){ref-type="supplementary-material"}). Mutants heterozygous of N-DADm and S-DADm were mated to generate mice that are heterozygous for both mutant alleles, and those mice were mated with each other to obtain male and female double homozygous mutant (referred to as NS-DADm) mice. Since this breeding strategy produces only 1 NS-DADm and 1 wild type (WT) of the same gender for every 32 pups, larger numbers of each genotype were generated by crossing WT and NS-DADm with each other. Interestingly, while loss of NCOR1, NCOR2, or HDAC3 is embryonically lethal^[@R26],[@R28],[@R38],[@R39]^, the NS-DADm mice exhibited no detectable embryonic lethality and lived to adulthood ([Table 1](#T1){ref-type="table"}). HDAC3 is enzymatically inactive in NS-DADm mice {#S4} ----------------------------------------------- We next studied the expression and function of HDAC3 in the tissues of mice bearing the DAD mutations in NCOR1 and SMRT. Gene expression of *Hdac3* is normal in the livers of NS-DADm mice ([Figure 1a](#F1){ref-type="fig"}). Similarly, levels of hepatic HDAC3 protein were indistinguishable from those of WT mice ([Figure 1b](#F1){ref-type="fig"}). Since expression of HDAC3, NCOR1, and SMRT was not significantly altered by the presence of the DAD mutations ([Supplementary Figures 2a and 2b](#SD1){ref-type="supplementary-material"}), we were able to test the hypothesis that endogenous HDAC3 requires NCOR1 or SMRT for its activity *in vivo.* Remarkably, whereas HDAC3 activity was readily measured in immunoprecipitates from WT liver, it was undetectable in liver from the NS-DADm mice ([Figure 2a](#F2){ref-type="fig"}). Importantly, this was not due to inefficient immunoprecipitation relative to WT ([Figure 2b](#F2){ref-type="fig"}). Similar loss of HDAC3 enzyme activity was observed in heart ([Figure 2c](#F2){ref-type="fig"}) and skeletal muscle ([Figure 2d](#F2){ref-type="fig"}). Moreover, no HDAC3 enzyme activity was detectable in embryos harvested on day 12.5 ([Figure 2e](#F2){ref-type="fig"}), demonstrating the importance of the NCOR1 and SMRT DADs in all tissues, and that no other factor substitutes prenatally as an HDAC3 activator. Due to the background of the HDAC enzyme assay, it is possible that a small amount of residual activity exists. However these data prove that the NCoR and SMRT are required for the vast majority of the HDAC3 enzymatic activity in the tissues examined. Thus, the nuclear receptor corepressors are required for HDAC3 enzyme *in vivo*. HDAC3 genome binding is reduced in the NS-DADm mice {#S5} --------------------------------------------------- Since HDAC3 is thought to be recruited to the genome by NCOR1 and SMRT, we hypothesized that this would be reduced in the NS-DADm mice. To test this, we located and quantitated the recruitment of HDAC3 to mouse liver using chromatin immunoprecipitation with HDAC3-specific antibody followed by massively parallel DNA sequencing (ChIP-seq). At 5 PM, when genomic recruitment is maximal in mouse liver^[@R16]^**,** we detected HDAC3 at 5799 sites in WT mice, the majority of which were distant from transcription start sites or present in introns ([Supplementary Figure 3](#SD1){ref-type="supplementary-material"}), consistent with prior findings^[@R40]^. By contrast, using the same stringent peak calling criteria, only 600 HDAC3 binding regions were detected in the NS-DADm liver, the majority of which overlapped with WT binding ([Supplementary Figure 4](#SD1){ref-type="supplementary-material"}). It should be noted that HDAC3 binding remained detectable at most sites. The strength of binding in the NS-DADm liver decreased \~62.4% on average ([Figure 3a](#F3){ref-type="fig"}), and individual HDAC3 binding sites reflect this decrease ([Figure 3b](#F3){ref-type="fig"}). The reduction of HDAC3 recruitment in the NS-DADm liver was validated at 10 sites by ChIP-qPCR ([Figure 3c](#F3){ref-type="fig"}). The partial genomic interaction of HDAC3 is likely due to another region of NCOR1 or SMRT^[@R33],[@R41]^, or to direct interaction between HDAC3 and other transcription factors, neither of which would activate the HDAC3 enzyme. Nevertheless, these data demonstrate that the NCOR1 and SMRT DADs are critical for normal genomic recruitment of HDAC3 recruitment *in vivo*. Local histone acetylation is increased in NS-DADm liver {#S6} ------------------------------------------------------- Deletion of HDAC3 has been shown to result in increased local histone acetylation in liver^[@R16]^ and macrophages^[@R18]^, which is consistent with its *in vitro* histone deacetylase activity. We next tested whether the loss of HDAC3 activity alter local histone acetylation *in vivo.* Indeed, histone H3 lysine 9 acetylation (H3K9Ac), an activating mark^[@R42]^, was locally increased at the 10 sites where loss of HDAC3 recruitment was validated by ChIP-qPCR, but not at control sites in the *Arbp* and *Ins* genes where HDAC3 is not bound ([Figure 4a](#F4){ref-type="fig"}). Similar results were obtained after analysis of H3K27Ac, another activating mark ([Figure 4b](#F4){ref-type="fig"}). Of note, the degree of elevated acetylation is comparable to that in the HDAC3 KO liver ([Supplementary Figure 5a](#SD1){ref-type="supplementary-material"}). Loss of HDAC3-DAD interaction did not alter the genomic recruitment of NCOR1 ([Figure 4c](#F4){ref-type="fig"}) or SMRT ([Supplementary Figure 5b](#SD1){ref-type="supplementary-material"}), nor did the complete absence of HDAC3 affect the genomic recruitment NCOR1 or SMRT ([Supplementary Figures 5c and 5d](#SD1){ref-type="supplementary-material"}). Thus, the loss of HDAC3-DAD interaction leads to increased local nucleosomal histone acetylation in NS-DADm mice, despite normal binding of NCOR1 and SMRT at these genomic locations. NS-DADm mice do not phenocopy mice lacking HDAC3 protein {#S7} -------------------------------------------------------- The viability of the NS-DADm, which carry the mutant alleles in every cell, was quite different from the embryonic lethality of mice with germ line deletion of HDAC3^[@R38],[@R39]^, despite the fact that HDAC3 activity was undetectable in embryos from NS-DADm mice ([Figure 2e](#F2){ref-type="fig"}). Since a major difference between these two genetic models is that HDAC3 protein is absent in the knockout but fully present in the NS-DADm mice, this result suggested that embryonic lethality of loss of HDAC3 is due to a deacetylase activity-independent function. We also compared the livers of NS-DADm mice with those lacking HDAC3 in liver. Deletion of hepatic HDAC3 ("Liver HDAC3 KO") was accomplished by injecting 10 week old male C57BL/6 mice bearing floxed *Hdac3* alleles (HDAC3^f/f^) with an adeno-associated virus^[@R43]^ in which the thyroxine-binding globulin (*Tbg*) promoter drives the liver-specific expression of either Cre recombinase or green fluorescent protein (GFP) as control. As previously described, the loss of HDAC3 in liver dramatically increased hepatic triglyceride levels^[@R10],[@R16],[@R17]^. Both mouse models exhibit undetectable hepatic HDAC3 activity ([Figure 5a](#F5){ref-type="fig"}) and increased local histone acetylation at the lost binding sites^[@R16]^ ([Figures 3](#F3){ref-type="fig"} and [4](#F4){ref-type="fig"}), as well as increased liver triglycerides. However, the \~2-fold increase of triglycerides activity in the NS-DADm livers was considerably more modest than the dramatic 5- to 10-fold increase observed in the mice lacking HDAC3 protein in liver ([Figures 5b and 5c](#F5){ref-type="fig"}). This suggests that the continued presence of the inactive HDAC3 protein in the NS-DADm mice serves a function that normally contributes to the overall biological activity of HDAC3. Indeed, whereas hepatic cholesterol levels increased in the liver HDAC3 KO**,** no significant change in cholesterol accumulation was observed in the NS-DADm mice ([Figures 5d and 5e](#F5){ref-type="fig"}), suggesting that the cholesterol accumulation in the absence of HDAC3, which has been also observed by others^[@R10]^, involves a deacetylation-independent function of HDAC3. Consistent with the physiological data, comparisons of transcriptomes revealed that the loss of HDAC3 in liver had more dramatic effects on gene expression than loss of HDAC3-DAD interaction in NS-DADm mice ([Figure 6a](#F6){ref-type="fig"}; the complete list of significantly changed genes is available at GEO). In keeping with previous results^[@R17]^, pathway analysis highlighted the upregulation of many genes associated with lipid metabolism in the livers lacking HDAC3 protein ([Figure 6b](#F6){ref-type="fig"}), many of which could be confirmed by RT-qPCR ([Figure 6c](#F6){ref-type="fig"}). By contrast, abrogation of HDAC3 enzyme activity but not the HDAC3 protein in NS-DADm liver caused upregulation of fewer genes involved in lipid metabolism ([Figures 6b and 6d](#F6){ref-type="fig"}). These results reinforce the conclusion that while the NCOR1 and SMRT DADs control HDAC3 activity and local histone deacetylation, these effects contribute relatively modestly to the total effects of HDAC3 protein on hepatic gene expression and the physiology of liver lipid metabolism. DISCUSSION {#S8} ========== We have introduced point mutations into both alleles of NCOR1 and SMRT that specifically abolish their ability to activate HDAC3. Using this unique mouse model we show that the DADs of nuclear receptor corepressors NCOR1 and SMRT are required for HDAC3 enzyme activity *in vivo*. This finding is of great importance as HDAC3 has been shown to play pivotal roles in transcriptional regulation^[@R1],[@R6],[@R7]^, cell cycle progression^[@R11]--[@R13]^, inflammation^[@R18]^, developmental events^[@R10]^, and metabolism^[@R16],[@R17]^. Our data clearly show that NCOR1 or SMRT is required for nearly all HDAC3 enzyme activity *in vivo*. HDAC3 clearly functions as an epigenomic modifier in the liver^[@R16]^, and indeed histone acetylation was increased in the NS-DADm liver at genomic locations where HDAC3 normally bind. This is most likely explained by the loss of HDAC3 catalytic activity. In addition, HDAC3 occupancy on the genome was significantly reduced in the NS-DADm, demonstrating the importance of the NCOR1 and SMRT DADs in recruiting HDAC3 to the genome, and providing a second mechanism for increased histone acetylation at site of endogenous HDAC3 recruitment in WT mice. Histone acetylation and deacetylation alters chromosome accessibility and affect functions of transcription factors (TFs) acting at the genome^[@R2]^. Many inhibitors of the enzyme activity of class I HDACs are being developed to treat diseases including several types of cancer^[@R44],[@R45]^. Moreover, approximately 10% of currently prescribed drugs directly target TFs^[@R46]^, including tamoxifen for breast cancer and bicalutamide for prostate cancer, which target nuclear receptors by regulating their interaction with HDAC3^[@R47],[@R48]^. Therefore our findings towards understanding basic HDAC biology have important therapeutic implications. The NS-DADm mice exhibited mild hepatic steatosis, and molecular analysis revealed reduced or absent HDAC3 binding and increased local histone acetylation at upregulated lipid metabolic genes. Therefore these effects are likely due to the absence of HDAC3-dependent histone deacetylation. Nevertheless, mice lacking HDAC3 in liver manifest much more severe hepatic steatosis^[@R10],[@R16],[@R17]^ along with disrupted cholesterol homeostasis^[@R10],[@R17]^, whereas NS-DADm mice exhibited no detectable alteration in hepatic cholesterol. Furthermore, while absence of cardiac HDAC3 causes lethality or diet-induced heart failure depending on when the HDAC3 is deleted^[@R17],[@R39]^, NS-DADm mice have normal hearts and are able to tolerate a high fat diet (data not shown). Thus, while our studies show for the first time that the nuclear receptor corepressors are required for the deacetylase activity of HDAC3 *in vivo*, they also suggest a DAD-independent or deacetylase-independent role of HDAC3. In addition to their roles in adult tissues, germ line deletion of NCOR1^[@R26]^, SMRT^[@R38]^, or HDAC3^[@R10],[@R39]^ all cause embryonic lethality. The viability of the N-DADm mice demonstrated that the DAD is not required for the essential functions of NCOR1^[@R34]^ during embryonic development. The viability of the S-DADm mice makes this point about SMRT, which has a variety of functions besides HDAC3 activation^[@R49]--[@R52]^. The phenotype of the S-DADm mice will be further characterized. More remarkably, however, although HDAC3 enzyme activity is diminished to an almost undetectable level in embryos of NS-DADm, the mice are viable and live to adulthood. They also suggest that HDAC3 has a DAD-independent or deacetylase-independent function that is required for life. The DAD-independent functions of HDAC3 are unlikely to be the regulation of the genomic recruitment of NCOR1 and SMRT, since the genomic occupancy of these corepressors was maintained in the livers of NS-DADm as well as in mice lacking hepatic HDAC3. Also, while it is possible that HDAC3 has nuclear receptor corepressor-independent ability to deacetylate non-histone substrates^[@R53]--[@R58]^, this is unlikely because the substrate used in the HDAC3 assay is a short peptide rather than a full-length histone protein. One caveat is that, due to the relatively high background of the HDAC activity assay, it is also possible that there is a small amount of residual HDAC3 deacetylase activity that is not detectable over background but may contribute to the modest phenotype of NS-DADm mice relative to mice lacking HDAC3 protein. It should also be noted that the genome-wide localization of HDAC3 was reduced but not abrogated in NS-DADm mice, potentially due to the second region in NCOR1 and SMRT that interacts with HDAC3 but does not activate the enzyme^[@R33]^. Nonetheless, the present study in NS-DADm mice demonstrates for the first time that corepressors NCOR1 and SMRT are required for endogenous HDAC3 activity. It also raises the possibility that the HDAC3 protein is critical for embryonic development as well as adult physiology through non-enzymatic mechanisms. ONLINE METHODS {#S9} ============== Mice {#S10} ---- NS-DADm were generated from crossing N-DADm and S-DADm and the generation of mice lacking HDAC3 in liver by injection of AAV-TBG-Cre into HDAC3^f/f^ mice has been previously described^[@R16],[@R17]^. Mice were housed under a 12 h light and 12 h dark cycle (lights on at 7 a.m. and lights off at 7 p.m.). We used adult male mice at the age of 3--7 months in all experiments, except where otherwise indicated. We harvested tissues at 5 p.m. without restricting the mice to food or water, unless otherwise indicated. All the animal care and use procedures followed the guidelines of the Institutional Animal Care and Use Committee of the University of Pennsylvania in accordance with the guidelines of the US National Institutes of Health. Antibodies and reagents {#S11} ----------------------- HDAC3 antibodies for ChIP-seq, immunoprecipitation, and western blot were purchased from Abcam (ab7030) and Millipore (05-813, clone 3G6). Acetylated H3K9 and H3K27 antibodies were purchased from Millipore (07-352) and Abcam (ab4729), respectively. IgG was purchased from Sigma (I8140). Immunoprecipitation and Western blotting {#S12} ---------------------------------------- Liver tissue was homogenized in modified radioimmunoprecipitation assay (RIPA) buffer supplemented with protease inhibitors. Lysates were incubated on ice for 1 hr and clarified by centrifugation. Supernatants were precleared and incubated with HDAC3 antibody at 4°C overnight followed by 1hr incubation with Protein A agarose beads at 4°C. Immunoprecipitates were washed 5X with modified RIPA, eluted by boiling in SDS-loading buffer and subject to immunoblot analysis. For the western blot of the total lysates, tissues were lysed in modified RIPA buffer supplemented with protease inhibitors, and the samples were resolved by Tris-glycine SDS-PAGE, then transferred to polyvinylidene fluoride (PVDF) membranes and blotted with the indicated antibodies. All antibodies were used at 1:1,000 to 1:5,000 dilutions. HDAC activity assay {#S13} ------------------- Immunoprecipitation with HDAC3 antibody and control IgG were performed as described above. Immunoprecipitates were washed 2X with modified RIPA buffer and 2X with 1X PBS followed by HDAC assay according to the manufacture's instruction (Active Motif, 56200) ChIP-seq {#S14} -------- ChIP was performed independently on liver samples from 4 or 5 different mice. Detailed procedures were previously described^[@R16],[@R18]^. High throughput sequencing was done by the Functional Genomics Core (J. Schug and K. Kaestner) of the Penn Diabetes Research Center using the Illumina Genome Analyzer IIx, and sequence reads were mapped to the mm8 mouse genome using ELAND pipeline. In each ChIP-seq sample, all the duplicate reads were removed except for one for each genomic position. Computational analysis for HDAC3 peaks {#S15} -------------------------------------- Peak calling was carried out using HOMER^[@R59]^ with a default option (FDR=0.001) on HDAC3 WT and NS-DADm samples against the matching input sample, and then 1 RPM cutoff was applied. In case of wild-type samples, two replicates were pooled into one before the peak calling, so the maximum tag counts per position were set to be two (-tbp 2). HDAC3 binding sites were annotated with the following priority. 1) pTSS (proximal promoter, from −1 kb to 100 b around TSS), 2) TTS ( from −100 b to 1 kb around TTS), 3) Exon, 4) Intron, 5) dTSS (distal promoter, from −10 kb to +1 kb around TSS) and 6) Intergenic (everything else). The heatmaps of normalized HDAC3 binding profiles surrounding peak centers were generated using heatmap.2 function in R package. Quantitative PCR {#S16} ---------------- Quantitative PCR was performed with Power SYBR Green PCR Mastermix and the PRISM 7500 instrument (Applied Biosystems), and analyses were performed by the standard curve method. Primer sequences provided in [Supplementary Table 1](#SD1){ref-type="supplementary-material"}. Tissue TG and cholesterol {#S17} ------------------------- Liver samples were homogenized in tissue lysis buffer (140mM NaCl, 50mM Tris and 1% Triton-X, pH8.0). TG and cholesterol concentration in the lysates were then quantified using LiquiColor Triglyceride Procedure No. 2100 (Stanbio) and Cholesterol LiquiColor Test (Stanbio), respectively. Microarray {#S18} ---------- Total RNA was extracted from liver using RNeasy tissue Mini kit (QIAGEN) according the manufacturer's instructions. Preparation of RNA for hybridization to Affymetrix MoGene 1.0 ST (Affymetrix, Santa Clara, CA) and scanning of the arrays were performed by the University of Pennsylvania Microarray Facility (<http://www.bioinformatics.upenn.edu/index.html>) according to the manufacturer's instructions. Robust Multiarray Averaging (RMA) signal extraction, normalization, and filtering were performed by the Microarray Facility's bioinformatics group (<http://core.pcbi.upenn.edu/>) using Partek Genomics Suite (Partek, St. Louis, MO). Supplementary Material {#S19} ====================== This work was supported by R37DK43806 from NIDDK, and by a Mentor Based Fellowship from the American Diabetes Association. We thank Theresa Alenghat for early versions of the SMRT DAD targeting construct, and K. Kaestner for help with gene targeting in C57BL/6 embryonic stem cells. We also acknowledge the Functional Genomics Core of the Penn Diabetes Research Center (DK19525), directed by K. Kaestner and J. Schug, for next generation sequencing. **ACCESSION CODES** These have been requested and will be added when available. **AUTHOR CONTRIBUTIONS** S-H.Y. and M.A.L. conceived of the hypothesis and designed the experiments. S-H.Y., Z.S., and M.B. performed the experiments. H-W.L. and K-J.W. analyzed bioinformatics data. S-H.Y., Z.S., and M.A.L. analyzed and interpreted the data. S-H.Y. and M.A.L. wrote the manuscript. {#F1} {#F2} {ref-type="supplementary-material"}. All error bars = s.e.m. \**p* \< 0.05, \*\**p* \< 0.01, and \*\*\**p* \< 0.001 by Student's two-tailed *t* test.](nihms424378f3){#F3} {ref-type="fig"}. (a) H3 lysine 9 acetylation. (H3K9Ac) (b) H3 lysine 27 acetylation (H3K27Ac). (c) NCOR1. N=4, All error bars = s.e.m. \**p* \< 0.05, \*\**p* \< 0.01, and \*\*\**p* \< 0.001 by Student's two-tailed *t* test.](nihms424378f4){#F4} {#F5} {#F6} ###### NS-DADm mice are viable without excess mortality. Genotype \# of matings \# of pups born Average pups/mating ---------- --------------- ----------------- --------------------- WT 12 77 6.4 NS-DADm 13 87 6.7

Introduction {#sec1_1} ============ Undiagnosed asthmatic patients who present with cough, wheezing, shortness of breath and other common respiratory symptoms are sometimes mistakenly diagnosed with upper or lower respiratory tract infections \[[@B1]\]. In children with asthma-like symptoms, such as recurrent episodes of wheezing, cough and shortness of breath, it has been reported that treatments more often consist of antibiotics and cough medicines than of asthma drugs \[[@B2]\]. Contrary to the recommendations in the international guidelines, antibiotics are often prescribed instead of drugs to treat asthma \[[@B1]\]. Excessive antibiotic usage by asthmatic patients has been reported in the literature \[[@B1],[@B2],[@B3],[@B4],[@B5]\]. Adoctor\'s education, training and experience can result in an appropriatediagnosis, which in turn can reduce antibiotic usage by asthmatics \[[@B6]\]. Knowledge of appropriate medication as recommended in the guidelines could result in fewer prescriptions of antibiotics for asthma patients \[[@B1]\], and the education and regular follow-up of patients also improve asthma self-management. This study compared the frequency of antibiotic usage and the number of asthma episodes before and after the diagnosis and treatment of asthma. Subjects and Methods {#sec1_2} ==================== Study Design and Patients {#sec2_1} ------------------------- Included in the study were 334 patients (2--16 years of age) who were diagnosed with asthma and admitted to the Divisions of Pediatric Allergy and Pediatric Pulmonology, Faculty of Medicine, Bezmialem Vakif University. Their demographic data and medical history for the year prior to diagnosis were obtained from parent reports at the first visit and retrospectively from a national database. After enrollment into the study, patients were prospectively followed up between September 2012 and September 2013. Those with immunodeficiency or on antibiotic prophylaxis for other diseases such as latent tuberculosis or repeating urinary tract infections were excluded. The study was approved by the Research Ethics Committee of the Bezmialem Vakif University. Signed informed consent forms were obtained from the families of all patients. Asthma Diagnosis {#sec2_2} ---------------- The diagnosis of asthma was based on symptoms and medical history supported by laboratory findings. In patients younger than 5 years, the modified Asthma Predictive Index and the 2011 Global Initiative for Asthma (GINA) report were used \[[@B7],[@B8]\]. Patients were considered positive for asthma if they had a history of ≥4 wheezing episodes in 1 year. Besides this primary threshold, they had to fulfill at least one major or two minor criteria. The major criteria included a parental history of asthma, physician-diagnosed atopic dermatitis and allergic sensitization to at least one aeroallergen. The minor criteria included wheezing unrelated to colds, peripheral blood eosinophils (≥4s%) and allergic sensitization to milk, eggs or peanuts. Skin prick tests, total immunoglobulin E levels, blood eosinophil counts and specific immunoglobulin E status (inhalant and food allergens) were used to evaluate the patients according to the criteria. For patients older than 5 years, the GINA recommendations were used. Besides a family history of asthma or atopic disease, the suspicion of asthma increased in the presence of signs and symptoms which responded to antiasthma therapy \[[@B8]\]. Pulmonary function tests were used to measure airflow limitation and its reversibility. Skin prick tests, total immunoglobulin E levels, blood eosinophil counts and specific immunoglobulin E status (inhalant and food allergens) were also performed to establish an asthma diagnosis. Evaluation of the Patients {#sec2_3} -------------------------- At the first visit, simple questionnaires were filled out by a specialist, either a pediatric pulmonologist or an allergologist. These forms contained the demographic characteristics of the patients, their age at diagnosis, the parents\' level of education, a history of smoking in the household, a history of atopy, the number of upper and lower airway infections (e.g. bronchitis and pneumonia) in the year prior to being diagnosed with asthma, the number of episodes of transient airway obstruction or respiratory distress and the frequency of antibiotic usage. Turkish children younger than 18 years have government health insurance. In all health institutions including public and private hospitals as well as primary settings, children can be examined and given prescriptions free of charge. In addition, these prescriptions have to be entered into a database controlled by the Ministry of Health. It is against the law to buy antibiotics from pharmacies without digital prescriptions. The database also contains patients\' International Statistical Classification of Disease and related problems (ICD) codes. Data about wheezing, transient airway obstruction or respiratory distress episodes in the year prior to the diagnosis of asthma were obtained from the database. Antibiotic usage and respiratory infection data for this period were taken from the same database and confirmed by the parents at the first visit. The antibiotics used were obtained from digital prescriptions and parent reports; most common were amoxicillin-cluvulanate, cefuroxime-axetil, clarithromycin, ampicillin-sulbactam and azithromycin. After diagnosis, patients were classified into 4 groups depending on their symptoms. Group 1 had mild and intermittent symptoms, group 2 had mild and persistent symptoms, group 3 had moderate and persistent symptoms and group 4 had severe symptoms. Patients received treatments based on the GINA recommendations \[[@B8]\]. The 80 (24s%) patients in group 1 received rapid-acting β~2~-agonist as needed. The 184 (55s%) patients in group 2 were given low-dose inhaled corticosteroids (ICS). According to the severity of the symptoms, low-dose ICS plus a leukotriene modifier or low-dose ICS plus a long-acting β~2~-agonist were used in the 53 (16s%) patients enrolled in group 3. The 17 subjects in group 4 were given medium- or high-dose ICS plus a long-acting β~2~-agonist. All patients were routinely evaluated every 3 months, during acute exacerbations and for other complaints. At every visit, for each patient, the number of asthma episodes, antibiotic usage and the occurrence of respiratory infections were obtained from the parent reports and database and recorded. In addition, patients were educated about asthma and asthma medication, and an experienced nurse educated the patients about techniques for the inhalation of asthma drugs. At the end of 1 year, the frequency of antibiotic usage, the number of asthma episodes and upper or lower airway infections were again recorded. Statistical Analysis {#sec2_4} -------------------- SPSS version 15.0 was used for analysis. The numerical parameters were described as the mean, median and standard deviation; distributions of the categorical measurements were determined by frequencies and percentages. The one-sample Kolmogorov-Smirnov test was used to evaluate the distributions of the number of episodes and frequency of antibiotic usage before and after diagnosis as well as during the treatment of asthma. As no normal distributions were found in either variable, the Wilcoxon signed-ranks test was used to compare differences between before and after diagnosis and during treatment with regard to the number of episodes and frequency of antibiotic use. Differences or change in the frequency of antibiotic use and the number of asthma episodes before and after the diagnosis and treatment of asthma were determined and written in two other columns in the SPSS. The Spearmans correlation test and the Mann-Whitney U test showed associations between other factors such as gender, age at diagnosis, number of people in the household, parents\' education levels, smoking status in the household, atopy history and any changes in the frequency of antibiotic usage or the number of asthma episodes. Multivariate linear regression analyses were used to evaluate the effect modification according to these factors. The multivariate analyses were adjusted for all factors simultaneously. p \< 0.05 was considered to be statistically significant. Results {#sec1_3} ======= The median age was 84 months (range 24--192) and the median age at diagnosis was 48 months (range 24--156). Of the 334 children, 212 (63s%) were at school or in day care centers. Household size was 4 ± 1, and there was at least one smoker in the households of 137 (41s%) of the patients. A family history of doctor-diagnosed asthma and atopy was present in 200 (60s%) of the patients. The mothers of 20 patients (6s%) and the fathers of 33 patients (10s%) had graduated from university. Before study enrollment, all patients had had at least four wheezing episodes in their lives, 207 (62s%) had inhaled-allergen positivity, 187 (56s%) had high immunoglobulin E levels and 160 (48s%) had eosinophilia. Allergic rhinitis symptoms were seen in 160 (48s%) patients, 77 (23s%) had atopic dermatitis and 40 (12s%) had food allergies. The patients were treated according to the GINA recommendations \[[@B8]\]; all took their drugs regularly and 301 (90s%) came to their evaluations every 3 months. The characteristics of the patients are shown in table [1](#T1){ref-type="table"}. For the year prior to diagnosis, 11 (3.2s%) parent reports about antibiotic usage and respiratory infections differed from what was recorded in the database. The number of times antibiotics were used differed by 1 in 9 of the patients and by 2 in 2 patients; such differences were due to the uncontrolled and spontaneous use of antibiotics prescribed at previous consultations. The agreement was 96.8s% between the parent reports and the database for the year prior to the diagnosis of asthma. For the year after the diagnosis, there were no differences, i.e. the agreement was 100s%. The median number of upper airway diseases before diagnosis of asthma was 5 \[interquartile range (IQR) = 6\] per year which decreased to 1 (IQR = 3) after treatment (p \< 0.001). The median number of lower airway diseases before diagnosis of 2 (IQR = 5) per year dropped to 0 after treatment (p \< 0.001). The annual median number of asthma episodes before the asthma diagnosis was 4 (IQR = 8), decreasing to 0 (IQR = 2) in the year after treatment (p \< 0.001; fig. [1](#F1){ref-type="fig"}). The median number of antibiotic administrations was 7 (IQR = 6) per year before diagnosis and 2 (IQR = 3) in the year following treatment (p \< 0.001; fig. [2](#F2){ref-type="fig"}). No statistically significant association was found between change in the frequency of antibiotic usage and number of asthma episodes before and after the diagnosis and gender, diagnosis age, number of household members, parents\' education levels, household smoking status or history of atopy (table [2](#T2){ref-type="table"}). Discussion {#sec1_4} ========== This study showed that the frequency of antibiotic usage and the number of asthma episodes decreased to a statistically significant level after patients received appropriate treatment and regular follow-up by specialists. The asthmatic children had more doctor visits, due to the fact that they are more symptomatic; this has been reported previously \[[@B3],[@B9]\]. The possibility of misdiagnosing an acute asthma episode together with symptoms of increasing severity, like bronchiolitis, atypical pneumonia and other respiratory tract infections, means clinicians may choose to treat multiple possible etiologies \[[@B1]\]. Generally, 35s% of asthmatics who experience episodes of respiratory tract infections are treated with antibiotics \[[@B10]\], even though these infections are most commonly viral in origin and trigger wheezing in young children \[[@B11]\]. Prior to admission to our clinic, the children had been frequently diagnosed with upper and lower airway infections, and so the use of antibiotics was high, which is similar to previous studies \[[@B1],[@B12]\]. Reasons for using antibiotics include diagnostic uncertainty, the prophylaxis of secondary infections, an attempt to utilize the anti-inflammatory properties of macrolide antibiotics and the belief that unknown or noncultivable bacteria may be important in some asthmatic patients \[[@B1],[@B12]\]. Marra et al. \[[@B6]\] demonstrated that children eventually diagnosed with asthma had a higher rate of antibiotic usage than those who had never been diagnosed with asthma. In contrast to the outcomes of routine examinations, international guidelines specify that antibiotics should not be used for chronic asthma therapy or for acute exacerbations \[[@B8]\]. After enrollment, patients experiencing exacerbations were examined without delay as well as at regular follow-up appointments. If no evidence of bacterial infection was detected, we did not use antibiotics; exacerbations were treated with asthma drugs. Numerous studies have assessed the role of antibiotics as a part of asthma therapy \[[@B13],[@B14]\]. De Boeck et al. \[[@B3]\] mentioned the strong tendency for health care providers to coprescribe antibiotics and asthma drugs. Paul et al. \[[@B1]\] determined that each year in the USA, antibiotics are prescribed for nearly 16s% of patients with asthma who present to pediatric ambulatory care settings. According to Knapp et al. \[[@B15]\], visits to an emergency department for moderate to severe asthma result in 29s% of the patients receiving an antibiotic prescription. Coprescribing antibiotics and asthma drugs might be due to asthma severity. However, some studies have shown that, compared to visits to pediatricians, visits to emergency departments by children with respiratory tract infections are less likely to result in a prescription for broad-spectrum antibiotics \[[@B16]\]. Marra et al. \[[@B6]\] also demonstrated that, for pediatric asthmatic patients, rates of prescribing antibiotics were higher in visits to physicians than to specialists. The latter resulted in increased diagnostic certainty of asthma exacerbation and less antibiotic usage. Likewise, our patients were followed prospectively by specialists, which resulted in a decreased use of antibiotics. Patient asthma education is increasingly being viewed as an important aspect of the ambulatory care setting \[[@B17]\]. Paul et al. \[[@B1]\] demonstrated that asthma education during patient visits is associated with a reduced number of antibiotic prescriptions. The results from this finding suggest other potential benefits for asthma education, as it seems to be associated with more judicious prescribing of antibiotics by providers. Parent-doctor communication during visits is associated with the prescription of antibiotics with the asthma medication. Some parents who believe their children have a disease which should or could be treated with antibiotics may directly request antibiotics \[[@B18]\]. If the patients and their parents are informed and educated about asthma during the visit to the specialist, they are less likely to request antibiotics. This, of course, leads to a general decrease in antibiotic usage \[[@B1]\]. Our patients and their parents were informed and educated at every visit. The limitation of this study was that for the year prior to the diagnosis of asthma, the patients were evaluated retrospectively. The number of episodes and the frequency of antibiotic usage were obtained from the database and parent reports. The use of a standard classification system for antibiotics may be more reliable. However, in the year after the diagnosis of asthma, patients were regularly followed up prospectively by the specialists and no patients were excluded from the study. Conclusion {#sec1_5} ========== This study showed that the appropriate diagnosis and treatment of childhood asthma significantly reduce the frequency of antibiotic usage and the number of asthma episodes. Diagnostic certainty about asthma exacerbations, appropriate therapies and regular follow-up contributed to these results as well as follow-up visits to a specialist. Disclosure Statement {#sec1_6} ==================== The authors have no conflicts of interest or funding to disclose. The authors appreciate the contributions and editorial assistance made by S. Delacroix, a native English speaker. {#F1} {#F2} ###### Characteristics of the patients ----------------------------------------- ---------------- Gender  Male 211 (63)  Female 123 (37) Median age (range), months 84 (24 -- 192) Mean age at diagnosis ± SD, months 61.40 ± 33.29 Attendance at school or day care center 212 (63) Household size 4 ± 1 At least one smoker in the household 137 (41) Family history of doctor-diagnosed asthma and atopy 200 (60) Inhaled-allergen positivity 207 (62) High immunoglobulin E levels 187 (56) Eosinophilia 160 (48) Food allergies 40 (12) Allergic rhinitis symptoms 160 (48) Atopic dermatitis 77 (23) ----------------------------------------- ---------------- Unless otherwise indicated values represent n (s%). ###### Associations between characteristics of patients and change in outcomes before and after the diagnosis Factors Asthma episode Antibiotic use -------------------------- ---------------- ---------------- -------- ------- Gender 0.021 0.726 0.061 0.321 Diagnosis age −0.119 0.051 −0.290 0.772 Household number 0.045 0.466 0.035 0.582 Mother\'s education −0.038 0.551 −0.031 0.626 Father\'s education −0.039 0.538 −0.079 0.214 Smoking in the household 0.064 0.296 0.007 0.915 History of atopy 0.042 0.489 −0.035 0.559

Introduction ============ Bipolar disorders are chronic conditions with lifetime prevalence rates of 0.6% for bipolar I and 0.4% for bipolar II.[@b1-ndt-13-733] Although bipolar disorders have a strong genetic etiology,[@b2-ndt-13-733] mood symptoms are significantly related to psychosocial events.[@b3-ndt-13-733] Thus, the personalized identification of individual psychosocial triggers is important to manage bipolar disorders. Studies of bipolar symptomatology often rely on self-report measures and fixed-interval weekly to monthly assessments,[@b4-ndt-13-733] but this approach is limited by recall bias. Several studies suggest that retrospective self-ratings are strongly affected by peak moments.[@b5-ndt-13-733] Because of these limitations, there is growing interest in the self-monitoring of mood and behavior on a daily basis. This monitoring includes observations of progression into episodic depressive or manic states.[@b6-ndt-13-733]--[@b8-ndt-13-733] Ecological momentary assessments involving repeated sampling of subjects' current behaviors and experiences in real time and in subjects' natural environments have been used in several studies.[@b7-ndt-13-733]--[@b9-ndt-13-733] This aims to minimize recall bias and maximize ecological validity. The ecological momentary assessments have been used in paper-and-pencil form to explore emotional reactivity to daily life stress among participants with affective disorder including bipolar disorders.[@b8-ndt-13-733] Frequent self-monitoring of symptomology associated with bipolar disorders including prodromal symptoms would allow psychiatrists to personalize treatment approaches in a more timely manner.[@b8-ndt-13-733] However, because the observation periods were less than 6 months in most studies,[@b8-ndt-13-733] there are few data about long-term recurrence prevention effects using ecological momentary assessments. Therefore, we reported three cases using self-monitoring daily mood charts to prevent recurrence episodes in higher risk individuals with bipolar disorders. We conducted the present study after obtaining approval from the Ethics Committee at the Hirosaki University School of Medicine. Participants provided written informed consent after receiving a full description of the study and consented for the report to be published. Cases ===== Case 1 ------ The patient was a 43-year-old Japanese male with a 6-year history of type I bipolar disorder with four admissions for a mixed state. He was treated with 100 mg/d of amoxapine, 30 mg/d of mirtazapine, and 5 mg/d of nitrazepam for more than 6 months. He was transferred to our hospital because of the poor control of his mental state. His medication was changed to 400 mg/d of lithium and 75 mg/d of quetiapine. He recorded his moods using a 5-point Likert scale ranging from −2, not at all happy, to 2, extremely happy, and paper-and-pencil mood charts including mood, motivation, thinking speed, and impulsivity. He recorded waking time and bedtime, and medication check, in addition to the four moods ([Figure 1](#f1-ndt-13-733){ref-type="fig"}). The patient submitted self-monitored mood charts every 2 weeks at each visit. The patient and psychiatrist discussed associations between mood and daily event or sleep--wake rhythms and looked for any patterns. In addition, the psychiatrist reported if he noticed any subclinical and prodromal depressive or manic moods and prescribed an order-made coping style for them. We made three copies of the self-monitoring mood charts and delivered them to the patient, his wife, and a manager at his job at the end of each visit. All of the original mood charts were kept with his medical chart. He made daily routines and decided on a final goal. Following monitoring of his ecological statements and case management, his psychiatric symptoms improved, and 1 year later, he returned to work and continued working at his job. He has continued the self-monitoring daily mood charts for more than 7 years, and his mental state has remained stable ([Table 1](#t1-ndt-13-733){ref-type="table"}). He visits us at a 6-week interval. Case 2 ------ The patient was a 38-year-old Japanese female with a 5-year history of type I bipolar disorder with admission of three times because of a manic state. Because the patient had exhibited aggressive behavior toward her parents, she was treated with 800 mg/d of lithium, 800 mg/d of valproate, 9 mg/d of aripiprazole, and 4 mg/d of flunitrazepam. In addition to self-monitoring daily mood charts, as described in Case 1, medication dosing was checked, and wake-up time, bedtime, medication time, and body weight were recorded. Her mood fluctuations decreased, and her mental state has remained relatively stable for more than 5 years. For work, she helped her parents at an apple farm. She has continued the self-monitoring daily mood charts for more than 6 years, and her mental state has remained stable ([Table 1](#t1-ndt-13-733){ref-type="table"}). She visits us at a 2-week interval. Case 3 ------ The patient was a 54-year-old Japanese female with a 3-year history of type II bipolar disorder with two admissions for a depressive state. She was treated with 800 mg/d of lithium and 0.25 mg/d of brotizolam. Because her mood swings were becoming too rapid, self-monitoring daily mood charts as described in Case 1 was initiated. In addition to self-monitoring daily mood charts, wake-up time, bedtime, and medication checks were recorded. Her mood fluctuations decreased, and her mental state has remained stable for more than 5 years. She got married 2 years after initiating self-monitoring daily mood charts, and she does housework. She has continued the self-monitoring daily mood charts for more than 5 years, and her mental state has remained stable ([Table 1](#t1-ndt-13-733){ref-type="table"}). She visits us at a 4-week interval. Discussion ========== The results of cases suggested that our paper-and-pencil form of self-monitoring daily moods was effective for preventing recurrence for bipolar disorders. We have maintained a stable mental condition for over 5 years using self-monitoring of daily mood in three cases, all of whom had a high risk of recurrence, although this number of subjects was very small and observation period was still short. We calculated the mean and coefficient of variation using 365 points ([Table 1](#t1-ndt-13-733){ref-type="table"}). Mood fluctuations decreased in all patients as the average score approached zero and coefficient of variation values decreased over time. As long-term treatment should be required for patients with bipolar disorder, our method of self-monitoring daily mood may be superior to current options in terms of utility and feasibility. Further studies are required to confirm the utility of our self-monitoring daily mood approach in larger samples. Many studies have suggested that the main factors of recurrence in bipolar disorders are noncompliance and abnormal sleep--wake rhythms.[@b10-ndt-13-733]--[@b12-ndt-13-733] Therefore, we added columns for wake-up time and bedtime and a medication check to the mood chart. In addition, we divided mood into four components, namely, mood, motivation, concentration or thinking speed, and impulsivity or aggression, because we aimed to capture the mixed futures of bipolar disorders based on mixed depression, as described by Kraepelin.[@b13-ndt-13-733] Most patients with successful outcomes wrote detailed routine events as well as special events in the remarks section of the self-monitoring daily mood chart. This may be a positive occurrence, as it suggests that they looked back at themselves and their behaviors for that day. Weekly summaries of patients' data can be discussed with them so they gain insight into the contexts in which they feel better or worse and can adjust their behavior accordingly. **Disclosure** Norio Yasui-Furukori has received grant/research support or honoraria from and has been a lecturer for Asteras, Dainippon, Eli Lilly, GSK, Janssen-Pharma, Meiji, Mochida, MSD, Otsuka, Pfizer, Takeda, and Yoshitomi; none of the funders had a role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The authors report no other competing interests or conflicts of interest in this work. {#f1-ndt-13-733} ###### Characteristics of cases and their outcomes Characteristics Case 1 Case 2 Case 3 ---------------------------------------------------------- -------- ---- -------- ---- -------- ---- Age (years) 43 38 54 Sex Male Female Female Diagnosis BP I BP I BP II Entry status  MADRS 4 9 9  YMRS 13 15 4 Current status  MADRS 5 8 4  YMRS 2 11 5 Number of episodes 1 2 2 Number of admissions 0 0 0 Mood[a](#tfn1-ndt-13-733){ref-type="table-fn"}  1st year −0.50 45 −0.76 48 −0.51 40  2nd year −0.57 43 −0.64 47 −0.31 29  3rd year −0.36 32 −0.35 33 −0.19 21  4th year −0.28 29 −0.36 34 −0.06 21  5th year −0.23 28 −0.44 37 −0.07 23 Motivation[a](#tfn1-ndt-13-733){ref-type="table-fn"}  1st year −0.69 47 −0.74 48 −0.53 39  2nd year −0.68 44 −0.69 44 −0.42 35  3rd year −0.32 31 −0.37 34 −0.12 26  4th year −0.30 32 −0.39 36 −0.17 26  5th year −0.23 25 −0.41 33 −0.11 22 Thinking speed[a](#tfn1-ndt-13-733){ref-type="table-fn"}  1st year −0.91 57 −0.67 47 −0.58 44  2nd year −0.73 51 −0.58 43 −0.38 39  3rd year −0.25 35 −0.39 40 −0.16 31  4th year −0.17 33 −0.30 38 −0.21 31  5th year −0.19 30 −0.32 37 −0.21 29 Impulsivity[a](#tfn1-ndt-13-733){ref-type="table-fn"}  1st year −1.32 82 −1.81 62 −1.81 51  2nd year −1.48 71 −1.48 70 −1.84 46  3rd year −1.92 40 −1.92 40 −1.92 40  4th year −1.83 50 −1.83 50 −1.69 51  5th year −1.93 34 −1.93 34 −1.80 41 **Notes:** Data are shown as mean and CV using 365 points. CV was calculated by standard deviation divided by mean after transformation from −2 to 2 Likert scale to 1 to 5 Likert scale. **Abbreviations:** BP, bipolar disorders; CV, coefficient of variation; MADRS, Montgomery and Asberg Depression Rating Scale; YMRS, Young Mania Rating Scale.

###### Strengths and limitations of this study - Strengths include a longitudinal design, with first assessment within 3 weeks after arrival to the host country, and repeated measures. - Use of computer-based assessment with audio-translations throughout the study. - Selection of participants was limited to the most common nationality groups arriving in Norway at the time of inclusion. - High attrition rate due to the fact that asylum seekers tend to move between and within countries and that many were told to leave the country. Introduction {#s1} ============ In 2015, more than 88 700 unaccompanied minors (UMs) fled to Europe,[@R1] putting considerable pressure on these countries to provide the necessary resources needed. Separated children who are no longer protected by parents or other caregivers, usually have to be under the age of 18 in order to be given the special protection and care that is granted unaccompanied refugee minors. In the countries of origin for UM, the civil registration service of their country often function poorly, and birth certificates can be lost, thrown away or falsified.[@R2] The scientific basis for assessing age is controversial, in that these tests only determine physical maturity and are most uncertain from the age of 15 to 21 years, where natural variation is at its greatest.[@R3] The consequences for many young asylum seekers assessed to be 18 years or older is that they will no longer be considered as minors and therefore not receive special protection in accordance with the United Nations.[@R2] Most studies investigating UM mental health have a cross-sectional design with a selection of youths with different levels of legal recognition and different durations of time in exile.[@R4] These studies show consistently that individual factors such as exposure to violence and other traumatic events prior to migration, correspond to elevated symptoms of psychological distress.[@R5] In some studies, the negative effects of exile-related stressors are also described,[@R6] yet they focus on youths with varying time in exile. There are different asylum procedures within the different countries,[@R7] and most UM endure some uncertainty before their legal status is defined. Most countries provide some form of shelter for UM while they are waiting for their case to be processed, but conditions vary greatly. Positive health effects have been shown to be associated with receiving a permanent residence permit,[@R8] but this process may take months and sometimes years. The impact of different levels of social support that UM are offered, especially after the first stage of reception and registration, has not been studied in detail.[@R9] The aim of our study was to examine UM's mental health during the asylum-seeking process and more specifically whether the official age assessed, level of support and the outcome of the asylum application were associated with UM's mental health at different stages of the asylum-seeking process. Methods {#s2} ======= Participants and procedures {#s2a} --------------------------- The sample in this study was recruited from an asylum reception centre for unaccompanied asylum-seeking adolescents between ages 15 and 18 years, which was the only one in Norway at this time. In this reception centre, all UMs who claimed to be in this age group stayed for the first weeks while asylum interviews and age-assessment procedures were performed. A research assistant kept track of all new arrivals, and each time our testing capacity allowed us to include some new participants, she was instructed to invite the ones who had arrived most recently. The study was conducted between September 2009 and March 2011. Altogether, the inclusion periods for this project were 12 weeks in 2009, 8 weeks in 2010 and 21 weeks in 2011. During these time periods, young asylum seekers came mainly from Afghanistan and Somalia. According to the statistics unit at the Norwegian Directorate of Immigration, 406 male UM from these language groups arrived in Norway during the inclusion periods. Unaccompanied males who had just arrived were contacted by the research assistant. Altogether, 216 adolescents were asked to participate, and 209 returned the informed consent and attended the study. Some participants were included in an Expressive Arts intervention group (n=71) that is not part of the present study. The remaining 138 are the focus of this article. Inclusion in the intervention group was based on a randomising procedure shortly after arrival in Norway. The participants in the present article were not significantly different from the intervention group in any baseline characteristics (p≥0.071). More about the whole project can be found on our home pages.[@R10] Information to participants included statements that participation would not impact the chances to stay in the country. Only one contact attempt was made for each individual, and no payment was offered. Participants followed the normal procedures in the asylum process. In Norway, all UMs receive assistance from a multidisciplinary professional staff (educators, social workers, psychologists, physicians and nurses) in the first reception centre while waiting for their 'official-age' to be assigned. Those defined as 18 or older can be moved to adult housing where less professional assistance is provided. The asylum seekers considered to be from 15 to 18 years are moved to specialised youth centres, with staff available 24 hours, every day. The youngest children stay in even more specialised orphanages. There are some exceptions to this pattern; according to variable housing capacity, some 18-year-old asylum seekers are allowed to stay in the youth centres for some time. The youth centres are located all over Norway and have language classes for all inhabitants. Food is prepared and served by the staff, and there are staff members available day and night. Most centres have recreational activities, and they give individualised support and medical follow-up if needed. In an adult centre, the asylum seekers are left to themselves most of the time. They buy and cook their own food, have no school or other scheduled activities and have no guardians or staff members to ask for advice. The first screening procedure was conducted within the first 3 weeks and later repeated at 4 months (n=101), 15 months (n=84) and 26 months (n=69) after arrival. At the last assessment, the population was almost halved, mainly because many of the informants were transported out of the country or had disappeared from the different living facilities. The participants who were deported were mostly individuals who had been registered as asylum seekers in another European country before coming to Norway or individuals suspected of having some connection to illegal activities. The ones who deflected were typically those who feared deportation after their asylum applications were turned down. It was, however, impossible to obtain exact numbers and reasons for the attrition in this project. When we compared those who have completed all four assessments with those who missed out at one occasion or more, there were no significant differences in any baseline demographic or symptom variables. Measures {#s2b} -------- Demographic data were registered with the aid of interpreters at the initial assessment. We asked for self-reported age, literacy, years of school attendance and whether their parents were still alive, deceased or if participants had lost touch with their parents and did not know. Later, we registered the results of official age assessments, especially which participants who were thought to be at least 18 years of age. We also determined the level of care offered according to placement in asylum centres for either adults or for youth. Before the last assessment, we registered the legal status, as participants were either given time-limited or permanent permission to stay or were refused legal residence in the country. New variables of interest such as level of care and legal status were included when they occurred prior to a new assessment. ### Exposure {#s2b1} Serious Life Events checklist (SLE) was developed by Tammy Bean and colleagues[@R11] in order to assess if an adolescent meet the criteria A1 (experienced a traumatic event) in the Diagnostic and Statistical Manual of Mental Disorders, Fourth edition (DSM-IV), for a diagnosis of post-traumatic stress disorder (PTSD). It is a self-report questionnaire which asks whether or not the participant has experienced 12 different kinds of traumatic events, such as separation from family, natural disaster, war and physical or sexual abuse. The instrument was scored by answering yes or no on each item. ### Psychological distress {#s2b2} Hopkins Symptom Checklist-25 (HSCL-25)[@R12] is a self-administered questionnaire designed to measure anxiety and depression. It has been validated in various clinical and community samples.[@R13] The HSCL-37 A version is an extension of the HSCL-25 and has also been applied in a number of refugee studies with minors.[@R15] The additional 12 items measuring externalising behaviour are not included in this paper. Each item was scored with 1 (not bothered) to 4 (extremely bothered). Scores≥2 was considered probably clinically significant.[@R17] ### Post-Traumatic Symptom Score (PTSS) {#s2b3} The Harvard Trauma Questionnaire[@R18] (HTQ) is a comprehensive instrument that was developed to assess potentially traumatic experiences and post-traumatic symptoms in various cultural contexts. Its psychometric properties were first established in a highly traumatised, clinical population, but it has also been evaluated with a larger community sample and with asylum-seeking adolescents.[@R6] The HTQ part IV comprises 30 symptom items, among which the first 16 items measure 'The symptoms of PTSD' according to the DSM-IV.[@R20] These 16 items are scored with 1 (not at all) to 4 (extremely). Scores≥2 was considered probably clinically significant.[@R17] ### Computer-based assessment {#s2b4} The chosen psychometric measurements were combined into a single questionnaire using the program MultiCASI.[@R21] The questionnaires were filled in by the participants themselves, in their native languages, Dari, Pashto, Farsi or Somali, using laptops with touch-screen function. Translations had been attained from earlier projects and were controlled by independent, native speaking, interpreters before they were added to the questionnaire. The items appeared one after the other on the screen, together with answering alternatives. All text had a sound-file connected to it that started as soon as the item appeared on the screen. The test could be used with any level of reading competence, and the sound of each item could be activated by touch as many times as necessary. Items could be skipped and left unanswered, but would then be repeated once more towards the end of the questionnaire. The first introduction to the computer-based self-screening was done shortly after arrival, with one language group at the time. An interpreter was present together with maximum five participants, as they were instructed in how to use the touch screen. They were encouraged to ask clarifying questions as they went on with answering the items, all in the same room, with earphones on, in order not to disturb each other. During the following waves of data collection, the same questionnaire was used and translating services were not necessary. The results were transported digitally to the SPSS files. Data analysis {#s2c} ------------- Differences in HSCL and PTSS between 0, 4, 15 and 26 months were assessed by linear mixed effects models by categorical time, including an interindividual random effect. Relationships between HSCL and PTSS at each time point ≥4 months and characteristics known at that time point were assessed by unadjusted and linear regression. Specifically, covariates were being literate, parents deceased, number of adverse events and age assessed as ≥18 years at 4 months. At 15 months, being placed in a reception centre for adults or youth was included, and at 26 months also asylum status: permanent, time-limited or refusal of asylum. Due to a low number of missing values in the independent variables in the regression analyses (at most three missing values on any independent variable), complete case analysis was considered appropriate. Non-response analysis during follow-up (4--26 months) used a generalised estimating equations (GEE) logistic regression by time and baseline HSCL score, reading ability, category for parents alive and number of serious life events. For descriptive analyses, we used the SPSS Version 22 for Windows. Beyond this, data were analysed using R (The R Foundation for Statistical Computing, Vienna, Austria) with the R package nlme for mixed effects models and gee for GEE analyses.[@R22] Results {#s2d} ------- Three-fourths of the population came from Afghanistan, while the remaining came from Somalia and Iran ([table 1](#T1){ref-type="table"}). There were no significant differences between the countries of origin and the variables included in this article. A minority (36%) were able to read in their own language. Mean number of serious lifetime events experienced was 6.1 (SD 2.3), range 1--11. Most of the participants (96%) had experienced at least one of the serious life events listed. The most frequently reported experiences were life-threatening events (82%), physical abuse (78%) and loss of a close relative (78%). The official age assessment found a mean age of 18.4 years (SD 2.4), range 15--28, which meant that 72 (56%) participants were considered to be adults. Of this 'adult' group, 36 participants were allowed to stay at the care centres for adolescents, while the rest had to move to centres for adults. None of the participants received psychiatric treatment during the study. Overall, there were no significant changes in the level of symptoms within the study period (p≥0.084), neither for HSCL ([table 2](#T2){ref-type="table"}) nor for PTSS. ###### Baseline characteristics of male unaccompanied refugee minors at arrival in Norway N=138 -------------------------------------- --------------- Age, self-reported (n=130)  Mean years (SD)  16.22 (0.84)  Range  14--20 Age, assessed by authorities (n=132)  Mean years (SD)  18.22 (2.27)  Range  15--27 Nationality  Afghan 102 (73.9)  Somalian  32 (23.2)  Iranian   3 (2.2)  Algerian   1 (0.7) Literacy, self-reported (n=136)  50 (36.8)  No loss of parent  30 (21.7)  Loss of father  60 (43.5)  Loss of mother   4 (2.9)  Loss of both parents  25 (18.5)  Unknown  16 (11.9) Psychological distress (n=131)  Mean HSCL (SD)   1.94 (0.58)  Caseness (n≥2.0)  29 (21.0) Post-traumatic stress (n=133)  Mean PTSS (SD)   2.16 (0.62)  Caseness (n≥2.0)  81 (58.7) Values are given as number (%) when others not specified. HSCL, Hopkins Symptom Checklist; PTSS, Post-Traumatic Stress Symptom Checklist. ###### MEC for time modelling the course of psychological distress (HSCL) and post-traumatic stress (PTSS) in unaccompanied refugee minors after arrival in host country HSCL PTSS --------------------------- ------- --------------- ------- ------- --------------- ------- Time 0.136 0.725 4 months versus 0 months 0.04 −0.09 to 0.16 0.557 0.02 −0.12 to 0.15 0.811 15 months versus 0 months 0.14 0.01 to 0.27 0.037 0.03 −0.11 to 0.17 0.671 26 months versus 0 months −0.02 −0.16 to 0.13 0.831 −0.06 −0.21 to 0.09 0.441 HSCL, Hopkins Symptom Checklist; MEC, mixed effect coefficients; PTSS, Post-Traumatic Stress Symptom Checklist. [Tables 3-5](#T3 T4 T5){ref-type="table"} show the associations between variables of interest and symptoms of psychological distress at different test points. Outcome of age assessment, which was known shortly after the first assessment, had no significant association with psychological distress at 4 months ([table 3](#T3){ref-type="table"}). However, those who were estimated to be 18 years or older, had higher levels of symptoms at 15 months ([table 4](#T4){ref-type="table"}) and at 26 months ([table 5](#T5){ref-type="table"}), but not when adjusted for the outcome of the asylum applications at the 26-month assessment. ###### Regression coefficients for literacy, premigration bereavement, serious life events and postmigration age assessment, related to course of psychological distress (Hopkins Symptom Checklist) in young male asylum seekers 4 months after arrival in host country; results unadjusted and adjusted for the other variables Unadjusted Adjusted ------------------------------ ------------ ----------------- ------- -------- ----------------- ------- Being literate 0.348 0.115 to 0.581 0.004 0.262 0.006 to 0.518 0.045 Parents deceased 0.245 0.457  Unknown versus both alive 0.175 −0.232 to 0.581 0.396 0.146 −0.254 to 0.545 0.472  One dead versus both alive 0.146 −0.166 to 0.457 0.355 0.182 −0.119 to 0.483 0.234  Both dead versus both alive −0.172 −0.564 to 0.219 0.384 −0.053 −0.442 to 0.337 Adverse events 0.066 0.015 to 0.116 0.012 0.046 −0.006 to 0.098 0.084 Age assessed≥18 years 0.126 −0.118 to 0.370 0.308 0.068 −0.191 to 0.326 0.604 ###### Regression coefficients for literacy, premigration bereavement, serious life events and postmigration age assessment, in addition to asylum seeker facilities, related to course of psychological distress (Hopkins Symptom Checklist) in young male asylum seekers 15 months after arrival in host country; results unadjusted and adjusted for the other variables Unadjusted Adjusted ------------------------------ ------------ ----------------- --------- ------- ----------------- ------- Being literate 0.054 −0.254 to 0.363 0.727 0.008 −0.296 to 0.313 0.957 Parents deceased 0.134 0.073  Unknown versus both alive 0.240 −0.278 to 0.757 0.359 0.346 −0.133 to 0.825 0.154  One dead versus both alive 0.253 −0.141 to 0.646 0.206 0.317 −0.051 to 0.684 0.090  Both dead versus both alive 0.581 0.097 to 1.065 0.019 0.626 0.157 to 1.094 0.010 Adverse events 0.039 −0.030 to 0.107 0.262 0.054 −0.010 to 0.119 0.099 Age assessed≥18 years 0.522 0.238 to 0.805 \<0.001 0.375 0.058 to 0.692 0.021 Adult reception centre 0.464 0.136 to 0.792 0.006 0.354 0.011 to 0.695 0.043 ###### Regression coefficients for literacy, premigration bereavement, serious life events and postmigration age assessment, asylum seeker facilities, in addition to asylum status, related to course of psychological distress (Hopkins Symptom Checklist) in young male asylum seekers 26 months after arrival in host country; results unadjusted and adjusted for the other variables Unadjusted Adjusted ------------------------------- ------------ ----------------- --------- -------- ----------------- ------- Being literate 0.025 −0.305 to 0.355 0.881 −0.040 −0.322 to 0.242 0.777 Parents deceased 0.043 0.038  Unknown versus both alive 0.591 0.021 to 1.162 0.042 0.562 0.076 to 1.047 0.024  One dead versus both alive 0.261 −0.130 to 0.652 0.187 0.384 0.049 to 0.719 0.025  Both dead versus both alive 0.670 0.160 to 1.180 0.011 0.532 0.088 to 0.976 0.020 Adverse events −0.059 −0.126 to 0.008 0.083 −0.041 −0.097 to 0.016 0.155 Age assessed≥18 years 0.392 0.086 to 0.697 0.013 −0.070 −0.428 to 0.288 0.696 Adult reception centre 0.717 0.372 to 1.063 \<0.001 0.272 −0.169 to 0.712 0.222 Asylum status (vs acceptance) \<0.001 0.017  Time-limited asylum −0.035 −0.391 to 0.320 0.844 −0.103 −0.498 to 0.292 0.602  Refusal of asylum 0.787 0.402 to 1.172 \<0.001 0.590 0.122 to 1.059 0.015 One-third of the participants were placed in a reception centre for adults. [Figure 1](#F1){ref-type="fig"} shows the trajectories of psychological distress for participants placed in a reception centre for adults or for youth. Those who were placed in a reception centre for adults had higher levels of psychological distress symptoms both at 15 months ([table 4](#T4){ref-type="table"}) and 26 months ([table 5](#T5){ref-type="table"}) compared with the remaining participants who were placed in reception centres for youth. However, when adjusted for the outcome of the asylum application at the 26-month assessment, the difference was not significant. {#F1} Final decision on the asylum claims was given between the last two test points. Refusal was highly associated with higher levels of psychological distress. Achieving time-limited residence permission was not significantly different compared with permanent asylum ([table 5](#T5){ref-type="table"}). Trajectories of psychological distress for those who received refusal or acceptance of their asylum application are illustrated graphically in [figure 2](#F2){ref-type="fig"}. Refusal was related to the official determined age of the asylum seeker. Among the participants who were considered to be 18 or more, 52 out of 72 (72.2%) were refused, compared with 15 out of 59 (25.4%) among the participants who were considered to be under 18 (seven missing). {#F2} The symptom scores of the PTSS (not illustrated in the tables) showed a similar association as the HSCL scores, with higher levels of PTSD symptoms for those placed in a reception centre for adults at 15 months (adjusted difference 0.34, 95% CI 0.06 to 0.63, p=0.017) as well as higher symptom scores for those who received a negative result for the asylum application at 26 months (adjusted difference 0.60, 95% CI 0.24 to 0.95, p=0.001). Loss to follow-up was not significantly related to initial levels of distress. Also, none of the baseline covariates were significantly related to non-response. Discussion {#s3} ========== The present study is a follow-up of unaccompanied refugee minors with four waves of assessment from within 3 weeks after arrival to more than 2 years spent in the host country. At the group level, the young asylum seekers reported high levels of psychological distress on arrival and symptom levels that stayed relatively unchanged over time. A low level of support during the asylum process and a negative outcome of the asylum application were associated with higher levels of psychological distress. Determination of the legal status of the asylum seekers involved age assessment procedures, with X-rays and dental examinations for all participants in this study. This resulted in a considerable gap between self-reported age and the official age estimates designated by the immigration authorities. On the basis of these examinations, 55% of the asylum seekers were considered to be at least the age of 18 and thus did not achieve a UM status. They risked being moved to a facility for adults, with low levels of support and care, and limited access to education and leisure activities. Also, the likelihood of being granted asylum was related to age, as illustrated by the numbers of children and adults in our study who got refusal of their claims. The results from our study are in agreement with other studies that have found that high-support housing, with sufficient supervision, was associated with lower levels of psychological symptoms.[@R5] Others have also described problems directly connected to the asylum process and have registered them as components in a list of postmigration stressors.[@R9] A weakness with most of these studies, are cross-sectional designs where there are no baseline measurements. Only a few studies have repeated assessments[@R6] where problems directly connected to the asylum process, such as age-assessment procedures, lack of adequate housing, low support, and so on, have been evaluated. The complexity of factors contributing to the increasing health risk make it difficult to draw specific conclusions within the total burden of stressors. In all studies with UM, it is likely that there will be some uncertainty concerning the participants' true chronological age.[@R3] Defined to be overage, in the present study, was not significantly related to the symptom scores at the 4-month assessment, and there was no indication that this process was stressful in itself. The age designated by the authorities, determined what type of housing and level of care that was offered during the remaining asylum procedure. This meant that many of the participants had to live in a reception centre for adults, where they had no guardian, no school, had to cook for themselves and budget their benefits. Our findings that this group had higher levels of psychological distress add further evidence that living conditions in the asylum seeking period may influence the mental health of young refugees.[@R6] It was probably known in the community and among the youth that being categorised as an adult increased the risk of asylum refusal. This factor is impossible to separate from the expectations associated with the placement in youth or adult reception centres. There should be a cautious interpretation of the results because of this clustering of risk factors. It is also possible that the asylum interviews were more adversarial for those who had adverse age assessments. These interviews happened early in the asylum trajectories, but these official age assessments may have been used to question testimonial credibility in the asylum process. The outcome of the individual asylum applications was revealed to the asylum seekers between 1 and 2 years after the arrival, and the negative impact of refusal was as expected since several studies have found that difficulties obtaining legal residence are associated with a range of psychological problems for this group.[@R6] We also know that longitudinal studies indicate a trend towards reduction of mental health symptoms for resettled refugees over time.[@R23] In a follow-up study of 131 young refugees in Denmark, the long-term effects of premigration adversity were mediated by a variety of factors connected to social life.[@R24] Another study suggests positive health effects on receiving permanent residence mediated through improved living conditions.[@R25] This, in association with our findings, emphasises the importance of a supportive postmigration environment for all refugees with premigratory experiences of serious trauma and human rights violations. Strengths of our study include a longitudinal design, with first assessment within 3 weeks after arrival to the host country and repeated measures. We used computer-based assessment with the same audio-translations throughout the study and did not need to use interpreters in order to complete the psychometric measures at follow-up. Due to a random selection of participants, we consider the sample to be representative for the refugees arriving to Norway in the beginning of the century. However, selection of participants was limited to the most common nationality groups arriving in Norway in this period and may limit the generalisation of our findings to refugees in general. High attrition rate due to the fact that asylum seekers tend to move between and within countries, and that many were told to leave the country, may have biased our findings. It is also possible that our research team was not viewed as independent from the authorities, even though we stressed this fact when we informed about the project. Finally, we have no data as to whether poor mental health might have affected the likelihood of asylum. Mental health is generally not an issue in the processing of asylum applications in Norway. Also, the baseline levels of mental health did not differ between participants that later received asylum and those who did not. Implications {#s3a} ------------ Our study shows that young asylum seekers may spend considerable time in a safe Western country, without recovering from the distress they have when they arrive in the host country. A reason for the continuing psychological health problems in this non-clinical group of youth can possibly be found in the living conditions and the level of care that is provided. Adolescence is a challenging transition period for most people. Fleeing to a foreign country without parents or other caregivers makes this life period even more challenging for young refugees and puts a considerable responsibility on the receiving countries. The burden of increasing numbers of asylum seekers challenges the political intentions of the UN Convention on the Rights of the Child to always give precedence to 'the best interest of the child'.[@R26] It is emphasised that safety and dignity in the use of medical assessments should be applied as a supplement to evaluations of the physical appearance and the psychological maturity of the child. An important objection to the use of dental/bone-age assessments is their lack of precision, especially around the time of puberty. The tests have been criticised for their large margins of error and their inadequacy in determining chronological age.[@R3] Professionals in various countries have differed with some doctors refusing to take part in such tests, while others have argued that these assessments are the best practice available. Needs of vulnerable adolescents and young adults in a stressful life situation deserve high priority and should be a main focus regardless of the outcome of age assessments.[@R27] It is noteworthy that access to psychiatric care was not evident for any of the participants although a majority of this sample had symptom levels suggesting a positive diagnosis of PTSD. This may reflect a lack of resources available for this population or reluctance to ask for healthcare. In our society, turning 18 is usually considered a transition point from child to adult. Yet with the limitations of the age-determining process, we cannot know for certain that this milestone has been reached. The consequences of this uncertainty can have legal, social and material implications.[@R28] If a child is put under difficult living conditions, where previous human support and education are withdrawn, this can have unintended negative effects on these young individuals transitioning into adulthood. Some child protection services argue that vulnerable young adults are still in need of support and care after the age of 18[@R29] and need to receive specialised care into their 20s.[@R30] Future studies should focus on how mental health and resilience evolve over a longer time span and evaluate specific interventions and appropriate levels of care for young refugees. Supplementary Material ====================== ###### Reviewer comments ###### Author\'s manuscript Gratitude to Liv Berit Løken for her care and assistance with all aspects of data collection. Thanks also to our very skilled interpreters and to all the young participants. **Contributors:** MJ has had the main responsibility for the drafting and writing of the article. TH was, in collaboration with MAMD and MJ, responsible for the literature review and the conception and design of the article. MAMD and MJ have been responsible for all phases of the data collection. Data analysis and interpretation of data were done in cooperation among MJ, TH and TW-L. All authors have contributed to the scientific writing and proof-reading of the article. The paper has been read and approved by all authors before submission. **Funding:** This work was supported by the Norwegian Directorate of Immigration. **Competing interests:** None declared. **Patient consent:** Obtained. **Ethics approval:** The Regional Medical Ethics Committee, South-East Norway approved this study. **Provenance and peer review:** Not commissioned; externally peer reviewed. **Data sharing statement:** No additional data are available. **Correction notice:** This paper has been amended since it was published Online First. Owing to a scripting error, some of the publisher names in the references were replaced with \'BMJ Publishing Group\'. This only affected the full text version, not the PDF. We have since corrected theseerrors and the correct publishers have been inserted into the references.

INTRODUCTION ============ Deficient androgen exposure during fetal masculinization results in smaller sized testes, prostate, seminal vesicles, and penis in adulthood and is associated with an increased risk of hypospadias and lower testosterone levels \[[@B1],[@B2],[@B3],[@B4]\]. It is well established that normal penile development is dependent on testosterone, its conversion via steroid 5-alpha-reductase to dihydrotestosterone, and a functional androgen receptor \[[@B5]\]. Penile length increases slowly until 4 years of age, after which follows a steady phase and then a rapid increase with puberty. Androgen deficiency is also known to be associated with a reduction in anogenital distance (AGD) in adults \[[@B6]\], and it is postulated that the ratio of the second to the fourth finger length (2:4 digit ratio) in adult men reflects fetal androgen exposure \[[@B7]\]. Studies on penile length in newborns have been very rarely conducted, especially in Korea. In 1987, a study on penile length and testicular size was conducted with 1,071 Korean children including 49 newborn infants \[[@B8]\]. In newborns, the mean and standard deviation (SD) of stretched penile length (SPL) was 3.3±0.5 cm. Early diagnosis of abnormalities in penile size is important both medically and psychologically \[[@B9]\]. The exact penile size is a most important factor in diagnosing penile problems such as micropenis, which is defined as an SPL\<2.5 SDs below the mean for age with normal function and structure \[[@B10]\]. Until now, no reports have addressed penile length according to birth weight and its relationship with birth weight and AGD or digit length in newborn infants. This study was preliminarily performed to update the normal SPL values that can be used for Korean newborns. We also intended to investigate the current status of penile length and the relationship of penile length with AGD or digit length according to birth weight in Korean newborn infants. MATERIALS AND METHODS ===================== 1. Patient characteristics -------------------------- This cross-sectional study was performed in the neonatal unit of our institution between May and January 2014. Of a total of 78 newborn male infants, 55 infants were prospectively enrolled in this study. Exclusion criteria were penile diseases, including hypospadias, concealed penis, cryptorchidism, and varicocele, and other growth problems such as chronic renal failure and endocrinologic disorders \[[@B11]\]. Therefore, most newborn infants who were considered to be normal and healthy except for their birth weights were included in the current study. The infants were divided into two groups. The normal weight group included newborns with a gestational age of 38 to 42 weeks and birth weight≥2.5 kg (NW, n=24). The low birth weight group included newborns with gestational age\<38 weeks and birth weight\<2.5 kg (LW, n=31). 2. SPL and variables -------------------- The penile length was measured as the SPL. The length was measured twice for each infant and the mean of the two measurements was recorded. The SPL was measured with a ruler by compressing the fat tissue with one end of the ruler through the pubic ramus; then the penis was fully stretched and the distance to the glans of the stretched penis was plotted \[[@B11]\]. None of the 55 infants had been circumcised. Foreskins of the uncircumcised infants were not involved in the measurement. Testicular size was measured by using Prader orchidometry (mL). Genital distances were measured with the male infant in a supine and frog-leg position \[[@B12]\]. AGD1 was the distance from the anterior aspect of the penis to the anal verge ([Fig. 1A](#F1){ref-type="fig"}). AGD2 was the distance from the posterior aspect of the penis to the anal verge. AGD3 was the distance from the posterior aspect of the scrotum to the anal verge as measured by use of a caliper \[[@B13]\]. Digit lengths were measured by using a ruler on the left hand. Digit lengths were measured twice for each infant and the mean of the two measurements was recorded. The measurement was taken from the basal crease to the tip on the ventral surface of the hand at a point midway across a line perpendicular to the base by using rulers ([Fig. 1B](#F1){ref-type="fig"}). For all infants, age, SPL, height, body weight, testicular size, digit lengths of the second and fourth fingers and the ratio between them, and AGDs and their ratios were measured. Penile size and other variables were also compared between the two groups. 3. Interobserver variability ---------------------------- SPL was measured in 55 infants by two observers (an urologist and a pediatric doctor) to estimate interobserver difference. Differences in the measurements of penile length were evaluated by using paired t-tests for the LW group and Wilcoxon\'s signed rank test for the NW group. 4. Statistical analysis and ethics statement -------------------------------------------- Data analysis was performed by using the software package SPSS ver. 17.0 (SPSS Inc., Chicago, IL, USA). Continuous variables were expressed as the mean±SD. Differences in SPL, height, body weight, and testicular size between the study in 1987 and the current study were evaluated by Student t-test. Variables in the NW and LW groups were compared by Mann-Whitney U test. A p-value of \<0.05 was accepted to be statistically significant. This study was approved by the Institutional Review Board of Ulsan University Hospital (IRB No. 2013028). Clinical data were prospectively collected and the medical records of all the participants were reviewed. RESULTS ======= 1. SPL compared with the previous study (1987) ---------------------------------------------- The SPL of the NW group (n=24, \>2.5 kg) was 3.3±0.2 cm. Compared with the previous study results reported in 1987, there was no significant change in the SPL (p=0.445). Among the anthropometric measurements of Korean children, there was a significant decrease in height (49.8±1.9 cm), whereas there were no significant changes in body weight (3.2±0.5 kg) or testicular size (1.1±0.4 mL) ([Table 1](#T1){ref-type="table"}). 2. Penile length, digit length, and AGD according to birth weight ----------------------------------------------------------------- The SPLs of the NW and LW groups were 3.3±0.2 cm and 2.9±0.4 cm, respectively (p\<0.001). All parameters including height, weight, penile length, testicular size, AGD1-3, and the length of the second and fourth fingers were significantly lower in the LW group than in the NW group. However, there were no significant differences in the AGD ratio or in the 2:4 digit length ratio between the two groups ([Table 2](#T2){ref-type="table"}). 3. Interobserver variability in penile length --------------------------------------------- In the NW group, there was no significant difference in SPL between two observers (3.3±0.2 cm and 3.2±0.3 cm, p=0.165); however, there was a significant difference between two observers in the LW group (2.9±0.4 cm and 2.7±0.4 cm, p=0.001). DISCUSSION ========== The penile length in children has increased significantly \[[@B11]\]; however, the penile length of newborn infants has not changed over the last quarter century. Biological and environmental changes and changes in feeding patterns during the rapid economic growth since 1987 may have affected the growth and development of the human body. The height and weight of Korean children have increased significantly compared with 1987 \[[@B11]\]. However, according to the current study results, there has been no significant change in the birth weight of newborn infants, whereas there has been a reduction in their heights compared with 1987. It may be that Korean mothers today intentionally keep fetal weight under control for a safe and problem-free delivery. The exact causes of the decreased height for infants are unknown and must be evaluated in the future. There is no apparent relationship between gestational age and penile length at 37 to 42 weeks \[[@B14]\]. Therefore, we compared the current data with similar data reported in 1987 for a gestational age of 38 to 42 weeks and birth weight≥ 2.5 kg. Although the two studies were conducted in geographically different places, one in Ulsan and the other in Seoul in Korea (1987), and penile length was measured by different individuals, both studies were conducted with Korean populations and penile length was measured by the SPL technique. In humans, AGD differs by sex; boys have a longer AGD than do girls \[[@B15]\]. Additionally, numerous studies have shown sex differences in the 2:4 digit length ratio, and males have a lower 2:4 digit length ratio than do females \[[@B16]\]. Longer male AGD may be determined by an androgen effect during the presumptive masculinization programming window before 11 to 13 weeks of gestation as in rodents \[[@B1]\]. No formal studies have yet reported AGD in patients with complete androgen insensitivity, which would provide definitive proof that fetal androgens determine the longer AGD in human males \[[@B13]\]. However, a study conducted with Caucasian infants reported a significant reduction in AGD in boys with hypospadias (42 control and 77 infants with hypospadias, p=0.002) \[[@B17]\]. In another study conducted with 116 adults, a significant positive correlation between AGD and testosterone levels was demonstrated \[[@B6]\]. Some problems with AGD measurement need to be resolved. AGD3 is the distance from the posterior aspect of the scrotum to the anal verge. AGD3 is also called the anoscrotal distance and appears to be the most reliable and repeatable measurement \[[@B15],[@B18]\]. In the current study, AGD1, 2, and 3 were measured as the reference to utilize basic epidemiologic data \[[@B13]\]. There is also some conflict as to which hand presents the most sexually different 2:4 digit length ratio. At birth, sex difference in the 2:4 digit length ratio is significant only for the left hand \[[@B19]\]. In the current study, we measured digit length for the second and fourth fingers on the left hand. The largest study to date, which included 360 young men from a normal population, found no relationship of 2:4 digit length ratio with testis or semen parameters \[[@B20]\]. Other studies showed significant negative associations between 2:4 digit length in men and reproductive success; in those studies, the 2:4 digit length ratio was higher in infertile men than in fertile men \[[@B21],[@B22]\]. There are no reports on the relationship between the 2:4 digit length ratio and the occurrence of cryptorchidism or hypospadias. Until now, there has been no report regarding the relationship between birth weight and AGD or digit length in newborn infants. We thus intended to investigate these variables. In the current study, all parameters including height, weight, penile length, testicular size, AGD1-3, and the length of the second and fourth fingers were significantly lower in the LW group than in the NW group. The difference may have been caused by a difference in total body size. In the current study, we excluded newborn infants with hypospadias, concealed penis, cryptorchidism, varicocele, and other growth problems such as chronic renal failure or endocrinologic disorders. We thus assumed that the testosterone exposure of the subjects in this study was at a normal level. The current study results suggested that birth weight is not associated with the AGD ratio or with the 2:4 digit length ratio. In terms of interobserver variability, there was no difference in newborn males with normal birth weight between the raters. However, interobserver variability cannot be completely excluded in the LW group. It is possible that there may have been a difference according to the time of measurement of penile length of newborn males \[[@B14]\]. The penile length measured within 12 hours after birth was 0.31 cm shorter than that remeasured at 1 to 7 days of age in 63 infants. CONCLUSIONS =========== The penile length in children has increased significantly, whereas that in newborn infants has not changed over the last quarter century in Korea. With a normal penile appearance, the AGD ratio and the 2:4 digit length ratio are consistent irrespective of birth weight, whereas the AGD, digit length, and penile length are significantly smaller in newborns with low birth weight. It is possible that the difference in penile length may have resulted from interobserver variability. This work was supported by the Priority Research Center Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (2009-0094050). The authors have nothing to disclose. {#F1} ###### Changes in stretched penile length and anthropometric data compared with a previous study conducted in 1987  Variable Present data (n=24) 1987 Data (n=49) p-value ------------------- --------------------- ------------------ --------- Penile size (cm) 3.3±0.2 3.3±0.5 0.445 Testis size (mL ) 1.1±0.4 1.2±0.2 0.248 Height (cm) 49.8±1.9 52.5±2.5 \<0.001 Weight (kg) 3.2±0.5 3.3±0.4 0.393 Values are presented as mean±standard deviation. Modified from Chung KH, et al. Korean J Urol 1987;28:255-8 \[[@B8]\]. ###### Differences in penile length, digit lengths, anogenital distances, and anthropometric data according to birth weight  Variable Normal birth weight group (n=24) Low birth weight group (n=31) p-value ------------------------ ---------------------------------- ------------------------------- --------- Penile size (cm) 3.3±0.2 2.9±0.4 \<0.001 Testis size (mL ) 1.1±0.4 0.8±0.4 \<0.001 Height (cm) 49.8±1.9 44.1±2.9 \<0.001 Weight (kg) 3.2±0.5 2.1±0.4 0.006 SPL/height (×10^-2^) 6.9±0.5 6.3±0.4 \<0.001 Digit 2 (cm) 2.6±0.2 2.2±0.2 \<0.001 Digit 4 (cm) 2.8±0.3 2.4±0.2 \<0.001 Digit 2/4 ratio 0.92±0.03 0.93±0.05 0.419 AGD 1^a^ (cm) 4.2±0.3 3.6±0.4 \<0.001 AGD 2^b^ (cm) 3.5±0.3 3.1±0.3 \<0.001 AGD 3^c^ (cm) 2.3±0.2 2.0±0.2 \<0.001 AGD 3/height (×10^-2^) 4.8± 0.5 4.6±0.4 0.056 AGD 1-3 (cm) 1.8±0.3 1.6±0.3 \<0.001 AGD 1-2 (cm) 0.6±0.1 0.5±0.1 0.003 AGD (1-2)/(1-3) ratio 0.35±0.05 0.34±0.05 0.478 Values are presented as mean±standard deviation. SPL, stretched penile length; AGD, anogenital distance. ^a^:AGD1 was the distance measured from the anterior aspect of the penis to the anal verge. ^b^:AGD2 was the distance measured from the posterior aspect of the penis to the anal verge. ^c^:AGD3 was the distance measured from the posterior aspect of the scrotum to the anal verge.

I[NTRODUCTION]{.smallcaps} {#sec1-1} ========================== The World Health Organization (WHO)in 1948 defined health as "A state of complete physical, mental, and social well-being and not merely an absence of disease or infirmity."\[[@ref1]\] It defined mental health as "a state of well-being in which every individual realizes his or her own potential, can cope with the normal stresses of life and is able to make a contribution to his or her community."\[[@ref2]\] There is a great public health significance for mental and behavioral disorders since they are among the most important causes of morbidity and burden in primary care and the lead to disability in affected individuals, loss of resources, and productivity.\[[@ref3]\] Psychological distress has been defined as a state of emotional suffering with predominant symptoms of depression and anxiety.\[[@ref4]\] The assessment of psychological distress is done using standardized self-administered or interviewer-administered scales such as General Health Questionnaire, Primary Health Questionnaire, Self- Reporting Questionnaire (SRQ), and Kessler scales.\[[@ref5]\] Screening for psychological distress can be an essential first step in planning of community mental health services. One of the barriers to provision of mental health services is availability of minimal resources to care for those suffering from mental disorders. This "treatment gap" can be best addressed by "decentralization of mental health services" as advocated by the WHO. The promotion of mental health forms one of the components of "Primary Health Care," and primary care for mental health must be coordinated with a network of existing or new services as required at different levels.\[[@ref6]\] M[ETHODOLOGY]{.smallcaps} {#sec1-2} ========================= A cross-sectional community-based survey was done from 2012 to 2013 in the village of Nitte of Udupi district with the help of medico social and psychiatric social workers of the psychiatry department that provides community mental health services at the Nitte Community Health center. The population statistics of the village and the village map was obtained from the village panchayat. The village has been divided into eight wards (I to VIII) and three divisions (Nitte A, B, and C) for electoral and administrative purposes. The sample size was calculated to be 784 with the earlier known prevalence of common mental disorders at 2%\[[@ref7]\] and an absolute error of 1%. Taking a nonresponse rate of 20%, a total of 940 persons were to be contacted. With the average number of adult members per household taken as three, the number of households surveyed in the present study was taken as 310. Systematic random sampling was done, and the sampling interval was calculated to be 8. All adults belonging to 18--65 years\' age group were included, excluding those with severe psychiatric and medical history who could not provide a reliable and adequate history. After obtaining approval from the institutional ethical committee, we conducted a door-to-door inquiry of each household as a unit for sociodemographic details and of each individual adult member of the family for screening. The details of the head of the family were obtained, and socioeconomic status was assessed by Udai Pareek Scale.\[[@ref8]\] They were screened with the help of psychiatry social worker who is adequately trained on the WHO Self-Reporting Questionnaire (SRQ) after obtaining informed consent. SRQ is an instrument to screen for psychiatric disturbance in primary health-care settings, especially in developing countries.\[[@ref9]\] A cutoff value of 8 in SRQ was taken as screening positive. A score of 8 has shown a sensitivity of 79%, specificity of 96%, positive predictive value of 75%, negative predictive value of 97%, and maximum (93.6%) cases were screened.\[[@ref10]\] Cross verification of the data was done by the principal investigator. Descriptive data were analyzed in percentages and proportions, whereas the associations were analyzed using appropriate tests of significance. R[ESULTS]{.smallcaps} {#sec1-3} ===================== A majority of the study population of 492 participants (52.2%) belonged to class 3 socioeconomic status. The study population was equally distributed among the various age groups ranging from 18 to 65 years and gender -- males (424, 45%) and females (519, 55%). Majority of the population (792, 84%) belonged to Hindu religion, were married and staying with their spouses (590, 62.6%), and were unemployed (423, 44.9%) (which also included the homemakers and the students). Most were educated up to middle school (210, 22.3%), 88 (9.3%) were illiterate, and majority (424, 45%) had no individual source of income \[[Table 1](#T1){ref-type="table"}\]. The most common physical co-morbidity was Diabetes mellitus and Hypertension \[[Figure 1](#F1){ref-type="fig"}\]. ###### Distribution of study population according to sociodemographic characteristics and the determinants of psychological distress (*n*=943) Psychological distress *P* --------------------------------- ------------------------ ------------ --------- SE status  Class 2 4 (4.7) 82 (95.3) 0.833  Class 3 19 (3.9) 473 (96.1)  Class 4 17 (4.7) 348 (95.3) Religion  Hindu 34 (4.3) 758 (95.7) 0.405  Muslim 2 (2.2) 89 (97.8)  Christian 4 (6.7) 56 (93.3) Age  18-25 4 (2.1) 188 (97.9) 0.209  26-35 5 (2.8) 175 (97.2)  36-45 9 (4.7) 183 (95.3)  46-55 13 (6.4) 191 (93.6)  56-65 9 (5.1) 166 (94.9) Gender  Male 10 (2.4) 414 (97.6) 0.04\*  Female 30 (5.8) 489 (94.2) Marital status  Unmarried 3 (1.3) 220 (98.7) 0.000\*  Married but staying separate 1 (14.3) 6 (85.7)  Divorced 2 (100.0) 0  Widow/widower 7 (5.8) 114 (94.2)  Married and staying together 27 (4.6) 563 (95.4) Occupation  Unemployed 17 (4.0) 406 (96.0) 0.287  Unskilled worker 12 (6.5) 174 (93.5)  Semiskilled worker 1 (1.1) 94 (98.9)  Skilled worker 2 (2.1) 93 (97.9)  Clerical, shop owner, farmer 7 (5.9) 111 (94.1)  Semiprofessional 1 (7.1) 13 (92.9)  Professional 0 12 (100.0) Education  Illiterate 11 (12.5) 77 (87.5) 0.000\*  Primary school certificate 12 (8.0) 138 (92.0)  Middle school certificate 8 (3.8) 202 (96.2)  High school certificate 5 (2.4) 203 (97.6)  Intermediate, post high school 1 (.5) 205 (99.5)  Graduate or postgraduate 3 (7.7) 36 (92.3)  Professional or honors 0 42 (100.0) Individual monthly income (Rs.)  \>20,000 0 12 (100.0) 0.152  10,000-20,000 1 (2.6) 37 (97.4)  5000-10,000 3 (2.3) 129 (97.7)  1000-5000 6 (2.8) 209 (97.2)  \<1000 10 (8.2) 112 (91.8)  None 20 (4.7) 404 (95.3) Physical morbidity  Absent 21 (3.6) 558 (96.4) 0.174  1 morbidity 14 (4.6) 293 (95.4)  \>1 morbidity 5 (8.8) 52 (91.2) Tobacco  Yes 9 (4.8) 179 (95.2) 0.402  No 31 (4.1) 724 (95.9) Alcohol consumption  Yes 5 (12.5) 117 (95.9) 0.583  No 35 (4.3) 786 (95.7) \**P* value statistically significant, SE: Socioeconomic {#F1} Of the 943 participants surveyed, psychological distress was present in forty with a prevalence of 42.4 per thousand. The most common individual item answered positive in the SRQ was "being easily tired" (332, 35.3%) and "feeling tired all the time" (266, 28.3%). The least commonly checked items were "the thought of ending one\'s life" (3.1%) and "feeling of being a worthless person" (4.1%). There was a significant association of gender with psychological distress (X^2^= 6.727, *P* = 0.009) (odds ratio \[OR\] females: males = 2.5), and the proportion of psychological distress among females (5.8%) was twice compared to males (2.4%). A highly significant association between marital status and psychological distress was also seen (X^2^= 52.367, *P* = 0.000) (OR married and separate: married and together = 3.47). Illiterates had the highest prevalence (12.5%), followed by those educated up to primary school (8%). This association of education with psychological distress was highly significant (X^2^= 31.977, *P* = 0.000) (OR of illiterate + primary: others = 4.3) \[[Table 1](#T1){ref-type="table"}\]. However, illiterates and married but staying separate/divorced were the only two groups that were found to have higher adjusted odds of psychological distress (20.007 and 6.617, respectively) on multinomial logistic regression analysis with bootstrapping \[[Table 2](#T2){ref-type="table"}\]. ###### Determinants of psychological distress in multinomial logistic regression analysis Psychological distress Adjusted odds (95% CI) *P* ---------------------------------------- ------------------------ --------- Gender  Female 1.852 (0.877-3.911) 0.106  Male Reference group Marital status  Unmarried 0.563 (0.159-1.990) 0.373  Married but staying separate/divorced 6.617 (1.459-29.996) 0.014\*  Widowed 0.708 (0.287-1.743) 0.452  Married and staying together Reference group Education  Illiterate 20.007 (1.7-247.2) 0.019\*  Primary school 5.835 (0.4-72.1) 0.169  Middle school 1.881 (0.15-24.1) 0.627  High school 1.627 (0.13-21.2) 0.710  Intermediate, posthigh school 1.169 (0.08-16.2) 0.907  Graduate or postgraduate 6.564 (0.47-92.7) 0.164  Professional or honors Reference group \**P* value statistically significant. CI: Confidence interval Females scored significantly higher in SRQ with higher mean SRQ rank value (508.42) compared to males (424). As the age progressed, median SRQ scores and mean SRQ rank values increased showing a significant positive correlation between SRQ score and age (Spearman\'s rho = 0.323, *P* = 0.000). The highest mean rank SRQ was seen in divorced persons (940.5), whereas it was least in unmarried individuals (347.76) with a significant association between marital status and SRQ score. D[ISCUSSION]{.smallcaps} {#sec1-4} ======================== Our study found the percentage of psychological distress to be 40 per thousand (4%). Drapeau *et al*.\[[@ref4]\] stated that it is difficult to pinpoint the prevalence of psychological distress and to compare the rates in different community surveys due to the differences in the scales assessing distress, of the time windows used in the documentation of symptoms and of the cut points applied to dichotomize and identify individuals with pathological distress. The prevalence rates range from 5% to 10% in the general population according to various studies in different settings,\[[@ref11][@ref12][@ref13]\] but it can be higher in population groups with exposure to some risk factors such as workers employed in stressful work conditions and immigrants. The prevalence is lower in the present study, and this could be because it is a rural, well-knit community with predominantly joint or extended families, income equity and sufficient education (colleges, anganwadis, and schools), and livelihood opportunities (cottage industries) as well as good access to health to the population (health centers). In the present study, there was a significant association of gender with psychological distress and higher SRQ scores. Similar to this, the prevalence of psychological distress is higher in women than in men in most countries as seen in earlier studies.\[[@ref12][@ref14]\] A study among men and women in Goa, India,\[[@ref15]\] found that moderate and high scores of psychological distress were detected in significantly more women than men. The present study found marital status to be a determinant for distress. Those who were married but separated or divorced had a significantly higher adjusted odds ratio of distress compared to those who are married and staying together. However, we did not find significantly higher rates of distress among the widowed. Earlier studies\[[@ref14][@ref16]\] have shown that those who are married have lower rates and those who are widowed or divorced have higher rates of distress. We found a highly significant association of education and distress similar to other studies.\[[@ref14][@ref17]\] As pointed out in these studies, education either directly or indirectly influences levels of distress -- having limited income or education may make one more vulnerable to social problems with distress. C[ONCLUSIONS]{.smallcaps} {#sec1-5} ========================= The prevalence of psychological distress in the current study was low, and somatic complaints were common presentations of psychological distress in the study population. Psychological distress was significantly more common in women, married but staying separate/divorced and illiterates. Recommendations {#sec2-1} --------------- This study highlights the need for further studies to explore the feasibility of providing training for developing skills among health workers with emphasis on detection of somatic complaints as indicator of psychological distress among the vulnerable groups. Interpretation of these symptoms has to be done based on specific cultural context, and we recommend further studies to validate the same. Limitations {#sec2-2} ----------- The SRQ instrument checks distress based on questions on the participant\'s experiences for the past 1 month only. Some studies have also shown a gender difference in the ideal cutoff scores for SRQ 20 which has not been addressed in the current study.\[[@ref18]\] Financial support and sponsorship {#sec2-3} --------------------------------- Nil. Conflicts of interest {#sec2-4} --------------------- There are no conflicts of interest. We would like to acknowledge Dr. Naveen Chandra Shetty, founder of Nitte rural psychiatry project, the staff of Nitte Rural Psychiatry team- Mrs. Madhavi, Mrs. Mallika, Mrs. Navya and Mr. Robin and the entire team of psychiatrists -- Dr. Shrinivas Bhat U, Dr. Nishi Guru, Dr. Smitha, Dr. Aneesh Bhat, and Dr. Satish Rao from the Department of Psychiatry, K S Hegde Hospital.

Background {#Sec1} ========== Chansu is a traditional Chinese medicine extracted from parotoid glands of the Chinese toad (*Bufo gargarizan*). It has been widely used for the treatment of inflammation, anaesthesia and arrhythmia in China, Japan and other Asian countries for centuries \[[@CR1]\]. Cinobufagin (CBF) is a major component isolated and purified in the last decade \[[@CR2]\]. The compound possesses a digoxin-like structure and is a type of sodium/potassium-ATPase inhibitor (Fig. [1](#Fig1){ref-type="fig"}). Recently, other sodium/potassium-ATPase inhibitors have also been reported to impair cancer cell migration through different signalling pathways \[[@CR3]--[@CR7]\].Fig. 1The molecular structure of CBF. CBF possesses a similar function as other typical sodium/potassium-ATPase inhibitors Cortactin is an important factor involved in cancer cell progression and invasion \[[@CR8]\]. It was initially discovered to localise to cortical actin at the cell periphery \[[@CR9]\]. During normal cell migration, cortactin is phosphorylated by Src kinase in the C terminal proline-rich domain, while Arp2/3 complex binds to the N terminal of cortactin. The binding of Arp2/3 complex promotes actin polymerisation and facilitates the stabilisation of branched actins \[[@CR10]\]. Then the formation of cell motility structures like lamellipodia supports the movement of certain types of cells, including osteoclasts and macrophages \[[@CR11]\]. Similarly, phosphorylated cortactin is also able to initiate actin assembly but to form invadopodia in cancer cells, followed by extracellular matrix (ECM) degradation. As a result, detached cells invade surrounding tissues. Thus, cortactin is used as a marker for detection of invadopodia and enrichment of cortactin indicates the metastatic level in a number of cancers \[[@CR12]\]. Colon cancer is one of the leading cancer deaths worldwide and most of the patients died from metastatic diseases \[[@CR13]\]. Previously, overexpression of cortactin was revealed in a few types of colorectal cancers \[[@CR14]\]. Compared with normal tissues, primary cancerous colorectal tissues showed assembly of cortactin in lymph nodes, suggesting a close involvement of cortactin in metastasis of colorectal cancer cells. In previous work, we demonstrated that CBF induced strong apoptosis in colon cancer cells and a hypoxia-regulated pathway was involved in the drug effect \[[@CR15]\]. In the current study, we used CBF as a surrogate marker for Chansu, and further determined the role of CBF in the modulation of cortactin expression in human colon cancer cells. Moreover, we established the effects of CBF treatment under both hypoxic and normoxic conditions, as several of these ion pump inhibitors, including digoxin and ouabain, have been shown to play an anti-cancer role by diminishing hypoxia-inducible factor 1 alpha (HIF-1α) expression \[[@CR16], [@CR17]\]. Subsequently, mice bearing HCT116 tumour were also used to reveal that CBF induced a reduction of cortactin synthesis in tumour tissues. Taken together, our findings suggest that the mechanism of Chansu for inhibition of colon cell invasion could be through blocking the interaction between cortactin and actin. Methods {#Sec2} ======= Cell culture and drug treatment {#Sec3} ------------------------------- HCT116 and HT29 colon cancer cell lines (provided kindly by Dr Albert Mellick, Griffith University, Australia) were maintained in DMEM (Gibco) supplemented with 10 % FBS (HyClone), 2 mM GlutaMAX (Gibco), 100 U/ml penicillin (Sigma), 100 μg/ml streptomycin (Sigma), 110 mg/L sodium pyruvate (Gibco) and 25 mM HEPES (Gibco), in an atmosphere of 5 % CO~2~ and 95 % air at 37 °C. CBF was purchased from Sigma and dissolved in DMSO (Sigma). 1 μM of CBF solution was applied in all cell line experiments. To achieve a hypoxic condition, cell lines were sealed in GasPak pouches (BD) and placed in an incubator at 37 °C. RNA extraction and real-time RT-PCR {#Sec4} ----------------------------------- In cell lines, total RNA was isolated at different time points using a PureLink RNA mini kit (Ambion), according to the manufacturer's instructions. For RNA extraction from mouse tumour tissues, fresh tissues (\<125 mm^3^) were initially submerged in RNA*later* solution (Ambion) and stored at 4 °C overnight. The subsequent RNA extraction was carried out by following the instructions of TRIzol reagent (Ambion). The concentration of all RNA samples was measured by NanoDrop ND-1000 (Thermo Fisher Scientific). cDNAs of interest were synthesised using SuperScript III RT kit (Invitrogen) according to the manufacturer's instructions. Real-time PCR was carried out in 20 μl of reaction solution, consisting of 0.4 μM primers (Sigma), 10 μl of Express SYBR GreenER qPCR SuperMixes (Invitrogen) and ddH~2~O. Real-time PCR was performed in iQ5 multicolour real-time PCR detection system (Bio-Rad). The reaction conditions were 50 °C for 2 min and 95 °C for another 2 min, followed by 40 cycles of 95 °C for 15 s, 60 °C for 1 min. Melting curves were monitored by heat-denaturing amplicons over a 35 °C temperature gradient at 0.5 °C/s from 60 to 95 °C. No genomic DNA contamination or pseudogenes were detected. Primers used in real-time PCR were: Human cortactin (Forward: 5′ - AGG TGT CCT CTG CCT ACC AGA A - 3′, Reverse: 5′ - CCT GCT CTT TCT CCT TAG CGA G -3′). Human GAPDH (Forward: 5′ - GTC TCC TCT GAC TTC AAC AGC G - 3′, Reverse: 5′ - ACC ACC CTG TTG CTG TAG CCA A - 3′). Western blotting {#Sec5} ---------------- Cells were scraped in cold PBS and centrifuged down (500 × *g*) to remove methanol. The pellet was resuspended in cold RIPA buffer (Pierce), supplemented with Complete protease inhibitor cocktail tablets (Roche). After centrifugation at 13,000 × *g* for 10 min, the supernatant was collected for further analysis. For protein extraction from mouse tumour tissues, frozen tissues were ground in a mortar and pestle and then immersed in cold RIPA buffer plus protease inhibitor. Further homogenisation was performed by passing the tissues 5--10 times through a 21-gauge needle. After centrifugation at 13,000 × *g* for 10 min, the supernatant was collected and mixed with 1X SDS sample buffer. Protein samples were loaded onto 7 % or 12 % SDS-PAGE gels, running in Mini Trans-Bolt module (Bio-Rad). After gel electrophoresis, proteins were transferred to PVDF membranes (Millipore). The membranes were incubated with primary antibodies against cortactin (1:2000, Abcam) and α-Tubulin loading control (1:5000, Abcam) overnight at 4 °C after a 45 min blocking. Horseradish peroxidise-conjugated goat anti-mouse and anti-rabbit (1:10000, Bio-Rad) secondary antibodies were applied afterwards. SuperSignal chemiluminescent substrate (Pierce) was added to the membranes which were visualised using a VersaDoc MP4000 system (Bio-Rad). Subcellular protein extraction {#Sec6} ------------------------------ 5 × 10^6^ treated HCT116 cells were detached, centrifuged down to form a cell pellet and snap frozen in liquid nitrogen. The subcellular protein extraction followed the manufacturer's instructions of ProteoExtract Subcellular Proteome Extraction Kit (Calbiochem). Briefly, frozen cell pellets were washed twice in wash buffer and exposed to Extraction Buffer I plus protease inhibitor cocktail. After centrifugation, the supernatant was collected as the cytosolic protein fraction and the pellet was resuspended Extraction Buffer II to isolate the fraction of membrane/organelle proteins. After a series of centrifugation and usage of specific extraction buffer III \~ IV, the nuclear and the cytoskeletal matrix protein fractions were separated. Immunocytochemistry {#Sec7} ------------------- Sterile coverslips were placed in a 24-well plate and cancer cells were seeded at a density of 1.5 × 10^5^ cells per well. After overnight incubation at 37 °C, cells were adherent to the coverslips. Treated cells were washed in ice cold PBS and fixed in 100 % methanol at −20 °C for 10 min, followed by washing in ice cold PBS twice, shaking gently. Usage of 0.2 % Triton X-100 (Sigma) in PBS to permeabilise samples was for no more than 10 min, followed by 3 times of wash in PBST. Cells were then blocked in block buffer (3 % normal goat serum and 0.5 % BSA in 0.01 M PBS) for 30 min and sequentially incubated with primary antibody cortactin (1:300, Abcam) overnight at 4 °C. The goat anti-rabbit Texas Red (1:1500, Abcam) secondary antibody was applied in dark for 1 h, followed a counter DAPI staining (Molecular Probes). All the coverslips were sealed onto microscope slides using ProLong Gold antifade reagent (Molecular Probes) and kept in dark for 24 h. Fluorescence images were visualised using confocal microscope FV1000 (Olympus). Xenograft establishment and CBF treatment {#Sec8} ----------------------------------------- To establish nude mouse models bearing HCT116 tumour, 5 × 10^6^ HCT116 cells in PBS (200 μl) were subcutaneously injected into each mouse using 30 - G needles. After tumour growth for 2 weeks, 14 female BALB/c nude mice (aged 8 weeks and weighing 16--18 g) were equally divided into two groups (7 mice per group, *N* = 7): intraperitoneal (i.p.) injection and control, with average tumour size of each group of about 0.32 cm^3^. CBF was dissolved firstly in absolute ethanol and diluted in 10 % propylene glycol solution. To prepare 50 ml injection solution, 10 mg of CBF was dissolved in 4 ml of absolute ethanol and diluted in 10 % propylene glycol solution to reach 0.2 mg/ml. The daily dose given to i.p. group was 1.5 mg/kg, while an equal amount of injection solution without CBF was given as a control. Mice were sacrificed when the tumour grew to 1 cm^3^. Tumour tissue specimens were taken for subsequent RNA and protein analyses. All experiments involving animals were approved by Griffith University (AEC No. MSC/01/08). Results {#Sec9} ======= CBF inhibits mRNA and protein expression of cortactin in HCT116 cells {#Sec10} --------------------------------------------------------------------- Cortactin is overexpressed in colon cancer tissues \[[@CR18]\] and has been shown to be an important factor in tumour progression and cancer invasion. In this study, we treated colon cancer cells (HCT116 and HT29) with CBF (1 μM). In order to mimic the colon cancer microenvironment, the colon cancer cells were also exposed to hypoxic and normoxic conditions, respectively (Fig. [2a](#Fig2){ref-type="fig"}). Our results showed that there was a decreased expression of cortactin mRNA in HCT116 cells under 1 % oxygen. Such a decrease was swift, but only lasted less than 24 h, followed by a significant increase of cortactin transcription. As the level of oxygen increased to 20 %, the transcription of cortactin, however, exhibited an overall reduction in HCT116 cells within 24 h. On the other hand, cortactin mRNA level in HT29 cells only slightly altered during the initial 12 h under different oxygen conditions. Subsequently, a sharp elevation of mRNA level was observed at 24 h with 2 folds increase in hypoxia and almost 5 folds increase in normoxia.Fig. 2CBF affected cortactin mRNA and protein expression in human colon cancer cell lines HCT116 and HT29. **a** The mRNA transcription of cortactin in CBF-treated HCT116 and HT29 cells. 1 μM of CBF was added to HCT116 and HT29 cells which then incubated under hypoxic and normoxic conditions. The cells were collected at different time points and analysed. Results are means with standard errors from four replicates. The level of GAPDH as a control was set to 1.0. **b** Cortactin protein inhibition only in HCT116 cells (1 % oxygen). The expressions of cortactin were significantly inhibited by CBF in HCT116 cells at 6 h and 12 h, but undetectable in HT29 cells The expression of cortactin protein under hypoxic conditions was consistent with mRNA message (Fig. [2b](#Fig2){ref-type="fig"}). The inhibition of cortactin at 6 and 12 h in HCT116 cells was significant, but not at 24 h. This inhibitory role was barely detected in CBF-treated HT29 cells at any time points. Taken together, our finding showed that CBF inhibits mRNA and protein expressions of cortactin in HCT116 cells and this inhibition is swift but not sustainable. In fact, all the increases of cortactin mRNA at 24 h time point implied an interference between CBF-induced inhibition and cortactin transcription. Such elevated mRNA levels might also result in no down-regulation of protein expression at 24 h. CBF affects distribution of cortactin in HCT116 cells under hypoxic conditions {#Sec11} ------------------------------------------------------------------------------ To further investigate CBF-induced cortactin inhibition, we examined the subcellular protein expression in HCT116 cells in hypoxia. Subcellular protein fractions indicated that the level of cortactin in cytoskeletal fraction was diminished after CBF treatment, suggesting the dissociation of cortactin and cytoskeletal proteins (Fig. [3a](#Fig3){ref-type="fig"}). Generally, cortactin is localised in the cell periphery. However, immunostaining showed that CBF exposure led cortactin to shift towards the nucleus (Fig. [3b](#Fig3){ref-type="fig"}).Fig. 3CBF changed the distribution of cortactin in HCT116 cells (1 % oxygen). **a** Subcellular protein extraction of four fractions, which are cytosolic, membrane/organelle, nucleic and cytoskeletal fractions. Cortactin expression was diminished significantly in the fraction of cytoskeletal proteins after exposure to 1 μM of CBF. **b** Co-localisation of cortactin in treated HCT116 cells. Under a hypoxic condition, HCT116 cells were incubated with 1 μM CBF for 24 h. The subsequent staining revealed a colour overlapping, indicating that CBF induced a nuclear translocation of cortactin. Scale bars equal 10 μm CBF inhibits mRNA and protein expression of cortactin in nude mouse models bearing HCT116 tumour {#Sec12} ------------------------------------------------------------------------------------------------ HCT116 cells were implanted in lateral right back of nude mice (Fig. [4a](#Fig4){ref-type="fig"}). When the average of tumour size reached 0.32 cm^3^, the CBF treatment started by i.p. injection. The mice were sacrificed once the tumour grew to 1 cm^3^ and tumour tissues were collected. Between two groups, the average of tumour size was found to be slightly suppressed by CBF in i.p. group (Fig. [4b](#Fig4){ref-type="fig"}). Moreover, cortactin mRNA level in i.p. group was significantly reduced (Fig. [4c](#Fig4){ref-type="fig"}). The RT-PCR was conducted using mouse tumour tissue samples from different groups. Interestingly, the protein inhibition was found in the treatment group (Fig. [4d](#Fig4){ref-type="fig"}), demonstrating that CBF repressed cortactin synthesis in xenografts.Fig. 4CBF inhibited cortactin synthesis in xenografts. **a** A nude mouse model bearing HCT116 tumour. The mouse was sacrificed when the tumour size reached 1 cm^3^. **b** Tumour size vs days of treatment (N = 7). Compared with the control, the i.p. group showed a slight suppression of tumour growth after the drug injection. **c** Analysis of cortactin mRNA level in tumour tissues. A significant decrease of cortactin mRNA level presented in i.p. group. Results are means with standard errors from four replicates. The level of GAPDH as a control was set to 1.0. **d** Abundant cortactin protein expression only in the control group. The i.p. group showed diminished expression of cortactin. The absence of cortactin in lane 6 of control group was an exemption with an unknown reason Discussion {#Sec13} ========== Recently, Chansu is being used as an anticancer agent. Clinical use has shown that it improved the quality of life in patients with advanced gallbladder cancer \[[@CR19]\]. As a major component in Chansu, CBF was used as surrogate marker to study the mechanism of anticancer activity of Chansu. We previously demonstrated that CBF (1 μM) induced effective apoptosis in colon cancer cell lines HCT116 and HT29, with 57 % and 30.5 % of cell death, respectively \[[@CR15]\]. We also showed that a key mediator of cell motility and metastasis, cortactin, was significantly inhibited by CBF in HCT116 cell line and in xenografts HCT116 tumours. Although the overexpression of cortactin in HCT116 cells was identified in previous studies \[[@CR14], [@CR20]\], this report also showed that there is a high level of cortactin in colon cancer cell line HT29 as well. The inhibition of transcription and protein levels of cortactin in HCT116 cells was observed in HCT116 but not in HT29 cells, reflecting that CBF appears to induce strong suppression in HCT116 cells but has limited efficacy on HT29 cells. Therefore, CBF is unlikely to strongly target cortactin in all colon cancer cells. On the other hand, the inhibition occurred swiftly but was overcome within 24 h. As the half-life of cortactin protein is about 8.9 h \[[@CR21]\], CBF seems to have a fast on-off effect on the protein. A sharp increase of cortactin mRNA under hypoxic conditions at 24 h revealed a possibility that the cancer cells in a hypoxic tumour microenvironment appear to be able to resistant to CBF. Moreover, the 24 h increase of cortactin mRNA was detected in both cell lines regardless to the oxygen levels, suggesting that it could be resulted from the up-regulation of mitogen-activated protein kinase (MAPK) pathways \[[@CR22], [@CR23]\]. Drugs like ouabain and digoxin have been shown to be capable of activating MAPK by binding to sodium/potassium-ATPases, resulting in the release of Src kinase \[[@CR22], [@CR23]\]. As cortactin is a substrate for Src \[[@CR9]\], the transcription level of cortactin could be elevated, followed by abundant expression of cortactin protein under our assay conditions. Apart from the *in vitro* experiments in 1 % oxygen, the suppression of cortactin was also shown in tissue samples of nude mice bearing HCT116 tumours. The results confirmed the inhibitory role of CBF in xenografts *in vivo*. Although cortactin inhibition did not sustain in HCT116 cells within 24 h, CBF is still able to block cortactin synthesis in HCT116 tumour tissues of mouse models. We hypothesized that this could be due to the down-regulation of nuclear factor kappa beta (NF-κB), as Hill *et al.* elucidated that inhibition of p65 Rel A subunit of NF-κB by IκKinase-2 led to a significant suppression of cortactin mRNA transcription in breast cancer MCF7F-B5 cells \[[@CR24]\]. Our previous results from multi-pathway arrays revealed that CBF significantly impedes NF-κB activity in HCT116 cells \[[@CR15]\]. Thus, the long-term inhibition of cortactin *in vivo* could be a consequence of CBF-induced deactivation of NF-κB. Taken together, our data establishes the inhibitory role of CBF in cortactin synthesis in a HCT116 cells. In addition to the general down-regulation of cortactin protein, the CBF treatment also altered the distribution of cortactin in HCT116 cells under hypoxic conditions. The nuclear translocation of cortactin in cancer cells was a new observation. Here, the cause for this shift is still unclear. Oddly, the total level of cortactin in nucleus did not boost after the drug exposure, which left a question whether cortactin was only dragged to the nucleic surface and then degraded swiftly. The pioneer work of Hering and Sheng (2003) showed a nuclear accumulation of cortactin mutant in dendritic spine morphogenesis \[[@CR25]\]. They discovered that there is a tandem repeat region within the N-terminal half of cortactin and the deletion of this region targets cortactin to the nucleus of dendritic cells rather than dendritic spines. Thus, one possible explanation for the nuclear import of cortactin in CBF treated cells is the interruption of the tandem repeats. Nevertheless, the nuclear translocation of cortactin was not detected in HT29 cells or the tissue samples from mouse models bearing HCT116 tumours. Therefore, more investigations of the nuclear import need to be done in the future. Conclusions {#Sec14} =========== In conclusion, CBF possess the unique capacity to inhibit the overexpression of cortactin in HCT116 cells and nude mouse models bearing HCT116 tumours. It seems likely that this is the mechanism for Chansu to be used as an anticancer agent, inhibiting colon cancer cell proliferation and metastasis. CBF : Cinobufagin HIF-1α : Hypoxia-inducible factor 1 alpha ECM : Extracellular matrix i.p. : Intraperitoneal MAPK : Mitogen-activated protein kinase NF-κB : Nuclear factor kappa beta **Competing interests** The authors declare that they have no competing interests. **Authors' contributions** CL drafted and prepared the manuscript. SMH, DG and WD proof-reading and corrections for the manuscript. SC help with the animal experiment. JQ help with preparation of the manuscript. MQW correction, proof reading and the final approval of the manuscript. All authors read and approved the final manuscript. This work was supported by the Dr. Jian Zhou smart state fellowship from the Queensland government, and grants from the National Health and Medical Research Council and Cancer council, Queensland to MQW. We would like to thank other members of Wei's Laboratory for their support and helpful comments.

Significance Statement {#s1} ====================== Cortico-basal ganglia circuits control a range of neurobiological functions, ranging from motor control and reward to cognition. The functional diversity of cortico-basal ganglia circuits rests on their diverse inputs from the cerebral cortex. It is likely that the highly heterogeneous nature of corticostriatal inputs makes the corticostriatal circuits vulnerable to a broad range of neurologic and psychiatric disorders. Here, we highlight the developmental progression and maturation of the morphology and physiology of corticostriatal pathways using neonatal and postnatal rodent brains. We also review the pathogenesis of various neurodevelopmental disorders that are related to dysfunctions of corticostriatal circuits. Exploring synaptic wiring of corticostriatal circuits should create a research window for the development of therapeutic approaches for treating basal ganglia-related neurologic disorders. Introduction {#s2} ============ An essential function of neural networks is the processing of sensory inputs and the generation of motor outputs. The cortico-basal ganglia circuitry is in a key hub that is involved in the integration of sensory and motor information by the brain. The striatum of the basal ganglia receives a large number of cortical inputs from the motor, sensory, association and limbic cortices ([@B71]; [@B76]). The corticostriatal afferents thus make up a wide range input of different nature into the basal ganglia. The corticostriatal inputs are highly heterogeneous and therefore the cortico-basal ganglia circuits are involved in processing a broad spectrum of neurobiological functions. These range from motor control at a basic level to high level adaptive learning and cognition ([@B5]; [@B63]; [@B123]; [@B68]). The integration of sensory and motor information is essential for behavioral performance and thus it is not surprising that dysfunction of cortico-basal ganglia circuits is well documented in a broad range of neurologic and mental disorders, including Parkinson\'s disease, Huntington\'s disease, attention-deficit hyperactivity disorder (ADHD), Tourette syndrome, obsessive-compulsive disorder (OCD), autism spectrum disorder (ASD), schizophrenia, and speech and language disorders ([@B31]; [@B138]; [@B66]). Many psychiatric disorders are rooted in developmental dysfunctions that affect neural circuits within the brain. Thus, it is imperative to understand how neural circuits are formed during development. In this review, we have focused on the construction at the cellular and molecular level during development of the corticostriatal circuits. We have summarized the developmental maturation timeline of corticostriatal innervations in [Figure 1](#F1){ref-type="fig"}. {#F1} Genetic studies using mouse models have shown that many genes important to the development of neural circuits are involved in the pathogenesis of neurologic and psychiatric disorders. The significance and biological functions of these building blocks of neural circuits are, in fact, best illustrated by pathologic studies of neuropsychiatric diseases. We have therefore also reviewed the pathophysiology of neuropsychiatric disorders where it is known that dysfunction of corticostriatal circuits is involved. A summary of the various animal model studies that have shown links to corticostriatal abnormalities and neuropsychiatric diseases is presented in [Table 1](#T1){ref-type="table"}. ###### Susceptible genes in neurodevelopmental diseases that are modeled in transgenic mice Gene Associated diseases Morphological and functional phenotypes in corticostriatal circuits of transgenic mice carrying defective or variant alleles of neurodevelopmental disorder-risk genes References ----------- --------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------ *Drd4* ADHD ↓ corticostriatal glutamate release [@B17] *Shank3b* ASDOCD ↑ dendritic arborizations↓ synaptogenesis/spinogenesis↑ precocious hyperactivity of corticostriatal inputs during development↓ corticostriatal synaptic transmission in adulthood [@B120] [@B121] *Fmr1* ASD ↑ inhibitory neurotransmission in the striatum↓ corticostriatal plasticity (LTD)↓ corticostriatal connectivity [@B25] [@B79] [@B169] *Foxp1* ASD ↑ excitability of iSPNs [@B7] *elf4e* ASD ↑ spinogenesis in Layer II/III of medial prefrontal cortex↑ corticostriatal plasticity (LTD) [@B136] *Nlgn1* ASD ↓ NMDA-mediated synaptic transmission of dSPNs↓ mEPSCs in iSPNs [@B16] [@B46] *Nlgn3* ASD ↓ inhibitory synaptic transmission in dSPNs of the ventral striatum↓ corticostriatal plasticity (LTD) [@B135] [@B99] *Tshz3* ASD ↑ corticostriatal LTP [@B24] *Met* ASDSLD ↑ neuronal activity of corticostriatal pyramidal neurons in Layer Vb [@B128] *Mef2c* ASDSLD ↑ synaptogenesis/spinogenesis in the striatum↑ mEPSCs in SPNs [@B26] *Cntnap2* ASDSLD ↓ number of striatal GABAergic interneurons [@B122] *Slitrk5* OCD ↓ BDNF-mediated neurite outgrowth of striatal neurons↑ neuronal activity of the orbitofrontal cortex↓ corticostriatal transmission↓ dendritic complexity of SPNs↓ GluR2, NR2B in the striatum [@B145] [@B140] *Sapap3* OCD ↓ postsynaptic density thickness↑ NR1, NR2B and↑NR2A in the striatal PSD fraction↓ corticostriatal field EPSP↓ corticostriatal quantal EPSP in the iSPNs↓ corticostriatal feedforward inhibition of fast-spiking interneurons [@B163] [@B161] [@B21] *Slc1a1* OCD ↓ neuronal activity in response to amphetamine in the dorsal striatum↓ NR2A and corticostriatal LTD in the dorsal striatum [@B170] [@B35] *ErbB4* Schizophrenia ↑ inhibitory synaptic transmission in the striatum↑ GABA~A~α1 in the striatum [@B56] *Zswim6* Schizophrenia ↓ neurite arborizations in the striatum↓ spinogenesis in the striatum [@B153] *Foxp2* SLD ↓ corticostriatal synaptic transmission↓ corticostriatal plasticity (LTD)↓ synaptogenesis/spinogenesis in the striatum↓ mEPSCs in SPNs [@B65] [@B45] [@B130] [@B26] ADHD: attention-deficit hyperactivity disorder; ASD: autism spectrum disorder; BDNF: Brain-derived neurotrophic factor; dSPN: direct striatonigral pathway neuron; EPSC: excitatory postsynaptic current; EPSP: excitatory postsynaptic potential; GABA: γ-aminobutyric acid; GABA~Aalpha1~: GABA~A~ receptor subunit alpha 1; GluR2: glutamate ionotropic receptor AMPA type subunit 2; iSPN: indirect striatopallidal pathway neuron; LTD: long-term depression; LTP: long-term potentiation; NR1: glutamate ionotropic receptor NMDA type subunit 1; NR2A: glutamate ionotropic receptor NMDA type subunit 2A; NR2B: glutamate ionotropic receptor NMDA type subunit 2B; mEPSC: miniature excitatory postsynaptic currents; OCD: obsessive-compulsive disorder; PSD: postsynaptic density; SLD: speech and language disorders; SPN: striatal projection neurons. Establishment over Time of Corticostriatal Projections {#s3} ====================================================== Corticofugal pathways form the major afferent inputs to the basal ganglia. In the developing mouse brain, corticofugal projection neurons are classified into two cell types: pyramidal tract (PT)-type cortical projection neurons and intratelencephalic (IT) projection neurons ([@B131], [@B132]). Previous studies have revealed the developmental projections of the PT-type corticostriatal pathway. [@B113] and [@B139] traced the developing corticofugal axons by placing DiI in fixed brain slices from a single hemisphere. The majority of the ipsilaterally DiI-labeled corticofugal axons are presumably derived from PT-type corticostriatal neurons, this is because the DiI-labeled axons can be traced and shown to exit the telencephalon to form part of the PT. However, because IT-type neurons project into both the ipsilateral and contralateral striatum, the possibility that some of the ipsilaterally DiI-labeled axons are derived from IT-type neurons cannot be excluded. Corticofugal axons have been found to enter the ipsilateral striatal anlage as early as embryonic day (E)12--E13. Corticofugal axons, presumably en route through the striatal analge during E12--E13, continue growing into the thalamus during E15--E17. By E18, clusters of collateral branches of corticofugal axons can be found in the ipsilateral striatum. Notably, these corticostriatal fibers are predominately co-localized with the tyrosine hydroxylase-positive nigrostriatal dopamine islands that are known to mark the loci of developing striosomes before the postnatal day (P)8 ([@B113]; [@B139]). To explore the developmental projections of IT-type cortical neurons, [@B144] performed retrograde and anterograde labeling of corticostriatal axons during the first two postnatal weeks, and found that IT-type corticostriatal axons enter the contralateral striatum during P3--P4, and these corticostriatal axons progressively refined their innervation over the first two weeks after birth ([@B144]). Immunostaining of growth-associated protein 43, a growth cone-specific protein, is markedly down-regulated in the dorsal striatum after P7, which suggests that there is a decrease in axonal growth in this region after P7 ([@B32]; [@B23]). Therefore, IT-type corticostriatal innervation of the mouse brain mainly starts during the early stage of the first postnatal week, and these innervations become mature by the second postnatal week. Topography of the Corticostriatal Pathways {#s4} ========================================== Corticostriatal projections are topographically organized in such a manner as to convey neural information from various different cortical regions into the striatum where this diverse information is integrated. The dorsolateral/sensorimotor striatum, which is known to be involved in motor sequencing and habit-related functions, mainly receives its cortical inputs from the motor and somatosensory cortex. The dorsomedial/associative and ventral/limbic striata, which are related to goal-directed motor learning and reward, respectively, receive cortical inputs from the frontal cortex, mesocortex and allocortex ([@B5]; [@B100]; [@B13]; [@B98]; [@B71]; [@B76]). The deciphering of the topographic corticostriatal projection map has laid the foundations for understanding the functional networks related to a range of physiologic and pathologic conditions. Corticostriatal projections into striatal compartments {#s4A} ------------------------------------------------------ Considering the axonal terminal fields in the striatum, it has been shown that corticostriatal axons from different cortical regions innervate the striosomal and matrix compartments with different weights. The striosomal compartment receives substantial cortical inputs from evolutionarily conserved areas of the cortex, including the prelimbic, anterior cingulate, orbitofrontal, and insular cortices. The surrounding matrix compartment receives cortical inputs from the neocortex, including the motor, somatosensory and visual cortices ([@B57]; [@B39]; [@B129]; [@B50], [@B51]; [@B43]; [@B83]). When the laminar distribution of the corticostriatal neurons is explored, the early-born PT-type corticostriatal projection neurons in the rat brain have been found to be primarily located in lower Layer V (Layer Vb), from where they project their axons into the striosomes. By way of contrast, the late-born IT-type corticostriatal neurons are located in Layer III and upper Layer V (Layer Va), from where they predominantly project their axons into the matrix compartment ([@B58]; [@B100]; [@B30]; [@B83]; [@B93]). Interestingly, a recent genetic study using virus-based axonal tracing has revealed that there is no preference regarding corticostriatal afferents entering into the striatal compartments of the mouse brain ([@B142]). Cre-dependent monosynaptic tracing via genetically modified rabies virus using compartment-specific Cre driver mice by [@B142] has found that both the patch/exo-patch (striosome) compartment and the matrix compartment receive cortical innervations from the limbic and sensorimotor cortices without any preference. Cortical neurons in the upper and deep layers were shown to project into both the patch/exo-patch (striosome) compartment and the matrix compartment ([@B142]). Nonetheless, single-cell RNA sequencing has identified genetically distinct cell populations in a single cortical area ([@B150]). It is possible that different populations of cortical neurons within a single cortical region may project differentially into the striatal compartments. A combination of the single-cell RNA sequencing linked with axonal tracing and behavioral studies may help to clarify the above issue. Corticostriatal projections linking to distinct striatofugal pathways {#s4B} --------------------------------------------------------------------- Evidence based on retrograde axonal tracing and electron microscopy have shown that more corticostriatal axons of the IT-type compared to the PT-type are linked to neurons of the direct striatonigral pathway (dSPNs). In contrast, more PT-type than IT-type corticostriatal axons are linked to neurons of the indirect striatopallidal pathway (iSPNs; [@B92]; [@B132]; [@B36]). When the efferent regions of the cortex are explored, it remains unclear as to whether there is differential innervation of dSPNs and iSPNs from distinct cortical regions ([@B14]; [@B160]; [@B67]). [@B160] have reported that axons from the sensory and limbic cortices preferentially innervate dSPNs, whereas axons from the motor cortex preferentially project toward iSPNs ([@B160]). However, [@B67] did not find that there was differential innervation of dSPNs and iSPNs by the various different cortical regions. This inconsistency may be the result of variation in the degree of virus infection present in a single cortical region across the two studies. Alternatively, it also may be due to differences in trans-synaptic retrograde efficiency across the various different types of corticostriatal synapses ([@B67]). Corticostriatal Axonal Outgrowth and Synaptogenesis {#s5} =================================================== A coculture study has demonstrated the specificity of corticostriatal innervation by prenatal and postnatal cortical afferents. During cortical and striatal coculture, neurites derived from the prenatal cortex homogeneously grow into co-cultured striatal tissue, while, on the other hand, neurites derived from perinatal stages specifically grow into the striosomal compartment of co-cultured striatum irrespective of the age of the striatal tissue. Only a small amount of neurite innervations can be found in striatal tissue when it is co-cultured with postnatal cortex ([@B143]). These results suggest a time-dependent interplay between the different cortical afferents and the striatum. Because corticostriatal axonal innervation occurs before synaptogenesis and functional connectivity is established during the postnatal period (see below), corticostriatal axonal innervation during the various embryonic stages may occur without much in the way of activity modulation from the cortex. Corticostriatal synaptogenesis occurs postnatally after corticostriatal axonal innervation has been established. A cell culture study has shown that cortical synaptic inputs into the striatum are important for maturation of the dendritic arborization of the SPNs ([@B20]). SPNs receives glutamatergic excitatory inputs from VGluT1-positive corticostriatal neurons and VGluT2-positive thalamostriatal neurons ([@B53], [@B52]). Corticostriatal and thalamostriatal axonal terminals form axospinous and axodendritic synapses with the SPNs, respectively ([@B40]). The synaptogenesis of the excitatory synapses occurs soon after axonal innervations during postnatal periods. Interestingly, VGluT1 and VGluT2 immunoreactivity are highly enriched in striosomal loci during the first postnatal week, suggesting that corticostriatal and thalamostriatal synapse formation start to occur in the neonatal striosomal compartment ([@B108]; [@B87]). The dendritic spines of SPNs are found low in the P6 striatum, and mature types of dendritic spines are first able to be detected during the period P8--P9 ([@B90]). Dendritic spines markedly increase in number during the period P10--P12, at which time the SPNs are at their most excitable during the postnatal period. Dendritic spines continue to increase until the fourth postnatal week ([@B151]; [@B26]; [@B121]). The numbers of asymmetric synapses found in the P21 striatum at this time are comparable to those found in the adult rat striatum ([@B137]). Considering the synaptic pruning, there is a dramatic decrease in dendritic spines that occurs within the dorsolateral rat striatum, but not the dorsomedial rat striatum from P18 to P25 ([@B155]). It should also be noted that synaptic pruning by microglia has also been documented in the postnatal brain ([@B116]; [@B167]). During the early postnatal stages, developmental maturation of the electrophysiological properties of the SPNs involves immature SPNs that are characterized by a depolarized resting membrane potential together with hyperexcitability and a lack of inward rectification ([@B151]; [@B121]). When developmental maturation of corticostriatal innervations of SPNs occurs, this happens in parallel concomitant with developmental corticostriatal innervations and dendritic spinogenesis, while at the same time cortical stimulation-evoked glutamate-mediated slow wave activity is able to be induced in the SPNs of P3 brain slices as shown by a voltage-sensitive dye ([@B77]). By P5--P6, electrical-evoked and optical-evoked EPSCs are able to be detected in SPNs, suggesting the presence of functional corticostriatal synapses at this stage ([@B151]; [@B121]). Optical-evoked EPSCs progressively increase with the gradual recruitment and stabilization of AMPA receptors during the period P8--P18. Optical-evoked EPSCs reach a mature level by P30 ([@B121]). Interestingly, a decrease in the release probability of glutamate during the development of the corticostriatal synapses has been observed during striatal long-term depression (LTD) over the period P10--P23, which correlates with the motor function developmental maturation. Moreover, it has been reported that a loss of NMDA NR2C/D-mediated corticostriatal inputs occurs concurrently with a decrease in an immature pattern of striatal activity before P10 and that this is correlated with the onset of locomotion by the neonatal mouse pups ([@B34]). Cortical synaptic inputs not only drive synaptic activity in SPNs, but are also able to indirectly inhibit SPN activity via fast-spiking interneuron-mediated feed-forward inhibition ([@B125]). Electrophysiological recordings of fast-spiking interneurons in the striatum of rat brain slices obtained during the periods P12--P14 and P19--P23 have shown that fast-spiking interneurons have received frequent cortical inputs by the end of the second week after birth, which is when corticostriatal synaptogenesis is ongoing. This suggests a potential role for fast-spiking interneuron-mediated cortical feed-forward inhibition during the development of corticostriatal circuits ([@B125]). In the dorsolateral striatum, corticostriatal long-term potentiation (LTP) is inducible as early as P9--P10, whereas LTD cannot be induced until P15 ([@B27]; [@B119]; [@B106]). This developmental transition affecting the synaptic plasticity of LTD is controlled by upregulation of the endogenous cannabinoid ligand anandamide ([@B2]). The release probability of glutamate from cortical axonal terminals has been shown to decrease from P10 to P23, which may account for the induction of striatal LTD. LTD is known to be a physiologic mechanism that underlies motor performance, learning, and memory. Taken together, the above findings delineate the functional maturation process during postnatal development of corticostriatal synaptogenesis. Neuronal Activity-Dependent Regulation of Corticostriatal Synaptogenesis during Development {#s6} =========================================================================================== Glutamate inputs {#s6A} ---------------- A previous study has suggested that recurrent activity in the closed loops of cortico-basal ganglia circuits is able to regulate the synaptogenesis of SPNs ([@B85]). Synaptic connectivity and the strength of corticostriatal pathways are subject to regulation by the outputs from basal ganglia circuits. Chronic inhibition of activity within dSPNs and iSPNs during the second postnatal week by chemogenetic manipulation has been shown to result in decreased and increased spinogenesis and miniature EPSCs (mEPSCs) in dSPNs and iSPNs, respectively ([@B85]). Acute and chronic elevation of cortical activity by the inactivation of cortical interneurons or the optogenetic activation of corticostriatal axonal terminals has been shown to increase the synaptic connectivity of the corticostriatal pathways during early development ([@B121]). Moreover, correlated increases in cortical and striatal activity has been observed during the period P10--P16 ([@B121]). These findings suggest that corticostriatal inputs are capable of modulating activity-dependent synaptogenesis in the SPNs. Dopamine inputs {#s6B} --------------- Dopaminergic inputs into basal ganglia circuits have been suggested to regulate corticostriatal maturation. Mesostriatal dopaminergic afferents start to innervate the striatum during early embryonic stages. At the perinatal stage, mesostriatal dopaminergic axonal terminals form "dopamine islands" that correspond to the developing striosomes ([@B114]; [@B62]; [@B156]). During the period P8--P13, activation of Gα~s~-coupled G-protein receptors in SPNs by a D1 agonist is able to increase corticostriatal activity and the number of dendritic spines ([@B86]). Furthermore, depletion of dopamine input into the P2 striatum weakens not only SPN activity in response to cortical stimulation, but also impairs the corticostriatal synchronization that accompanies locomotion induced hyperactivity during the period P21--P25 ([@B55]). Moreover, the imbalances in dopamine D1 and D2 signaling in SPNs that occur before the first two postnatal weeks has been shown to lead to significant alterations in corticostriatal innervation and spinogenesis ([@B85]). A recent study has further shown that nigrostriatal dopamine release during a critical period between P18 and P28 is required to reduce the intrinsic hyperexcitability of neonatal dSPNs to the level found in adults ([@B96]). Taken together, the above compelling evidence highlights the importance of dopamine transmission, not only in the regulation of the postnatal maturation of SPNs, but also it associated with the pathogenic mechanisms related to neurodevelopmental disorders. Brain-derived neurotrophic factor (BDNF) {#s6C} ---------------------------------------- BDNF is one of the most studied neurotrophins and has been shown to be involved in the pathogenesis of various neurodevelopmental and neuropsychiatric disorders ([@B10]; [@B117]). BDNF binds to tyrosine receptor kinase B (TrkB) to transduce signals within developing neurons. Evidence from BDNF and TrkB conditional knock-out mice has indicated that BDNF-TrkB signaling regulates the neuronal survival, morphogenesis and synaptogenesis of striatal neurons ([@B11]; [@B95]); furthermore, presynaptic BDNF secretion is required for activity-dependent corticostriatal LTP ([@B118]). Notably, TrkB expression is enriched in the striosomal compartment during the first postnatal week when corticostriatal axons are selectively innervating striosomal cells ([@B29]). These findings raise an intriguing possibility that BDNF-TrkB signaling may be involved in setting up within the striatal compartments the temporal order during activity-dependent corticostriatal synaptogenesis; this could involve initiating the initial steps of synapse formation within the striosomal compartment. Neuropsychiatric Diseases Related to Dysfunction of the Corticostriatal Circuits {#s7} ================================================================================ Attention-deficit hyperactivity disorder (ADHD) {#s7A} ----------------------------------------------- ADHD is a neurodevelopmental disease that is characterized by symptoms including inattention and/or hyperactive/impulsive behaviors that persist for more than six months ([@B9]). Based on neuroimaging and genetic studies, abnormal neural connectivity and abnormal neurologic functioning are believed to underlie the pathology of ADHD brains ([@B104]). Aberrant neural circuits, including dorsal frontostriatal, orbitofrontostriatal, prefrontostriatal, and frontoparietal circuits, have been identified as being associated with ADHD ([@B42]; [@B74]; [@B104]). Many ADHD studies have centred on dysfunction of the dopaminergic system ([@B94]). Dopamine neurotransmission is essential for motor control, reward learning, and motivation ([@B15]; [@B59]). In addition to regulating neurotransmission within the adult brain, dopamine also plays an important role in neural development. Notably, depletion of dopamine by intraventricular injection of 6-hydroxydopamine (6-OHDA) into neonate rodent brains has been shown to induce an ADHD-like behavioral phenotype; furthermore, abnormal development of corticostriatal pathways and synaptogenesis has been found in 6-OHDA-treated brains ([@B55]; [@B19]). Moreover, altered frontostriatal functional connectivity, which has linked to a reduction in the dendritic arborizations of SPNs, in particular iSPNs, has been also observed in juvenile 6-OHDA-treated mice. These neonatal 6-OHDA-induced pathologic changes become more severe in adult mice compared to juvenile mice ([@B19]). Transgenic knock-in mice expressing a human DRD4 polymorphic variant associated with ADHD have been shown to exhibit a decrease in corticostriatal glutamate release ([@B17]). These different lines of evidence suggest that the corticostriatal development controlled by dopaminergic transmission may underlie ADHD pathophysiology. Autism spectrum disorder (ASD) {#s7B} ------------------------------ ASD is a highly heterogeneous disease. The core symptoms of ASD include impaired social communication functioning and self-interest related repetitive behaviors ([@B9]; [@B165]). Aberrant synaptogenesis, plasticity and excitatory/inhibitory balance are believed to be involved in ASD pathophysiology ([@B154]; [@B18]; [@B110]). The complex yet specific symptoms of ASD pathogenesis are presumably caused by aberrant wiring of specific neural circuits in the ASD brain. Evidence suggests that dysfunction of the basal ganglia circuits plays an important role in ASD pathogenesis and is related to repetitive behavior and defective social communication ([@B126]; [@B44]; [@B54]). Clinical studies have reported that there is an increase in the volume of the striatum that is positively correlated with the repetitive behaviors, the social deficits and the communicational deficits of the patients ([@B73]; [@B133]). Neuroimaging studies have shown a correlation between inward surface deformation affecting distinct striatal regions and impaired motor skills, praxis and poorer social communication ([@B127]). Moreover, reduced long-range functional connectivity between the right inferior frontal cortex and the right caudate has been observed in brain of children suffering from ASD ([@B91]). These findings imply that there are dysfunctions that affect the corticostriatal and striatofugal circuits of ASD brains. Many genes that have been associated with ASD have been found to be related to corticostriatal development and synaptogenesis, including the *Shank* gene family and the *Neuroligin* (NLGN) gene family. *Shank1*, *Shank2*, and *Shank3* are members of a postsynaptic scaffolding SH3 and multiple ankyrin repeat domains protein family. Mutations of the *Shank* family proteins are associated with autism and a number of other neuropsychiatric disorders ([@B124]; [@B89]; [@B103]). Mutation of the ASD-risk gene *Shank3B,* which is the most well-studied isoform, has been found to be enriched in corticostriatal regions, and to be associated with an increase in dendritic arborizations and a decrease in dendritic spinogenesis and synaptogenesis of SPNs ([@B120]). Mutation of *Shank3B* also leads to precocious hyperactivity of corticostriatal inputs due to the presence of cortical hyperactivity during P14; while, in contrast, lower levels of corticostriatal connectivity have been found in adult brains with this mutation ([@B120]; [@B121]). Mutations affecting the NLGN gene family induce corticostriatal synaptopathy and this seems to be related to the ASD pathogenesis ([@B147]). For example, *NLGN3* mutant mice exhibit increased inhibitory synaptic transmission in dSPNs of the ventral striatum and impaired corticostriatal LTD in the dorsal striatum ([@B135]; [@B99]). *NLGN1,* another NLGN member that has recently been identified as an ASD-risk gene ([@B109]), has been shown to regulate NMDA-mediated synaptic transmission in dSPNs and mEPSC frequency in iSPNs ([@B16]; [@B46]). In addition to genes in the Shank and NLGN families, other ASD risk genes also seem to alter corticostriatal connectivity and activity. Human transcriptome analysis has identified Teashirt zinc-finger homeobox family member 3 (*TSHZ3*) as a hub gene that is involved in the development of cortical projection neurons. Patients with deletion mutation of the *TSHZ3* gene exhibit ASD symptoms, and an animal model study has shown that there is an increase in corticostriatal LTP in *Tshz3* heterozygous mice ([@B24]). Furthermore, *Fmr1* encodes the fragile X mental retardation protein that is involved in autistic Fragile X syndrome. In *Fmr1* knock-out mice, an increase in inhibitory neurotransmission and defects in LTD have been found in the striatum and this has been linked with hypoconnectivity of the corticostriatal pathways ([@B25]; [@B79]; [@B169]). Eukaryotic translation initiation factor 4E (*eIF4E*) is another gene associated with ASD susceptibility ([@B111]). Overexpression of *eIF4E* in transgenic mice results in increased dendritic spines in Layers II and III of the medial prefrontal cortex as well as enhanced striatal LTD ([@B136]). Mutation of transcription factor forkhead box p1 (*Foxp1*), an ASD-risk gene, is known to increase the neuronal excitability of iSPNs during P18 ([@B7]). Finally, a reduction in putative cortico-striosomal synaptogenesis has been found in the P8 striatum of a ASD mouse model involving maternal treatment of mice with valproic acid ([@B87]). Collectively, the above animal model studies suggest that dysfunction of corticostriatal synaptic homeostasis and disruption of synaptogenesis during development may contribute to the pathologic mechanisms of ASD. Obsessive-compulsive disorder (OCD) {#s7C} ----------------------------------- OCD is a neurodevelopmental disease wherein dysfunction of corticostriatal circuits at the neural circuit level is implicated in the pathology of the disease. Neuroimaging studies have found abnormalities of corticostriatal pathways, including the orbitofrontal cortex, prefrontal cortex, anterior cingulate cortex and striatum in the brains of patients with OCD ([@B69]; [@B134]; [@B6]; [@B75]; [@B80]). The pathologic alterations in these corticostriatal circuits have been correlated with the severity of the patient's OCD symptoms ([@B70]). Moreover, repetitive transcranial magnetic stimulation and deep-brain stimulation-induced changes in corticostriatal activity seem to be able to cause alterations in the symptoms in OCD patients ([@B48]; [@B41]), which supports a causal relationship between dysfunction of corticostriatal circuits and OCD pathogenesis ([@B22]; [@B47]). Animal model studies have shown that chronically optogenetic activation of the orbitofronto-ventral striatum pathways is sufficient to bring about a progressive increase in the obsessive grooming behavior of mice, and that fluoxetine, a clinical drug used to treat OCD, is able to alleviate the optogenetic-induced obsessive grooming behaviors ([@B3]). At the genetic level, mutations of several genes have been associated with OCD. It is notable that corticostriatal circuitry appears to be a convergent pathologic locus that is targeted by several OCD-risk genes, these include synapse-associated protein 90/postsynaptic density protein 95-associated protein 3 (*SAPAP3*) and members of the *Slitrk* gene family. SAPAP3 is an excitatory postsynaptic scaffold protein. Mutations of *SAPAP3* have been found in patients with OCD ([@B171]). *Sapap3* knock-out mice exhibit defects affecting the structure of the postsynaptic complex and show a OCD-like behavioral phenotype. Moreover, the synaptic activity of the corticostriatal pathways, but not the thalamostriatal pathways, are altered, which indicates that the corticostriatal pathways seem to be specifically relevant to OCD pathophysiology ([@B163]; [@B161]). Compulsive grooming, an abnormality associated with *Sapap3* knock-out mice, is able to be rescued by optogenetic activation of parvalbumin-positive fast-spiking interneuron-mediated feed-forward inhibition of the orbitofrontal-striatal circuits ([@B21]). The *Slitrk* gene family is known to regulate synaptogenesis ([@B148]; [@B168]). Mutations of *Slitrk* gene family members are associated with various neuropsychiatric diseases, including Tourette syndrome and OCD ([@B1]; [@B115]). Slitrk5 is known to regulate the recruitment of TrkB into the postsynaptic regions to bring about BDNF-mediated neurite outgrowth of striatal neurons ([@B145]). Genetic deletion of *Slitrk5* not only results in neuronal hyperactivity in the orbitofrontal cortex, but also causes a reduction in corticostriatal transmission and dendritic complexity that is associated with OCD-like behaviors ([@B140]). In addition to mutation of the *SAPA3* and *Slitrk5* genes, mutations of the *SLC1A1* and *Hoxb8* genes have been shown to be related to abnormalities in corticostriatal activity and to changes in corticostriatal connectivity that have in turn been associated with OCD phenotypes. The *SLC1A1* gene encodes neuronal glutamate/aspartate/cysteine transporter excitatory amino acid transporter 3 (EAAT3). An abnormal increase in *SLC1A1* expression with a parallel decrease in EAAT3 activity, have been found in the brains of patients with OCD ([@B157]; [@B38]; [@B164]). EAAT3-deficient mice exhibit diminished basal ganglia-dependent stereotypic behavior ([@B170]). A recent study has further reported that NMDA receptor subunit composition and NMDA-dependent synaptic plasticity are altered in mice overexpressing EAAT3 and this is linked to OCD-like behavioral deficits; these deficits can be rescued by antipsychotic treatment ([@B35]). Interestingly, Hoxb8 function in microglia has been suggested to be involved in the pathogenesis of OCD via regulation of the corticostriatal circuits. *Hoxb8* gene knock-out has been shown to induce an expansion in cortical synapses and contraction in striatal synapses. These changes then lead to an enhancement of corticostriatal activity in the *Hoxb8* mutant mice, which can be seen to exhibit compulsive grooming behavior ([@B64]; [@B107]). Schizophrenia {#s7D} ------------- Schizophrenia is a neurodevelopmental disease. Imbalanced excitatory/inhibitory transmission and abnormal synaptic function have been implicated in the pathophysiology of schizophrenia. Genetic linkage studies have indicated that the *AKT1* and *PRODH* genes are associated with schizophrenia ([@B4]; [@B82]; [@B152]). Elevated frontostriatal connectivity, particular in the dorsolateral prefrontal cortex, is present in subjects carrying AKT1 and PRODH variant alleles ([@B102]; [@B82]; [@B149]). Corticostriatal dysfunction affecting cognitive learning and reward processing have been identified among patients with schizophrenia ([@B159]; [@B33]). In animal studies, NRG1-ErbB4 signaling ([@B78]) has been proposed to underlie the pathogenesis of schizophrenia. Conditional deletion of ErbB4 in Dlx5/6 cell lineage cells, including SPNs and cortical interneurons, results in enhanced inhibitory synaptic transmission in the striatum ([@B56]). Loss of zinc-finger SWIM domain-containing protein 6 (*ZSWIM6*), another schizophrenia-risk gene, has been shown to decrease neurite arborization and the number of dendritic spines in SPNs ([@B153]). Abnormal dopamine transmission has been a prevailing theory for the etiology of schizophrenia ([@B141]). In this context, it is interesting to note that mesostriatal dopamine input during development is able to modulate corticostriatal innervations, spinogenesis and activity during development ([@B55]; [@B85]; [@B96]). These findings suggest that abnormal synaptic wiring within the corticostriatal circuits may be a potential pathologic mechanism that underlies abnormal dopamine neurotransmission in schizophrenia brains. Speech and language disorders {#s7E} ----------------------------- Speech and language are unique and fundamental to human beings and their social communication. The prevalence rate for language delay during the development of school children is high, which calls for a better understanding of the neurobiology of speech and language and the development of new therapeutic approaches to such problems. Cortico-basal ganglia circuits are important to speech and language ([@B162]; [@B37]; [@B61]; [@B84]). Several genes related to speech and language have been identified using genome-wide linkage/association sequencing studies ([@B112]; [@B81]; [@B61]; [@B84]). A well-studied speech and language-related gene is *FOXP2*. The *FOXP2* R553H missense mutation caused severe level of speech and language disorder in the KE family ([@B88]). Neuroimaging studies have found structural abnormalities as well as functional abnormalities that affect the cerebral cortex and striatum of patients with a *FOXP2* mutation. This suggests there is involvement of cortico-basal ganglia circuits in the pathology of this disorder ([@B162]; [@B97]). Consistent with the above, animal studies have indicated that the cortico-basal ganglia circuits are critical to vocal communication in songbirds and rodents ([@B45]; [@B49]; [@B8]; [@B28]; [@B84]). Increasing evidence indicates that *Foxp2* plays a crucial role in the neural development and plasticity of corticostriatal circuits. Mutation of the *Foxp2* gene decreases synaptic transmission and plasticity in corticostriatal synapses ([@B65]; [@B45]; [@B130]). Humanized *Foxp2* gene has been shown to increase synapse formation and synaptic functioning in the SPNs from mouse brains ([@B45]; [@B26]). It is important to note that developmental deficits in spoken language function are a comorbidity associated with various other psychiatric disorders, including ASD in which vocal communication can also be severely affected. Interestingly, Foxp2 has been shown to negatively regulate several ASD-risk genes, including *Cntnap2*, *Met*, and *Mef2c* ([@B158]; [@B105]; [@B26]). Our recent study has uncovered a molecular mechanism by which Foxp2 promotes the synaptogenesis of corticostriatal circuits via suppression of Mef2c, a negative regulator of synapse formation. The Foxp2-Mef2c signaling-mediated synaptic wiring of corticostriatal circuits has a causative role in vocal communication ([@B26]). Cntnap2 may modulate corticostriatal inputs via regulation of the development of striatal GABAergic interneurons ([@B122]). Finally, *Met* knock-out mice have been shown to exhibit hyperactivity of corticostriatal pyramidal neurons in cortical Layer Vb, which may affect the development of corticostriatal pathways ([@B128]). Cortico-basal ganglia circuits engage with other neural circuits to control complex neurobiological functions such as speech and language. The corticostriatal circuits may be a good entry point by which one may explore the developmental basis of vocal communication and this should be able to provide insights into the pathogenesis of various speech and language disorders. Tourette syndrome {#s7F} ----------------- Tourette syndrome is a childhood-onset neurodevelopmental disease that is characterized by involuntary vocal and motor tics. Tourette syndrome is categorized as one of the three tic disorders in DSM-V ([@B9]). Tourette syndrome is a heterogeneous disorder that shows comorbidity with other neuropsychiatric diseases, including OCD and ADHD ([@B72]). The pathophysiology of Tourette syndrome is poorly understood. At the circuit level, increasing evidence had implicated pathologic changes in cortico-striato-pallido-thalamic circuits within the brains of patients with Tourette syndrome ([@B101]). Neuroimaging studies have found abnormal increases in structures associated with corticostriatal connectivity, and these are further correlated with the severity of the tics ([@B60]; [@B166]), At the genetic level, twins siblings studies of individuals with Tourette syndrome have indicated that the concordance rate of monozygotic twins is higher than that of dizygotic twins ([@B101]). Furthermore, genetic studies have identified a number of candidate genes that are associated with Tourette syndrome. One possible candidate gene is the Slit and Trk-like 1 (*SLITRK1*) gene. Patients carrying *SLITRK1* variant alleles have known to exhibit Tourette syndrome ([@B1]). Slitrk1 is expressed at high levels in cortical Layers III, V, and VI, which is where the corticostriatal and corticothalamic projection neurons are located in the mouse brain. Slitrk1 is also highly expressed in postnatal striosomes. Expression of slitrk1 has been shown to be significantly down-regulated after the second postnatal week, which is when corticostriatal innervation and synaptogenesis occur. The expression pattern of *SLITRK1* in the cerebral cortex and striatum of the human brain is similar to that found in the monkey brain and the mouse brain ([@B146]). Previous studies have shown that Slitrk1 positively regulates dendrites outgrowth and synaptogenesis in cortical pyramidal neurons and hippocampal neurons ([@B1]; [@B168]; [@B12]). Given that Slitrk1 is expressed during the development of corticostriatal pathways, a potential role for Slitrk1 in the regulation of corticostriatal connectivity and its possible involvement in Tourette syndrome warrants further investigation. Conclusion {#s8} ========== A broad repertoire of genetic components participates in the synaptic wiring of the corticostriatal circuits during development. Activity-dependent machinery is also adopted to fine tune synaptogenesis to develop precise functionality within the cortico-basal ganglia network. Characterization of the development and functioning of corticostriatal circuits may not only help us to understand the developmental basis of motor control, skill development and habit learning as well as complex cognitive functions such as speech and language, but may also provide insights into the pathophysiology of corticostriatal circuits-related neurologic and psychiatric disorders. This, in turn, may lead to the identification of potential therapeutic approaches to the treatment of these important diseases. Synthesis {#s9} ========= Reviewing Editor: Juan Burrone, King\'s College London Decisions are customarily a result of the Reviewing Editor and the peer reviewers coming together and discussing their recommendations until a consensus is reached. When revisions are invited, a fact-based synthesis statement explaining their decision and outlining what is needed to prepare a revision will be listed below. The following reviewer(s) agreed to reveal their identity: NONE. This review provides a comprehensive and detailed description of the wiring of cortico-striatal circuits in health and disease. The review was well organised and well discussed. It is a welcome summary of the large body of work that exists in this area. Overall, I find that this is an important and timely review. However, there were some minor issues that the authors should address. For example, the review contained multiple sentences that were either not clear or grammatically incorrect. The authors should take time to revisit the review in detail and iron out these issues. Below are specific comments on the review. The introduction has many sentences that need rewriting. Below are some examples, but the authors should carefully read through this section and modify so that it reads more clearly. Delete 'the' form first line of introduction to read: 'are critical to network function'. Second sentence (starting with 'The cortico-basal ganglia\...') is not clear. Fourth sentence of introduction (starting with 'The cortico-striatal afferents\...') is also not clear. 13th line of introduction: insert 'be' so that sentence reads '\...to be involved in a broad\...' Second paragraph of introduction, first sentence: sentence does not read well. Second paragraph of introduction, second sentence: change to 'at the cellular and molecular level' Second paragraph of introduction, third sentence does not read well. Page 4, 5th line: insert 'The' to start sentence 'The majority\...' Page 4, 2nd paragraph, first sentence, change to: '\...Sohur et al (2014) performed retrograde and anterograde labelling of\...' Page 4, 2nd paragraph, last sentence: what do you mean by establish the innervations? Page 6 second paragraph: sentence starting 'Previous studies\...' is confusing. Page 7, line 3: start sentence with 'A previous cell culture\...' Page 7: what does Asymmetric synaptogenesis mean? Is this the synaptogenesis of asymmetric synapses? If so, please define asymmetric synapses or refer to them as excitatory synapses. Page 8, second paragraph, line 2: switch to 'interneuron-mediated'. Page 8, second paragraph, line 5: change to 'by the end of the second week'. Page 8, second paragraph, line 6: change 'are undergoing' with ' is ongoing'. Page 8, last paragraph, line 1: start with 'Previous studies' Page 9, first paragraph: sentence starting with 'Moreover, correlated increases in cortical\...' is unclear. Page 10: In ADHD section, what is meant by 'physiologic function'? Page 11, last paragraph: This sentence does not make much sense, as the genes listed have not been mentioned before. Page 12: this whole section on genes related to ASD may benefit from a common thread that links genes together. As it stands it reads like a long list of genes involved in ASD but it is not always clear why the review jumps for one gene to another. Pages 13-14: a similar comment applies to genes involved in OCD. [^1]: The authors declare no competing financial interests. [^2]: Author contributions: H.-Y.K. and F.-C.L. wrote the paper. [^3]: This work was supported by Ministry of Science and Technology-Taiwan Grants MOST107-2321-B-010-002 and MOST107-2320-B-010-041-MY3, the Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education in Taiwan (F.-C.L.), and Postdoctoral Fellowship Grants MOST107-2811-B-010-011 and MOST107-2321-B-010-010-MY3 (to H.-Y.K.).

The authors confirm that all data underlying the findings are fully available without restriction. All relevant data are within the paper and its Supporting Information files. Introduction {#s1} ============ The stoichiometric proportions of elements in plant and animal tissues differ substantially, which causes all herbivores to be confronted with a high "stoichiometric mismatch" [@pone.0115104-Denno1], [@pone.0115104-Hessen1]: the proportions of C:x (C indicates the content of carbon, x indicates the content of another element) in the consumer\'s body are much lower than in its food. To build up and maintain its body composition, a herbivore must develop a strategy enabling the selective consumption or assimilation of particular elements. The strategies for how to cross such a stoichiometric threshold remain poorly understood. Invertebrates feeding on dead wood comprise an extreme case. C, H and O comprise 99% of wood mass. How can an organism build its body based on a substrate consisting almost exclusively of these elements? These compositional differences cause an extreme stoichiometric mismatch between the most common organic tissue on Earth and all the organisms exploiting this resource. The problem of the stoichiometric mismatch between plant biomass and its potential non-microbial consumers in terrestrial ecosystems has not been fully investigated. Fagan et al. [@pone.0115104-Fagan1] observed the discrepancy in nitrogen and phosphorus content between living autotrophs (plant foliage and algae) and typical herbivores; the C:N and C:P ratios are 10 to 20 times higher in autotrophs than in herbivores. Schneider et al. [@pone.0115104-Schneider1] demonstrated a high stoichiometric mismatch in the P content between cave-dwelling invertebrates and the plant detritus they consume. Wood-boring (xylophagous) insects living in the xylem of dead trees constitute another prominent example but have not previously been studied. As estimated from the available data on the elemental content in wood [@pone.0115104-Meerts1]--[@pone.0115104-Palviainen2] and insects [@pone.0115104-Fagan1], [@pone.0115104-Schneider1], [@pone.0115104-Elser1], [@pone.0115104-Rumpold1], the stoichiometric discrepancy may reach two orders of magnitude for living wood and three orders of magnitude for dead wood. This discrepancy concerns not only nitrogen and phosphorus but also K and Mg. Many of the studies on the nutritional mismatch between terrestrial herbivores and their food have focused exclusively on two major nutrients: N and P; however, other elements are also essential for the growth and survival of consumers, and a low level of these elements in food plants may cause a nutritional imbalance [@pone.0115104-Rumpold1]--[@pone.0115104-Cohen1]. No data are available for the content of minor essential nutrients in dead wood; however, it may be assumed that their concentrations are also negligible. To date, only a few papers have tackled the stoichiometry of microelements, including studies of freshwater or marine pelagic systems [@pone.0115104-Karimi1]--[@pone.0115104-Twining1], of terrestrial saprobiotic microorganisms [@pone.0115104-Schneider2], and of prairie grasshoppers [@pone.0115104-Joern1]. However, the majority of those studies concerned the consumption of living plant tissues, which are relatively nutrient-rich [@pone.0115104-Fagan1], [@pone.0115104-Joern1], [@pone.0115104-Sterner1]. Bark beetles, feeding on living phloem, are the most unbalanced herbivores studied to date [@pone.0115104-Hessen1], [@pone.0115104-Elser1], [@pone.0115104-Sterner1], [@pone.0115104-Ayres1]. Sterner and Elser [@pone.0115104-Sterner1] suggested that termites may possibly be the most unbalanced consumers, with a discrepancy of two orders of magnitude between the nutrient content of their bodies and the nutrient content of their food. Stoichiometric mismatch is not the only problem faced by xylophagous organisms. The poor digestibility of cellulose, hemicelluloses and lignin limits the energy budgets of wood-eating invertebrates and slows their growth [@pone.0115104-Haack1]--[@pone.0115104-Walczyska2]. The life strategy of xylophages includes an extremely long development time, which is possible due to the relatively low mortality risk of larvae living deep in the xylem and may compensate for the low digestibility of food [@pone.0115104-Walczyska2], [@pone.0115104-Walczyska3]. However, it remains an open question whether the fundamental effect on the life history of a xylophage is caused by carbon (energy source) or other nutrients (assembling molecules of crucial functional importance). Although the symbiotic interactions of numerous xylophages with microorganisms may ease the cellulose digestibility constraint [@pone.0115104-Douglas1], the supply of nutrients in wood cannot be increased this way; their deficit can only be ameliorated by an external supply. However, despite the stoichiometric mismatch, wood-boring beetle larvae are capable of extracting from their food all of the elements necessary for growth and for controlling their metabolic processes. Therefore, the wood consumed must be supplemented with some nutrient-rich material. The obvious candidates are saprobiotic fungi that invade dead wood, which are capable of transferring large quantities of elements to the food source (see [@pone.0115104-Mooshammer1] for a review). The aim of this study was to determine how xylophagous insects can manage the drastic stoichiometric imbalance of major and minor nutrients. We tested the hypothesis that rather than offsetting nutritional constraints with a prolonged development period, wood-boring larvae may balance their nutritional demands by the import of nutritional elements from outside the system by the action of fungi. Thus, avoiding the stoichiometric mismatch may differentially shape the life histories of dimorphic sexes and various species exploiting the same resources. To address this hypothesis, we compared the elemental contents in the bodies of three species of wood-boring beetles inhabiting the same pine stumps, differing in body size and life histories, with the elemental content of wood (potential food during larval development) at various stages of decay. We examined the levels of essential macro- and micronutrients (C, N, P, K, Ca, Mg, Fe, Zn, Mn, Cu, and Na). In decaying wood, in addition to elemental concentrations, we also estimated the amount of fungal tissue using ergosterol content as a proxy [@pone.0115104-Klamer1]. Materials and Methods {#s2} ===================== Three common species of pine-xylem-feeding beetles were used: *Stictoleptura rubra* Linnaeus 1758 ( =  *Corymbia rubra* Nakano and Obayashi 1957;  =  *Aredolpona rubra* Viliers 1974), *Arhopalus rusticus* Linnaeus 1758 ( =  *Criocephalus rusticus* Haldeman 1847; Coleoptera, Cerambycidae) and *Chalcophora mariana* Linnaeus 1758 ( =  *Buprestis mariana* Linnaeus 1758; Coleoptera, Buprestidae). The development times for these species reported in the literature [@pone.0115104-Dominik1] are 3 years in the smallest beetle, *S. rubra*; 2--4 years in *A. rusticus*, which is of intermediate size; and 5--6 years in the largest of these beetles, *Ch. mariana*. Pine stumps potentially inhabited by larvae of these beetles were collected in the Niepołomice Forest, approximately 20 km east of Cracow (southern Poland, 50°05′N, 20°21′E, elevation 184--212 m a.s.l.) in spring, summer and autumn during 2010--2012. The stumps were collected from approximately 80-year-old pine stands, one to four years after felling. The stumps (diameter measured at the top was approximately 40 to 90 cm; height measured at the center was approximately 10 to 50 cm) were cut slightly below the ground and were hand-split to collect wood samples, pupae and larvae. The adult beetles leaving the stored stumps were also collected. In addition, *S. rubra* adults were captured in the forest. The field studies did not involve endangered or protected species. No specific permissions were required for collecting beetles and stumps in this location. The wood samples were classified by degree of decay (after [@pone.0115104-Esseen1]): (1) undecayed wood -- hard and healthy, without visible changes caused by microorganisms; (2) moderately decayed wood -- considerably changed by microorganisms, colored (purple or dark brown), wet and softer than (1) but still difficult to tear apart with a knife; (3) highly decayed wood -- many visible changes, ample layers of white or brown rotting fungi, wet and soft, easily torn apart by knife or even by hand; and (4) corridor wood -- a thin layer of wood from the walls of corridors made by xylophagous larvae, together with content ('frass', i.e., wood fragments and feces). Prior to elemental content determination, the insects and wood samples were freeze-dried (using Christ Beta 1--8 LD plus). All the samples were dried using a two-stage pressure/temperature regime: main drying at 0.34 mbar/−31°C and final drying at 0.0010 mbar/−76°C. We dried insects for five days and wood samples for seven days. For CHNS analyzer, we dried, ground and homogenized the material (one insect or wood sample taken from one stump per measurement sample), and for the mineralization (samples for AAS and colorimeter) we dried intact insects and ground and homogenized wood (wood sample taken from one stump per measurement sample). C and N levels were determined using a Vario EL III automatic CHNS analyzer. K, Ca, Mg, Fe, Zn, Mn, Cu and Na levels were determined by atomic absorption spectrometry (Perkin-Elmer AAnalyst 200 and Perkin-Elmer AAnalyst 800), and P content was determined colorimetrically (MLE FIA flow injection analyzer). Prior to analysis, the samples were mineralized by acid digestion: beetles in HNO~3~, pupae in a solution of HNO~3~, HClO~4~ and H~2~SO~4~, and wood samples in a solution of HNO~3~ and HClO~4~. Insect samples consisted of one or a few individuals (as described above), and pupae, adult males and adult females were considered separately. Sulfanilic acid was used as the reference material for C and N analyses, and Certified Reference Materials (bush -- NCS DC 733348, chicken -- NCS ZC73016 and pork muscle -- NCS ZC 81001) were used for the other elements. Sample sizes differed for particular elemental analyses of beetles and variously decayed wood because of the uneven abundance of particular species, sexes and developmental stages and the necessity to pool small specimens and match the separate requirements for every individual element measured (for details, see [S1 Table](#pone.0115104.s001){ref-type="supplementary-material"}). Ergosterol content was measured in wood samples using a GC/MS Clarus 600 chromatograph (Perkin-Elmer). We used 73 wood samples collected from the same site as the samples for the elemental content analysis, but from different stumps. The degree of decay of the stumps was categorized in the same way as for elemental content analysis, but the 4th category (corridors) was omitted. We express the degree of stoichiometric mismatch between the insects and their food for element x most simply as the ratio of the stoichiometric ratios in food and in the consumer\'s body (Trophic Stoichiometric Ratio  =  *TSR*):where *C* -- carbon content and *X* -- content of element *x*. This index does not depend on the units used for stoichiometric ratios C:x (molar or mass units). Values of *TSR~x~\>1* indicate stoichiometric mismatch, with severe mismatch indicated by a *TSR* value substantially different from unity. Because the variation of the *TSR* values cannot be directly measured, we used bootstrapping to estimate the mean *TSRs* for specific elements and food categories with confidence limits. First, the *TSR* values were computed from raw data on a given elemental content in the food and consumer. The values were drawn 15 times, and a mean was calculated. The procedure was repeated 500 times, giving 500 means (bootstrap samples), from which the final averages with confidence intervals were calculated. Principal component analysis (PCA) was employed to compare the multi-elemental stoichiometric relations among species, sexes and developmental stages. The data were log-transformed, centered and standardized by species but not by sample; thus, PCA was performed on a correlation matrix. To check for differences between the indicated clusters, we computed ANOVA independently for the 1st and 2nd axis scores. The Mann-Whitney U test and Kruskal-Wallis test were used for significance testing (p\<0.05) of the differences between species and sexes in the elemental composition and body size (dry mass). Statistica 10 was used for all statistical analyses except for PCA (Canoco 4.5). Results {#s3} ======= Body composition of beetles {#s3a} --------------------------- The adult beetles differed in body dry mass between species and showed sexual dimorphism of body size, with females being larger than males ([Fig. 1](#pone-0115104-g001){ref-type="fig"}; for details, see [S2 Table](#pone.0115104.s002){ref-type="supplementary-material"}), but the difference in body size between sexes was significant only for *S. rubra* (Mann-Whitney U test, p\<0.05). {#pone-0115104-g001} The sexes of the pupae were not determined; their mass was intermediate between the adult masses of the two sexes or larger than the masses of either ([Fig. 1](#pone-0115104-g001){ref-type="fig"}; for details, see [S2 Table](#pone.0115104.s002){ref-type="supplementary-material"}). The complete data on elemental content are presented in [S1 Table](#pone.0115104.s001){ref-type="supplementary-material"}. The carbon mean content in imagines ranged from 51.7 to 63.3% d.m. and did not differ significantly between species or sexes, except for *A. rusticus* females, which had a significantly higher carbon content than males and females of other species. *Ch. mariana* pupae had a significantly lower C content than *S. rubra* pupae ([Table 1](#pone-0115104-t001){ref-type="table"}). 10.1371/journal.pone.0115104.t001 ###### Average concentrations of elements in the three species of xylophagous beetles and in samples of variously decayed wood from pine stumps inhabited by larvae (see text for definitions of wood decay categories). {#pone-0115104-t001-1} C N P K Na Ca Mg Fe Zn Mn Cu ------------ ----------------------- ---------------- ---------------- -------------- -------------- ---------------- -------- ---------- -------------- -------------- ------- -------------- ------ Xylophages \% d.m. **mg/kg d.m.** *S. rubra* mean 52.9 10.3 0.6 6866.8 711.6 770.3 1428.8 102.9 112.8 23.8 37.2 SD 3 1.3 0.1 931.8 259.7 433.2 266.8 41.6 18.6 5.4 18.6 sex diff. n.s. M\>F n.s. n.s. n.s. M\<F M\<F n.s. n.s. n.s. n.s. mean 59.2 7.8 0.6 6831 794.1 1421.4 1615.9 128.3 175.9 18.8 41.6 *A. rusticus* SD 5.1 2.2 0.1 1081.8 207.8 1124.3 284.3 69.2 67.3 6.5 10.9 sex diff. M\<F M\>F M\>F n.s. n.s. M\>F n.s. M\>F n.s. n.s. n.s. mean 54.6 7.9 0.6 7837.1 818.1 808 2202.6 42 78 25.3 6.8 *Ch. mariana* SD 1.8 1 0.1 890.1 499.4 136.6 441.9 10.8 12.1 6.3 3 sex diff. n.s. M\>F n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. Pooled mean 54.1 9.4 0.6 7137.9 752.2 879.4 1684.3 80 109.6 23.6 28.7 SD 3.7 1.8 0.1 1027.7 339 556.7 476.3 51.8 44.4 6.2 20.5 males species diff. n.s. Sr\>Cm n.s. n.s. Sr\<Cm Sr\<Ar Sr\<Cm Sr = Ar\>Cm Sr = Ar\>Cm n.s. Sr = Ar\>Cm females species diff. Sr = Cm\<Ar Sr = Cm\>Ar n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. pupae species diff. Sr\>Cm Sr\>Cm Sr\<Cm Sr\<Cm n.s. n.s. Sr\<Cm Sr\>Cm Sr\>Cm n.s. Sr\>Cm \% d.m. **mg/kg d.m.** Wood Undecayed mean 55 84.7 10.1 147.7 16.4 986.1 116.3 13.9 9.9 57.7 0.4 SD 2.8 53.1 6.2 47.5 2.2 230.2 53 1.4 3.5 28.9 0.1 Moderately decayed mean 50.8 370 54 274.7 15.5 1027.3 122.1 29.2 10.8 48.9 3 SD 1.6 177.6 21.3 121.4 5.7 314.2 60.2 13.2 4 13.9 2.6 Highly decayed mean 48.9 2166.5 149.6 735.2 22.3 1301.4 196.3 32.8 11.9 68 2.9 SD 0.7 863.5 74.7 371.4 6 323 81.7 18.3 2 22.1 1.8 Corridors mean 49.6 1178.3 102 744.8 41.4 1187 296.3 30.1 17.1 62 2.5 SD 1.2 687 46 423.6 17.8 497.6 177.4 15.7 5.2 23.2 1.7 wood categories diff. 1 = 2\>3; 4\<1 1 = 2\<3; 4\>1 1 = 2\<3 = 4 1 = 2\<3 = 4 1 = 2 = 3\<4 n.s. 4\>1 = 2 1\<2 = 3 = 4 1 = 2 = 3\<4 n.s. 1\<2 = 3 = 4 sex diff. -- significant differences between sex categories (Mann-Whitney U, p\<0.05), M -- male, F -- female; species diff. -- significant differences between species of one sex/age category (males, females, pupae) (Kruskal-Wallis test or Mann-Whitney U, p\<0.05); Sr -- *Stictoleptura rubra*, Cm -- *Chalcophora mariana*, Ar -- *Arhopalus rusticus*; wood categories diff. -- significant differences between wood decay categories (Kruskal-Wallis test, p\<0.05). The mean nitrogen content ranged from 6% to 10.9% and differed significantly between species and sexes. The N content was significantly lower in females of all the species, and the pupae showed the lowest values ([Table 1](#pone-0115104-t001){ref-type="table"}). Males of all three species had similar P levels (av. 0.64%, SD 0.1; [Table 1](#pone-0115104-t001){ref-type="table"}). Male and female *S. rubra* did not differ in P content, but their pupae had a significantly lower P content than adults. The P content differed significantly between the sexes in *A. rusticus*, and the pupae of *Ch. mariana* had a significantly higher P content than *S. rubra* pupae ([Table 1](#pone-0115104-t001){ref-type="table"}). The levels of the eight other elements did not differ significantly within species, with only three exceptions: Ca (lower in male than in female *S. rubra*, higher in male than in female *A. rusticus*, Mann-Whitney U test, p\<0.05; [Table 1](#pone-0115104-t001){ref-type="table"}), Mg (lower in male than in female *S. rubra*; [Table 1](#pone-0115104-t001){ref-type="table"}), and Fe (higher in male than in female *A. rusticus*; [Table 1](#pone-0115104-t001){ref-type="table"}). The interspecific differences appear to reflect taxonomic relationships at the level of the subfamilies Buprestidae and Cerambycidae ([Table 1](#pone-0115104-t001){ref-type="table"}): *Ch. mariana* pupae had higher K and Mg contents and lower Fe, Zn and Cu contents than *S. rubra* pupae. *Ch. mariana* males had higher Na and Mg contents and lower Fe, Zn, ad Cu contents than *S. rubra* males. *A. rusticus* males had higher Ca content than *S. rubra* males. PCA allowed for a simultaneous comparison of multi-elemental stoichiometries in all sex, age and species categories of the beetles studied. On the plane determined by the first two axes (54.4% of the total variance), the beetles tended to group according to taxonomy (species and subfamily) and sex ([Fig. 2](#pone-0115104-g002){ref-type="fig"}). The 1st component was loaded mostly by the variance of Fe, Cu, N, and Mg, and the 2nd component was loaded by K, Mn and P levels. The clusters for both sexes of *S. rubra* greatly overlapped but differed from the pupae of this species ([Fig. 2](#pone-0115104-g002){ref-type="fig"}). The other two species tended to differ between themselves and to maintain the body multi-elemental stoichiometry across developmental stages and across sexes; no overlap occurred between clusters for *Ch. mariana* pupae, *Ch. mariana* males, and the clusters for the representatives of Cerambycidae ([Fig. 2](#pone-0115104-g002){ref-type="fig"}). These tendencies were partly confirmed as statistically significant by the ANOVA computed independently for the 1st and 2nd axis scores ([Fig. 3](#pone-0115104-g003){ref-type="fig"}). {#pone-0115104-g002} {#pone-0115104-g003} Elemental content and ergosterol content in wood {#s3b} ------------------------------------------------ The relative concentrations of all elements except carbon tended to increase during the decay process. In the material from corridors, the elemental concentrations were similar to those in highly decayed wood, with the exception of Na, Zn and Mg (concentrations highest in corridors; [Fig. 4](#pone-0115104-g004){ref-type="fig"}, see [S1 Table](#pone.0115104.s001){ref-type="supplementary-material"} for complete data set). The relative increments of the increase in concentration during wood decay (from category 1 to 3) were highest for nitrogen (23-fold), phosphorus (14-fold), copper (6.3-fold) and potassium (4-fold); the other elemental concentrations increased by 18% (Mn) to 136% (Fe). {ref-type="supplementary-material"}.](pone.0115104.g004){#pone-0115104-g004} Ergosterol content in dead wood significantly increased along the decay gradient; each wood category significantly differed from the two others (Kruskal-Wallis test, p\<0.05, [Fig. 5](#pone-0115104-g005){ref-type="fig"}). The respective median values were 39.6 µg/g (dry mass) for the undecayed wood, 169 µg/g for the moderately decayed wood and 385.7 µg/g for the highly decayed wood (see [S4 Table](#pone.0115104.s004){ref-type="supplementary-material"} for more details). {#pone-0115104-g005} Stoichiometric mismatch expressed as the Trophic Stoichiometric Ratio (TSR values) {#s3c} ---------------------------------------------------------------------------------- The *TSR* values were highest for the undecayed wood for all the nutrients ([Table 2](#pone-0115104-t002){ref-type="table"}, [S3 Table](#pone.0115104.s003){ref-type="supplementary-material"}). After pooling the data for the three species and both sexes ([Table 2](#pone-0115104-t002){ref-type="table"}), the stoichiometric mismatch of undecayed wood as food, represented by the *TSR* values, reached three orders of magnitude for N, two orders of magnitude for P, one order of magnitude for K, Na, Mg, Zn and Cu, and less than one order of magnitude for Fe. No stoichiometric mismatch was shown for Ca and Mn: the *TSR* values for these nutrients were lower than one. The stoichiometric mismatch was lower in moderately decayed wood (two orders of magnitude for N and P, one order of magnitude for K, Na, Mg, Zn and Cu) and even lower for highly decayed wood (one order of magnitude for N, P, K, Na, Zn and Cu). The *TSR* values calculated for corridors were between those for moderately decayed wood and highly decayed wood for most of the elements but were the lowest for Na, Mg and Zn ([Table 2](#pone-0115104-t002){ref-type="table"}). 10.1371/journal.pone.0115104.t002 ###### Trophic stoichiometric ratios (*TSR  =  (C:x)~wood~/(C:x)~beetle~*); averages (lower bound -- upper bound) of the bootstrapped distributions. {#pone-0115104-t002-2} Wood decay categories Elements ----------------------- ------------------ ----------------- -------------- -------------- ---------------- ---------- ------------ ---------- ---------------- --------------- Undecayed **2246** **861** **54** **50** *0.99* 18 7.4 14 *0.48* **86** **(1306--3327)** **(681--1088)** **(46--63)** **(42--61)** *(0.72--1.33)* (13--23) (5.6--9.5) (11--18) *(0.38--0.58)* **(63--112)** Moderately decayed **364** **139** **28** **51** *0.93* 14 4 13 *0.47* **32** **(203--656)** **(92--199)** **(22--34)** **(36--68)** *(0.52--1.43)* (10--19) (2.3--6.2) (8--21) *(0.34--0.62)* **(11--60)** Highly decayed **50** **49** **12** **34** *0.69* 9 3.4 10 *0.33* **15** **(40--61)** **(39--62)** **(9--14)** **(27--42)** *(0.50--0.94)* (7--11) (2.5--4.5) (8--12) *(0.27--0.40)* **(10--22)** Corridors **128** **87** **13** **23** *0.85* 7 3.8 7 *0.39* **25** **(77--189)** **(55--146)** **(9--16)** **(16--33)** *(0.58--1.21)* (5--9) (2.7--5.0 (6--9) *(0.31--0.48)* **(14--38)** **Bolded letters** indicate the most limiting nutrients (*TSR\>10* for all the wood categories); *italics* -- nutrients in excess (*TSR \<1* for all the wood categories); see text for definitions of wood decay categories; pooled data for 3 species, both sexes (number of specimens as in [Table 1](#pone-0115104-t001){ref-type="table"}); see text for explanation of bootstrap calculation and [S3 Table](#pone.0115104.s003){ref-type="supplementary-material"} for detailed results. Discussion {#s4} ========== Our results show that dead wood tissue as the sole source for xylophagous beetles cannot provide sufficient nutrients for growing larvae to compose their bodies. In addition to N and P, we found that K, Na, Mg, Fe, Zn and Cu are also limiting nutrients. The larval diet is apparently supplemented by fungal tissues gradually infecting the decaying wood and transporting nutritional elements into the stump. Thus, "xylophages" (also called "wood eaters") are in fact "fungivores", as they feed on fungi. The nutritional demands differ among species belonging to different families, as determined by the content of several elements in the adult bodies: Fe, Zn, Cu, Mg, Na, and N. Within a family, the species do not differ significantly in elemental composition, but the sexes do differ in their elemental composition. The females require more time to develop because they are larger. The pine stumps in this study were inhabited mostly by *S. rubra* and *Ch. mariana*, which belong to different families and differ significantly in their body stoichiometries and life histories. However, the specimens of *A. rusticus*, a species quite similar to *S. rubra* in stoichiometry and in the length of development time, only rarely co-occurred in the stumps. Stoichiometry of xylophages {#s4a} --------------------------- For the three major body components (C, N, P), the body composition of adult xylophagous beetles falls within the broad range of values found in the few other coleopteran taxa studied to date [@pone.0115104-Fagan1], [@pone.0115104-Elser1], [@pone.0115104-Gonzlez1], [@pone.0115104-Sun1]. It has been suggested that besides a taxonomic idiosyncrasy or a presumed body mass allometry, N and P content may reflect the feeding strategy of invertebrates, with predators having higher concentrations of N and P than herbivores and detritivores [@pone.0115104-Fagan1], [@pone.0115104-Gonzlez1]. The xylophagous beetles studied here do not confirm this generalization: their N content is close to the high values reported for carnivores, and their P content is intermediate between herbivores and carnivores [@pone.0115104-Fagan1], [@pone.0115104-Gonzlez1]. However, our results are consistent with species-specific data provided by Fagan et al. [@pone.0115104-Fagan1], who found high N levels in several cerambycid and buprestid beetles. Females of all three species studied here have significantly lower N concentrations than males of the same species, possibly associated with the higher fat content in the females. Because the sex of pupae could not be determined, and the samples most likely contain individuals of both sexes, their average body composition should be intermediate between males and females. The position of pupae of the two species studied on the PCA plot ([Fig. 3](#pone-0115104-g003){ref-type="fig"}) suggests that they have relatively high C levels and low N and P levels. This result may be attributed to the higher C:N ratio of the chitinous exuvium left behind at eclosion and the fat reserves exhausted during the pupal stage. Stoichiometry of wood decay {#s4b} --------------------------- The concentrations of the nutrients (N, P, K, Ca, Mg) measured in the sapwood and heartwood of living gymnosperm trees [@pone.0115104-Meerts1] are an order of magnitude greater than in the undecayed wood of the dead pine stumps that we studied. The stumps, cut one to four years before the wood samples were taken for analysis, were remnants of trees cut when they were at least 80 years old. Palviainen et al. [@pone.0115104-Palviainen1], [@pone.0115104-Palviainen2] reported similarly low nutrient concentrations measured in pine stumps during the first five years of decay after cutting. Meerts [@pone.0115104-Meerts1] suggested that mineral nutrients in a living tree may be recycled from senescing sapwood. After cutting, the stumps are exposed to weathering and may be further depleted of nutrients until the fungal mycelium growing into the wood enriches it with nutrients, particularly N and P, imported from outside the system [@pone.0115104-Boddy1]--[@pone.0115104-Clinton1]. Even so, there is still a significant difference in the stoichiometries between partly decayed wood and the insects feeding on it. The nutritional composition of corridor material is similar to that of the wood from which the samples were taken (moderately and highly decayed), except for Na, Zn and Mg, which are present in the corridors at higher concentrations ([Fig. 4](#pone-0115104-g004){ref-type="fig"}). Most of the corridors that we measured originated from moderately decayed wood and less commonly from highly decayed wood. It is not clear whether the chemical composition of the wood is nonuniform and the larvae bore selectively in the more nutritious areas or whether the chemical composition of the corridors changed due to the more intense penetration of fungal mycelium following larval activity. The dynamics of nutrient content in decaying wood {#s4c} ------------------------------------------------- Two mechanisms contribute to the decrease of C:x ratios during decay, that is, to the enrichment of the wood with elements other than C, H and O: (i) the liberation of C as CO~2~ due to microbial and animal respiration, and (ii) the import of nutrients from outside the system by fungal tissue (mycelium) growing into the decaying wood. To determine whether elements are imported in significant amounts during wood decay, we assessed the relative contributions of the two mechanisms from the stoichiometric proportions of specific elements in the wood at various stages of decay. The initial (*SRI*) and final (*SRF*) stoichiometric ratios in decaying wood for a given element x may be expressed as: where *pC~I~* and *pX~I~* are initial relative concentrations of carbon and of element *x*, respectively; *C~I~*, *C~F~*, *X~I~* and *X~F~* are the initial and final absolute amounts of carbon and of element x, respectively; α represents the proportion of carbon not released as CO~2~ during decay; and *β* represents the coefficient of enrichment of element *x*. From [equation (3](#pone.0115104.e003){ref-type="disp-formula"}), it follows that A value of *β\>\>1* would indicate an increase of the absolute amount of element *x* in the wood during decay. The coefficient *α* is defined as:where *M~I~* and *M~F~* are initial and final masses of decaying wood, respectively. The concentration of carbon decreases from 55% to 48.9% during decay ([Table 1](#pone-0115104-t001){ref-type="table"}); thus, *pC~F~/pC~I~  = 0.489/0.550 = 0.889*. The unknown proportion *M~F~/M~I~* can approximated from literature data concerning the rates of decay of coarse woody debris (e.g. [@pone.0115104-Russell1], [@pone.0115104-Fukasawa1]), including those of conifer stumps [@pone.0115104-Tobin1], [@pone.0115104-Garrett1]. The proportion of the remaining mass of wood *d* after *t* years of decomposition is usually described with exponential model . The experimentally evaluated constant *k* may range between 0.020 and 0.1101, depending on the tree taxon, environmental conditions and the methods used [@pone.0115104-Russell1]--[@pone.0115104-Garrett1]. Samples of wood at advanced stages of decay were taken from the stumps of trees cut 3 or 4 years before sampling. The model solved for 3 years of decay using these figures yields the proportion of remaining mass between 0.72 and 0.94 and solved for 4 years 0.64 to 0.92. Based on these estimates, the possible values of the coefficient α (eq. 5) may range from 0.57 to 0.84. The range of *β* values can be estimated for each element *X* using *β~x~ = α×SRI~x~/SRF~x~*, where *SRI~x~* and *SRF~x~* are the measured initial and final *C/X* ratios. The results ([Table 3](#pone-0115104-t003){ref-type="table"}) show that independently of the assumed value of α, the amounts of Na, Ca, Mg, Zn, Mn and Fe do not increase during decay (coefficients of enrichment do not deviate substantially from 1.0, ranging from 0.7 to 1.4 for *α* = 0.57 and from 1.0 to 2.1 for *α* = 0.84), but the contents of N, P, Cu and K increase several fold, independent of the assumed value of *α* ([Table 3](#pone-0115104-t003){ref-type="table"}). Thus, the change in the relative concentrations of these elements is not only the result of the carbon escape from decaying wood but also the result of the net import of these elements from outside of the system. 10.1371/journal.pone.0115104.t003 ###### Coefficient of enrichment of elements in decaying wood (see text for explanation). {#pone-0115104-t003-3} N P K Na Ca Mg Fe Zn Mn Cu ---------------------------- ------- ------------ ---------- ---------- --------- ------ --------- ----- ----- ----- ------- --------- C:x in undecayed wood *SRI* 6111 55000 0.4 3.4 0.1 0.5 4 5.6 1 137.5 C:x in highly decayed wood *SRF* 253.5 3667 0.1 2.5 0.04 0.3 1.7 4.6 0.8 19 *α = 0.84* **20.1** **12.5** **3.3** 1.1 **2.1** 1.4 2 1 1 **6** Coefficient of enrichment *β* *α = 0.57* **13.7** **8.5** **2.3** 0.8 1.4 0.9 1.3 0.7 0.7 **4.1** Bolded letters indicate the coefficients of enrichment exceeding 2, indicative for the elements imported to the decaying wood from outside. Although part of the additional N may be supplied by nitrogen-fixing bacteria, which are known to occur in decayed plant matter [@pone.0115104-Roskoski1], fungi appear to be the only agents translocating other nutrients from outside of the system to the decaying wood. In fact, the ergosterol content (the fungal tissue proxy) increased along the decay gradient and differed significantly between dead wood categories ([Fig. 5](#pone-0115104-g005){ref-type="fig"}). The measured ergosterol contents cannot be simply recalculated into fungal biomass because the conversion factors are strongly dependent on the fungal species [@pone.0115104-Niemenmaa1]. However, the contents of N, P. K and Cu in the mycelium of wood decaying fungi [@pone.0115104-Vetter1], [@pone.0115104-Campos1] and in the fruiting bodies of mushrooms [@pone.0115104-Vetter1], [@pone.0115104-Rudawska1] are two to three orders of magnitude higher than in undecayed wood (this study), while the contents of the other elements studied here differ less than one order of magnitude. Thus, the increase in nutrient concentration in decaying wood may be attributed to the action of fungi. An experiment in which xylophagous beetle larvae were fed with fungi instead of wood [@pone.0115104-Tanahashi1] showed that fungi can be an adequate food source for these insects. Fungi inhabiting dead wood have been described as nutrient immobilizers [@pone.0115104-Dighton1], and our data support the view that fungi may serve as nutrient deliverers [@pone.0115104-Palviainen1], [@pone.0115104-Palviainen2], [@pone.0115104-Watkinson1], [@pone.0115104-Clinton1], [@pone.0115104-Lindahl1], [@pone.0115104-Lindahl2]. The translocation of elements by fungi in the forest floor is a well-known phenomenon [@pone.0115104-Boddy1], [@pone.0115104-Watkinson1], [@pone.0115104-Cairney1], [@pone.0115104-Boddy2]. Nutritional imbalance in wood-boring insects {#s4d} -------------------------------------------- The concentrations of elements in the imagines and pupae were one or more orders of magnitude greater than in the potential food of the larvae, except for Ca, Mn and Fe, which showed a less pronounced discrepancy ([Table 1](#pone-0115104-t001){ref-type="table"}). The mismatch, expressed as the *TSR* value, is most striking for undecayed wood (three orders of magnitude for nitrogen) and diminishes as wood decay proceeds, but the differences remain large ([Table 2](#pone-0115104-t002){ref-type="table"}; for full data set, see [S3 Table](#pone.0115104.s003){ref-type="supplementary-material"}). The nutrient content of wood from corridors is close to that of highly decayed wood ([Table 2](#pone-0115104-t002){ref-type="table"}). The most extreme differences are for N and P, which are, respectively, 1500--2000 and 500--900 times less concentrated in undecayed wood than in the beetles. The concentration differences for Cu, K and Na are also significant. Cu is approximately 86 times higher in beetles than in undecayed wood. The K and Na concentrations are 54 and 50 times higher, respectively, in the beetles than in undecayed wood ([Table 2](#pone-0115104-t002){ref-type="table"}). Only Ca and Mn are available in excess. The other nutrients (K, Na, Mg, Fe, Zn, Cu) are much more scarce in dead wood, such that they may constrain the development of wood-eating larvae. All the elemental concentrations tend to increase as wood decay proceeds ([Table 2](#pone-0115104-t002){ref-type="table"}, [Fig. 4](#pone-0115104-g004){ref-type="fig"}), although the *TSR* values remain quite high for N and P. Considering wood as the exclusive source of nutrients, xylophagous beetles seem to be faced with the most unbalanced diet of all organisms studied to date. Even termites live on a less unbalanced diet [@pone.0115104-Sterner1]. The N and P stoichiometric mismatches were similar in all three studied species. We found differences concerning other nutrients: the Na mismatch for *Ch. mariana* is almost twice as high as that of the Cerambicidae beetles, whereas the Fe and Cu deficiencies for *Ch. mariana* are two to one order of magnitude lower. Thus, according to this study, different taxa of xylophagous beetles occupying the same nutritional niche may be faced with different stoichiometric mismatches concerning elements other than N and P. Nutritional limitations of xylophage life history and stoichiometric compensation by fungi {#s4e} ------------------------------------------------------------------------------------------ Walczyńska [@pone.0115104-Walczyska1], [@pone.0115104-Walczyska2], using experimental measurements of consumption, assimilation and growth efficiences of larvae feeding on pine wood (with stoichiometric conditions identical to the ones used in the present study), demonstrated that the low digestibility of wood may affect the life history of *S. rubra* by prolonging the development time. This result poses the following question: is the development time long enough to concentrate essential nutrients to the levels required in the body? We calculated the minimum growth period needed to collect sufficient essential element *x* (GP~x~, years) aswhere *B* -- average mass of a larva, *A~X~* -- concentration of element *x* in the body of an imago or pupa, *K* -- daily food consumption rate averaged through the development period (after Walczyńska [@pone.0115104-Walczyska1]), *F~X~* -- concentration of element *x* in food. The conversion efficiency of limiting nutrients was assumed at 100%, and seasonal changes in the food consumption rate were incorporated (nothing was consumed for half the year). Eating an exclusive diet of undecayed wood would lengthen the time needed to form body tissues to an implausible 40 years for males and 85 years for females ([Table 4](#pone-0115104-t004){ref-type="table"}), and the overall assimilation efficiency would drop to improbably low values. Based on field observations, the maximum development time for *S. rubra* was estimated at three years [@pone.0115104-Dominik1]. Only highly decayed pine stumps or the material from corridors could provide the beetles with enough of the most limiting nutrients. Larvae found in highly decayed stumps were quite well developed at the age of at least 1.5 years. The stumps were likely not highly decayed during early larval development. The great majority of larvae collected from stumps were found in moderately decayed wood, which is relatively poor in nutrients ([Tables 2](#pone-0115104-t002){ref-type="table"}, [4](#pone-0115104-t004){ref-type="table"}). The higher content of nutritive elements in material from corridors may suggest that larvae are capable of selecting areas of wood more heavily infected with fungi and thus providing adequate amounts of nutrients or that the activity of the larvae facilitates fungal infection through the corridors ([Fig. 2](#pone-0115104-g002){ref-type="fig"}). Nonetheless, the enriched chemical composition of material from corridors permits the beetles to complete their life cycle within their maximal lifetime in males ([Table 4](#pone-0115104-t004){ref-type="table"}). The females must further prolong their life history to be able to assimilate the amount of nutrients sufficient to build up their bodies. 10.1371/journal.pone.0115104.t004 ###### Estimated minimum number of years a larva would need to spend feeding on wood of different decay categories to gather the amounts of the most limiting essential elements present in the bodies of adult beetles and pupae. {#pone-0115104-t004-4} N P K Na Cu -------------- -------------------- -------- -------- ------- ------- ------- Male imago Undecayed **40** **21** 2 1 3 Moderately decayed **10** **4** 1 1 0 Highly decayed 2 1 0 1 0 Corridors **3** 2 0 1 1 Female imago Undecayed **85** **51** **4** **4** **8** Moderately decayed **21** **10** 2 **4** 1 Highly decayed **4** **3** 1 **3** 1 Corridors **6** **5** 1 2 1 Pupa Undecayed **79** **36** **3** **4** **4** Moderately decayed **19** **7** 2 **4** 0 Highly decayed **3** 2 1 **3** 0 Corridors **6** **4** 1 1 1 Bolded values are the estimated periods exceeding the maximum reported lifetime of *Stictoleptura rubra* larvae (3 years, see text for definitions of wood decay categories). Wood-boring insects, enhancing the decomposition process of dead wood by mechanical grinding and fragmentation of the solid stumps, depend in turn on the fungal supplementation of their food with nutrients. Thus, elemental transport by fungi plays a pivotal role in the function of forest ecosystems: matching the stoichiometric balance between the trophic links. Conclusions {#s5} =========== 1\. During larval development, xylophagous beetles are confronted with a severe nutritional imbalance caused by poor digestibility of food and its stoichiometric mismatch with the beetles\' bodies. 2\. These nutritional constraints are partly offset by the adjusted life histories of xylophages, with a development time extended to a couple of years. The life histories of dimorphic sexes and various species exploiting the same resources may differ, but computational simulations show that the prolongation of the development time is not sufficient to accumulate nutrients in adequate amounts. 3\. The nutritional balance of growing xylophagous larvae can be maintained due to the substantial enrichment of dead wood with nutrients imported from the outside by the mycelium of saprotrophic fungi. 4\. The fungal transfer of essential nutrients from the soil into the wood of dead trees is of fundamental importance for maintaining the detrital food web in forest ecosystems. Supporting Information {#s6} ====================== ###### **Element content in adult beetles, pupae, and wood samples from pine stumps inhabited by larvae.** Wood decay categories: 1 undecayed - hard and healthy; 2 moderately decayed -- colored, moist and softer than (1) but too hard for knife; 3 highly decayed -- visible changes, layers of white or brown rotting fungi, wet and soft, easily torn apart by knife or even by hand; 4 corridors -- walls of corridors made by feeding larvae, together with their content. (XLSX) ###### Click here for additional data file. ###### **Dry body mass of the studied beetle species.** (XLSX) ###### Click here for additional data file. ###### **Trophic stoichiometric ratios (TSRX  =  (C:X)wood/(C:X)beetle, where C -- content of carbon, X -- content of element x for beetles and the potential food of their larvae.** Bolded numbers indicate the most limiting nutrients; italics -- nutrients in excess. Means (white background) and confidence limits (grey background) estimated using bootstrap resampling. Wood decay categories as in [Table S1](#pone.0115104.s001){ref-type="supplementary-material"}. (XLSX) ###### Click here for additional data file. ###### **Ergosterol content in wood.** Wood decay categories: 1 undecayed - hard and healthy; 2 moderately decayed -- colored, moist and softer than (1) but too hard for knife; 3 highly decayed -- visible changes, layers of white or brown rotting fungi, wet and soft, easily torn apart by knife or even by hand. (XLSX) ###### Click here for additional data file. The authors are indebted to Ola Walczyńska, Filip Kapustka, Justyna Kierat, Łukasz Sobczyk, Ulf Bauchinger, January Weiner III and anonymous reviewers for constructive critical comments. We also thank Maciej Choczyński, Paweł Dudzik and Patrycja Gibas for assistance during the analyses. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: JW MF. Performed the experiments: MF. Analyzed the data: MF JW. Contributed reagents/materials/analysis tools: JW MF. Wrote the paper: MF JW.

All relevant data are found within the paper and its Supporting Information files. Introduction {#sec001} ============ The human DEK oncoprotein is a predominantly chromatin-bound factor that regulates nuclear processes such as chromatin architecture, epigenetics, transcription and DNA repair \[[@pgen.1007227.ref001]--[@pgen.1007227.ref018]\]. DEK was originally identified as a fusion protein with the CAN/NUP214 nucleoporin in a patient with acute myeloid leukemia harboring the chromosomal translocation (t6;9)(p23;q34) \[[@pgen.1007227.ref019]\]. Since its discovery, DEK was also shown to be increased in acute myeloid leukemia types that do not harbor the DEK-NUP214 fusion protein \[[@pgen.1007227.ref020]--[@pgen.1007227.ref022]\] and to be frequently overexpressed in solid tumors including colon, breast, gastric adenocarcinoma, ovarian carcinomas, bladder cancer, retinoblastoma, lung, pancreatic, neuroendocrine prostate cancer, hepatocellular, skin cancer, head and neck cancer squamous cell carcinoma (HNSCC), and esophageal squamous cell carcinoma (ESCC; **[S5 Fig](#pgen.1007227.s005){ref-type="supplementary-material"}**) \[[@pgen.1007227.ref023]--[@pgen.1007227.ref040]\]. Additionally, high DEK expression is associated with poor prognosis in melanoma, gastric, ovarian, breast, prostate, bladder, lung, pancreatic, skin cancer, and head and neck SCC \[[@pgen.1007227.ref025], [@pgen.1007227.ref026], [@pgen.1007227.ref030], [@pgen.1007227.ref031], [@pgen.1007227.ref033]--[@pgen.1007227.ref035], [@pgen.1007227.ref040]--[@pgen.1007227.ref043]\]. Esophageal carcinomas are the sixth most common cause of cancer related death worldwide, and eighth in incidence worldwide \[[@pgen.1007227.ref044]--[@pgen.1007227.ref046]\]. Esophageal carcinoma occurs as either SCC or adenocarcinoma \[[@pgen.1007227.ref047]\]. Esophageal SCC accounts for one third of esophageal cancer cases in the United States but represents more than 90% cases of esophageal cancer worldwide \[[@pgen.1007227.ref047], [@pgen.1007227.ref048]\]. The most common risk factors for ESCCs, similar to HNSCC, include tobacco smoke, heavy alcohol consumption, and infection with human papillomavirus \[[@pgen.1007227.ref049], [@pgen.1007227.ref050]\]. Several studies have additionally revealed that ESCC and HNSCC harbor similar genetic and molecular alterations \[[@pgen.1007227.ref044], [@pgen.1007227.ref051]--[@pgen.1007227.ref054]\] and are treated with similar regimen of surgery and chemoradiation \[[@pgen.1007227.ref050]\]. However, the 5-year survival rate for patients with HNSCC is over 50%, while for patients with ESCC it remains at a dismal 5--15% \[[@pgen.1007227.ref045], [@pgen.1007227.ref046], [@pgen.1007227.ref048]\]. Current treatment regimens frequently result in irreparable tissue damage and disfiguration that additionally highlight the need for continued identification of oncogenic drivers and targeted therapies \[[@pgen.1007227.ref055]\]. SCC arises from keratinocytes in squamous epithelium, and the overexpression of DEK has been shown to promote cell survival, proliferation, and transformation in combination with classical oncogenes while inhibiting apoptosis, cellular differentiation and senescence \[[@pgen.1007227.ref016], [@pgen.1007227.ref056]--[@pgen.1007227.ref059]\]. DEK overexpression occurs through various mechanisms including gene amplification, increased transcription, and mutations in microRNAs and ubiquitin ligases responsible for DEK mRNA and protein degradation, respectively \[[@pgen.1007227.ref060]--[@pgen.1007227.ref070]\]. Several *in vivo* studies demonstrate the critical role of human and murine Dek in driving benign and malignant tumor growth. For example, *Dek* knockout (*Dek-/-*) mice are partially resistant to the formation of benign skin papillomas when treated with DMBA and TPA, a tumor initiator and promoter, respectively \[[@pgen.1007227.ref016]\]. In a breast cancer mouse model, *Dek-/-* mice bred to Ron receptor tyrosine kinase transgenic mice, displayed a delayed onset of mammary tumors compared to *Dek+/+* mice \[[@pgen.1007227.ref071]\]. In another study, *Dek* knockout *(Dek-/-) HPV E7* oncogene transgenic mice were protected from 4-nitroquinoline 1-oxide (4NQO)-induced HNSCC and ESCC tumor growth, but not initiation, when compared to their *Dek+/+* counterparts \[[@pgen.1007227.ref039]\]. Taken together, these studies support the possible importance of Dek overexpression as a key driver of uncontrolled cellular growth and tumor development. Historically, most of the data that links DEK overexpression to oncogenic phenotypes were obtained from knockdown and knockout model systems. Only recently has DEK overexpression been investigated *in vivo*. In a 2017 report, Nakashima et. al. generated tetracycline inducible, whole body, Dek over-expressing mice \[[@pgen.1007227.ref072]\]. The mice were treated with 4NQO in the drinking water for 28 weeks to induce oral lesions, then induced to overexpress Dek for 4 weeks before sacrifice. 4NQO is a chemical carcinogen that mimics the effects of tobacco smoke by forming DNA adducts and mutations similar to those seen in human HNSCC and ESCC \[[@pgen.1007227.ref073], [@pgen.1007227.ref074]\]. When administered in drinking water, 4NQO stimulates susceptibility to squamous cell carcinomas in the tongue, oral cavity, and esophagus \[[@pgen.1007227.ref073], [@pgen.1007227.ref075]\]. In the study, the mice over-expressing Dek for 4 weeks, post 4NQO treatment, harbored significantly increased hyperplasia in the tongue with a trend toward increased tongue tumor incidence. Interestingly, Dek overexpression significantly decreased tongue tumor diameter \[[@pgen.1007227.ref072]\]. This suggests that a short term induction of Dek overexpression after long term carcinogen exposure has pro- and anti-tumorigenic effects. Importantly, this study demonstrated that Dek overexpression promotes cellular proliferation in tissues exposed to carcinogens. However, whether these effects are due to high Dek expression in keratinocytes as the cell of origin, and/or other cell types, remains unknown. Therefore, we targeted long term induction of the Dek transgene to the stratified squamous epithelium, and monitored resulting tumor phenotypes. To this end, a tetracycline responsive *Dek* and *luciferase* transgenic *Bi-L-Dek* mouse model was newly generated. *Bi-L-Dek* transgenic mice harbor a tetracycline response element (TRE) that controls the bi-directional expression of Dek and firefly luciferase. The TRE allows for temporal and tissue specific control of Dek overexpression, thus making it a versatile mouse model wherein the Dek transgene expression is controlled by tetracycline or its more stable derivative, doxycycline (dox). *Bi-L-Dek* mice were crossed to keratin 5 tetracycline transactivator *(K5-tTA*) transgenic mice to target Dek and luciferase expression to basal keratinocytes that serve as progenitor cells for stratified squamous epithelium and are the cell of origin for squamous cell carcinoma (SCC) of the tongue and esophagus. The tTA protein produces a tet-off system where expression of the Dek transgene is repressed by dox. Dek overexpression and transgene repression by dox was verified in the skin, tongue and esophagus. Once validated, *Bi-L-Dek_K5-tTA* mice were subjected to 4NQO treatment in the presence or absence of dox. Dek caused a trend toward increased proliferation in tongue and esophageal epithelium after 4NQO treatment. Furthermore, Dek overexpression was sufficient to increase the incidence of gross esophageal, but not oral, SCC tumor formation in this system. This data suggests that Dek contributes to ESCC tumorigenesis at least partially through keratinocyte intrinsic pathways which promote cellular and tumor growth. Results {#sec002} ======= Generation of a conditional *Dek* transgenic tetracycline-off mouse model {#sec003} ------------------------------------------------------------------------- In order to overexpress Dek conditionally in a tissue specific manner, we utilized a construct *Bi-L-Dek* wherein Dek and luciferase gene expression were driven by a tetracycline response element (TRE). To generate the *Bi-L-Dek* transgene, Dek cDNA was cloned into the Bi-L-Tet plasmid expression vector (Clontech, Mountain View, CA, USA) as published by others \[[@pgen.1007227.ref076]\], described in the Materials and Methods, and illustrated in **[S1A Fig](#pgen.1007227.s001){ref-type="supplementary-material"}**. Following validation of TRE-dependent Dek and luciferase expression in vector transfected cells **([S1B--S1E Fig](#pgen.1007227.s001){ref-type="supplementary-material"})**, the transgene was excised and injected into the pronucleus of fertilized mouse eggs for generation of *Bi-L-Dek* transgenic founders (**[S1F Fig](#pgen.1007227.s001){ref-type="supplementary-material"}**). *Bi-L-Dek* mice harbor a TRE that controls two mini cytomegalovirus (CMV) promoters driving bi-directional transcription of *Dek* and *luciferase* (**[Fig 1A](#pgen.1007227.g001){ref-type="fig"}**). Four *Bi-L-Dek* transgenic founder lines were assessed for transgene stability over four generations before screening for doxycycline responsive expression of Dek (**[Fig 1B](#pgen.1007227.g001){ref-type="fig"}**; the data for founder line \#317 used in subsequent experiments is shown). {#pgen.1007227.g001} To determine which lines expressed Dek and luciferase under control of the TRE, *Bi-L-Dek* mice were crossed to *K5-tTA* mice (**[Fig 1C and 1D](#pgen.1007227.g001){ref-type="fig"}**). The keratin 5 promoter targets tTA protein expression to the basal layer of stratified squamous epithelium including that of the esophagus, tongue, and skin (**[Fig 1E](#pgen.1007227.g001){ref-type="fig"}**). In this system, administration of doxycycline (dox) represses tTA binding to the TRE to inhibit *Bi-L-Dek* transgene expression (**[Fig 1E](#pgen.1007227.g001){ref-type="fig"}**). Founder \#317 was chosen for subsequent experiments, harbors approximately three copies of the transgene (**[Fig 1B](#pgen.1007227.g001){ref-type="fig"}**), and is referred to as *Bi-L-Dek* from here on. *Bi-L-Dek_K5-tTA* mice overexpress Dek conditionally in stratified squamous epithelium {#sec004} -------------------------------------------------------------------------------------- *Bi-L-Dek* transgene expression in *Bi-L-Dek_K5-tTA* mice was validated with multiple methodologies. These included an *in vivo* imaging system (IVIS), and the detection of Dek mRNA and protein expression by quantitative, real time polymerase chain reaction (RT-qPCR), western blot analysis, *in situ* immunohistochemistry (IHC) and immunofluorescence (IF) **([Fig 2](#pgen.1007227.g002){ref-type="fig"})**. IVIS and *ex vivo* imaging confirmed luciferase expression in the skin of *Bi-L-Dek_K5-tTA* bi-transgenic mice and in the esophagus (**[Fig 2A and 2B](#pgen.1007227.g002){ref-type="fig"}**). Dek mRNA levels were induced 3.5 fold over endogenous levels in the skin of *Bi-L-Dek_K5-tTA* mice, and repression to endogenous levels was achieved by feeding with dox chow for seven days (**[Fig 2C](#pgen.1007227.g002){ref-type="fig"}**). Dek protein expression in the skin also increased approximately 3 fold over the levels of endogenous Dek in the *Bi-L-Dek_K5-tTA* mice (**[Fig 2D](#pgen.1007227.g002){ref-type="fig"}**). Dek overexpression in the tongue was detected by IHC along with the expected decrease in the corresponding mice on dox chow (**[Fig 2E](#pgen.1007227.g002){ref-type="fig"}**). Finally, we isolated keratinocytes from *Bi-L-Dek_K5-tTA* skin for cell culture and performed IF with antibodies against Dek and K5, then stained with DAPI to detect DNA. As expected, Dek expression was higher in the *Bi-L-Dek_K5-tTA* derived keratinocytes compared to those treated with dox, or compared to keratinocytes from single transgenic control mice (**[Fig 2F and 2G](#pgen.1007227.g002){ref-type="fig"}**). Exogenous Dek localized to the nucleus as expected, and co-localized with endogenous Dek and DAPI (**[Fig 2F](#pgen.1007227.g002){ref-type="fig"}**). Altogether, *Bi-L-Dek_K5-tTA* mice overexpressed Dek in the squamous epithelium of the skin, tongue, and esophagus, and Dek expression was repressed by doxycycline. ![*Bi-L-Dek_K5-tTA* mice express luciferase and overexpress Dek in stratified squamous epithelium.\ (**A**) *In vivo* imaging system (IVIS) analysis depicts a single (*Bi-L-Dek*) and a bi-transgenic (*Bi-L-Dek_K5-tTA)* mouse after intraperitoneal injection of luciferin for luciferase detection in the skin of *Bi-L-Dek_K5-tTA* mice. (**B**) *Ex vivo* IVIS analysis of single transgenic (*K5-tTA*) versus bi-transgenic *(Bi-L-Dek_K5-tTA)* flank skin, ear, and esophagus following injection of luciferin, sacrifice, and dissection. (**C**) RT- qPCR of Dek mRNA levels in skin epithelium obtained from the flank of mice show a 3 fold induction of Dek transcript levels that is repressed to endogenous levels after seven days on dox chow. Primers detect endogenous and exogenous *Dek*. Error bars represent three mice for each genotype excluding the *Dek-/-* negative control which represents one mouse repeated in triplicate. (**D**) Representative western blot analysis for the detection of Dek protein levels in flank skin epithelium demonstrates increased levels of Dek protein in *Bi-L-Dek_K5-tTA* mice over those on dox and single transgenic controls. (**E**) Immunohistochemistry (IHC) with DEK antibodies (BD Biosciences, San Jose, CA, USA) in tongue epithelium confirms Dek protein overexpression in *Bi-L-Dek_K5-tTA* mice that is repressed within seven days of dox chow. (**F**) Immunofluorescence (IF) of cultured skin keratinocytes isolated from newborn *Bi-L-Dek_K5-tTA* pups with or without dox and their single transgenic littermates. Dox treated keratinocytes were cultured with 1ug/ml of dox for 48 hours before fixation. IF images of keratinocytes were taken at the same magnification and exposure after being probed for Dek, keratin 5 (K5), and stained with DAPI. (**G**) The mean fluorescent intensity of Dek staining in **2F** was quantified using ImageJ software (National Institutes of Health, Bethesda, Maryland, USA) \[[@pgen.1007227.ref089]\].](pgen.1007227.g002){#pgen.1007227.g002} To assess the extent of Dek overexpression in the *Bi-L-Dek_K5-tTA* mice, we quantified transgene expression in the context of Dek knockout mice. *Bi-L-Dek* and *K5-tTA* transgenic mice were interbred with *Dek-/-* mice to generate *Dek-/-*\_ *Bi-L-Dek_K5-tTA* offspring (**[Fig 3A](#pgen.1007227.g003){ref-type="fig"}**). IVIS confirmed luciferase expression (**[Fig 3B](#pgen.1007227.g003){ref-type="fig"}**), and RT-qPCR and western blot analysis confirmed Dek mRNA and protein expression, respectively, in the epidermis (**[Fig 3C](#pgen.1007227.g003){ref-type="fig"} and [S2 Fig](#pgen.1007227.s002){ref-type="supplementary-material"}**). Dek mRNA levels were induced by approximately four fold in the *Bi-L-Dek_K5-tTA* compared to control mice (**[S2 Fig](#pgen.1007227.s002){ref-type="supplementary-material"}**) and Dek protein levels were induced by approximately 2--3 fold in the *Bi-L-Dek_K5-tTA* over control mice. These levels are similar to the levels of DEK expression that can be routinely achieved in normal epithelial cells transduced with retroviral or lentiviral DEK expression vectors and are within the range of DEK levels observed in cancer cells \[[@pgen.1007227.ref016], [@pgen.1007227.ref042], [@pgen.1007227.ref056], [@pgen.1007227.ref077]--[@pgen.1007227.ref081]\]. {#pgen.1007227.g003} Brain-specific and global *Bi-L-Dek* transgene expression is achievable via *Dlx5/6-tTA* and *Rosa-tTA* drivers {#sec005} --------------------------------------------------------------------------------------------------------------- To explore the broader utility of this model system, conditional *Bi-L-Dek* mice were bred to *Dlx5/6-tTA* mice to target Dek expression to neurons originating from the ventral forebrain (**[S3 Fig](#pgen.1007227.s003){ref-type="supplementary-material"}**). As expected, robust Dek protein overexpression was detectable in cortical interneurons and striatal projection neurons from the ventral forebrain, as previously demonstrated for other Dlx5/6-driven transgenes \[[@pgen.1007227.ref082]\]. Additionally, we crossed the *Bi-L-Dek* mice with *Rosa-tTA* mice to produce global Dek overexpressing mice. IVIS demonstrated luciferase expression from the *Bi-L-Dek* transgene throughout the body, which was repressed with dox chow (**[S4 Fig](#pgen.1007227.s004){ref-type="supplementary-material"}**). No overt phenotypes were observed in the mice similar to results in previously published *Tet-O-Dek*\_ *Rosa26-M2rtTA* mice. In all, these results further validate *Bi-L-Dek*-mediated transgene expression in murine epithelia, and demonstrate broad utility of this genetic mouse model for studies of Dek overexpression in other organ systems. Dek overexpression promotes esophageal squamous cell carcinoma {#sec006} -------------------------------------------------------------- Based on data in the cancer genome atlas (TCGA), DEK is more highly expressed in ESCC compared to normal tissue and in ESCC compared to esophageal adenoma ([S5 Fig](#pgen.1007227.s005){ref-type="supplementary-material"}). However, the contribution of Dek overexpression to ESCC development is unknown. To determine if DEK contributes to ESCC development or progression, we utilized the *Bi-L-Dek_K5-tTA* mice with Dek overexpression targeted to basal keratinocytes that form the epithelium. *Bi-L-Dek_K5-tTA* mice overexpressed Dek in stratified squamous epithelium of the tongue and esophagus, and *Bi-L-Dek_K5-tTA* mice on dox expressed only endogenous levels of Dek. Exposure of mice to drinking water containing the soluble quinoline derivative 4NQO promotes the development of oral and/or esophageal cancer. Therefore, we exposed two groups of mice +/- Dox to 4NQO in order to determine whether Dek overexpression in basal keratinocytes is sufficient to promote SCC and if early onset of Dek overexpression increases oral and/or esophageal tumor incidence or tumor burden. The experimental design is illustrated in **[Fig 4A](#pgen.1007227.g004){ref-type="fig"}**. At six weeks of age, *Bi-L-Dek_K5-tTA* mice in the absence of dox (n = 7) or in the presence of dox (n = 5) were exposed to 10ug/mL of 4NQO in their drinking water to promote SCC susceptibility. After 16 weeks, mice were given normal water, sacrificed at 45 weeks of age or when moribund, and analyzed after experimentally induced death or sacrifice. One hour prior to sacrifice, mice were injected with the thymidine analog Bromodeoxyuridine (BrdU) to quantify proliferation. In the absence of 4NQO treatment, Dek overexpression did not significantly increase cellular proliferation in the tongue or the esophagus when compared to control mice on dox (**[Fig 4B](#pgen.1007227.g004){ref-type="fig"}**). However, following 4NQO treatment there was a trend toward increased proliferation in the epithelia of Dek overexpressing tongue (p = 0.07) and esophagus (p = 0.15) (**[Fig 4C and 4D](#pgen.1007227.g004){ref-type="fig"}**). These results are in line with recently published data wherein global Dek overexpressing mice did not exhibit hyperplasia or other phenotypes under normal conditions, but displayed tongue hyperplasia after 4NQO treatment \[[@pgen.1007227.ref072]\]. *Bi-L-Dek_K5-tTA* mice exposed to 4NQO or not were then analyzed for the presence of tumors in the tongue and esophagus. Detailed results are shown for each mouse in **[Fig 5A](#pgen.1007227.g005){ref-type="fig"}**. After 4NQO treatment, *Bi-L-Dek_K5-tTA* mice continued to express higher levels of Dek protein in esophageal epithelium compared to the control group on dox (**[Fig 5B](#pgen.1007227.g005){ref-type="fig"}**). In addition, Dek overexpressing mice had a significantly higher incidence of gross esophageal tumors (**[Fig 5C](#pgen.1007227.g005){ref-type="fig"}**). Specifically, all of the *Bi-L-Dek_K5-tTA* mice developed at least one visible esophageal tumor (100%), in contrast to only one of five *Bi-L-Dek_K5-tTA* mice on dox (20%). Furthermore, one Dek overexpressing mouse harbored an excessively large tumor, while two others harbored two separate grossly apparent tumors (**[Fig 5A, 5C and 5D](#pgen.1007227.g005){ref-type="fig"}**). From published studies, the 4NQO protocol utilized was expected to result in a 10% incidence of gross tumors in *Dek* wild type mice \[[@pgen.1007227.ref075], [@pgen.1007227.ref083]--[@pgen.1007227.ref085]\]. This compares roughly to the 20% cancer incidence in mice exposed to dox. Overall, the number of invasive tumors and mice with multifocal tumors was not significantly different between the two groups (**[Fig 5C](#pgen.1007227.g005){ref-type="fig"}**); however, survival of Dek overexpressing mice was less than 60% while all dox-treated mice survived until sacrifice at week 45 (p = 0.11; **[Fig 5E](#pgen.1007227.g005){ref-type="fig"}**). Taken together, these data provide evidence that Dek overexpression promotes esophageal squamous cell carcinoma growth. ![Dek overexpression in 4NQO-treated mice leads to trends in increased cellular proliferation.\ (**A**) The water soluble carcinogen 4-nitroquinoline 1-oxide (4NQO) confers susceptibility to head and neck squamous cell carcinoma (HNSCC). At six weeks of age, mice were given 4NQO drinking water for 16 weeks in the presence or absence of dox chow. After 16 weeks, the mice were returned to normal water and were monitored until death or sacrifice at 45 weeks of age, or earlier if moribund. Prior to sacrifice, mice were injected with BrdU to measure proliferative differences in tissues. Abbreviations: time (T), weeks (wks). (**B-C**) Quantification of BrdU positive cells per millimeter of squamous epithelium in (**B**) normal and (**C**) 4NQO treated tongue and esophagus of *Bi-L-Dek_K5-tTA* mice in the absence or presence of dox. Three to six mice were examined for each tissue in the no dox/dox treatment groups and a student's t-test was used to determine statistical significance. A minimum of 100mm of epithelium was quantified per mouse and epithelial distance was measured using ImageJ \[[@pgen.1007227.ref089]\]. (**D**) Representative IHC images of increased BrdU incorporation in the tongue and esophagus of *Bi-L-Dek_K5-tTA* mice treated with 4NQO compared to those on dox.](pgen.1007227.g004){#pgen.1007227.g004} {#pgen.1007227.g005} Esophagi and tongues from all mice were microscopically examined to define tumor phenotypes and quantify microscopic lesions. Histological analysis confirmed that gross tumors in Dek overexpressing mice were squamous cell carcinomas with stromal invasion confirmed histologically in 67% of the mice (**[Fig 5A, 5C and 5H](#pgen.1007227.g005){ref-type="fig"}**). Additional multifocal microscopic squamous cell lesions were detected in 50% of Dek overexpressing mice with all mice developing 1--3 squamous cell lesions including at least one grossly apparent tumor. In contrast, microscopic lesions predominated in dox treated mice, with one mouse harboring no lesions and the other mice harboring 1--2 squamous cell lesions including a single grossly apparent tumor (**[Fig 5A, 5G and 5I](#pgen.1007227.g005){ref-type="fig"}**). The single tumor apparent at necropsy in this group was a well differentiated squamous cell carcinoma with abundant keratin production which differed from the moderate to poorly differentiated squamous cell carcinomas that predominated in Dek overexpressing mice (**[Fig 5I and 5H](#pgen.1007227.g005){ref-type="fig"}**). Microscopic tumors in dox treated mice consisted primarily of papillary squamous cell lesions with a single focus of very superficial invasion in one lesion. This differed from the more extensive invasion and necrosis in tumors that arose in Dek overexpressing mice (**[Fig 5I and 5H](#pgen.1007227.g005){ref-type="fig"}**). Dek levels appeared to be high in all tumors regardless of whether these originated in the Dek overexpressing group or the dox control group, thus suggesting strong selection for the upregulation of endogenous Dek during tumorigenesis (**[Fig 5F and 5G](#pgen.1007227.g005){ref-type="fig"}**, bottom row). With regards to the one tumor that arose in the dox control group, endogenous upregulation or leaky transgenic expression of Dek could be responsible. No tongue tumors were identified in either group of mice. Taken together, we demonstrate for the first time that Dek overexpression promotes the growth of esophageal SCC *in vivo*. Discussion {#sec007} ========== A number of studies have linked DEK overexpression in various malignancies to cellular growth, motility/invasion and chemoresistance \[[@pgen.1007227.ref016], [@pgen.1007227.ref036], [@pgen.1007227.ref037], [@pgen.1007227.ref043], [@pgen.1007227.ref058], [@pgen.1007227.ref071], [@pgen.1007227.ref077]\]. Relevant mechanisms have not been fully elucidated in each case. However, DEK loss has been shown to attenuate proliferation and survival, while inducing senescence or apoptosis, depending upon the cell type and model system studied. Required signaling pathways included those controlled by p53 and ΔNp63 to inhibit apoptosis and promote proliferation, respectively \[[@pgen.1007227.ref017], [@pgen.1007227.ref039], [@pgen.1007227.ref058]\], Wnt/beta-catenin to drive invasion and cellular proliferation \[[@pgen.1007227.ref032]\], VEGF to foster angiogenesis \[[@pgen.1007227.ref007]\], Rho/ROCK/MLC to support migration \[[@pgen.1007227.ref086]\], and NFkB to regulate cellular survival and growth \[[@pgen.1007227.ref006], [@pgen.1007227.ref008], [@pgen.1007227.ref059]\]. One caveat regarding cancer-related interpretation of these results is that many of the experiments are based upon DEK loss of function, and thus only address the requirement for DEK in tumor cell growth and not the contribution of DEK overexpression to tumor growth. For instance, Dek knockout mice are viable and resistant to chemically induced papillomas and HPV E7 driven HNSCC. *In vitro*, DEK overexpression in primary keratinocytes extends life span, stimulates transforming activities of classical oncogenes, and de-regulates cellular metabolism \[[@pgen.1007227.ref016], [@pgen.1007227.ref017], [@pgen.1007227.ref078]\]. These data are in line with, but do not prove, oncogenic activities that promote cancer development at the organismal level. Here we demonstrate that Dek overexpression targeted to the epithelium stimulates proliferation specifically in the presence of 4NQO in the tongue and also in the esophagus. Furthermore, concurrent Dek overexpression and 4NQO exposure increased the incidence of gross esophageal tumors demonstrating for the first time that Dek overexpression contributes to ESCC tumor growth *in vivo*. The observed increase in hyperplasia in the tongue is similar to that seen with sequential 4NQO exposure followed by ubiquitous Dek overexpression reported by Nakashima et. al.(72). Interestingly, and in contrast to our data, Nakashima et. al. reported that Dek overexpression decreased the volume of resulting tongue tumors \[[@pgen.1007227.ref072]\]. Key differences in the experimental designs between the two models likely account for the observed differences in tumor location and size in. Specifically, in the current study: 1) Dek overexpression was targeted to the basal epithelium as opposed to ubiquitous Dek overexpression including immune and stromal cells that modulate cancer cell growth, 2) 4NQO and Dek overexpression were concurrently administered rather than sequential exposure to 4NQO followed by Dek overexpression, 3) 4NQO exposure duration and dosage was 16 weeks at 10 μg/ml compared to 28 weeks at 20 μg/ml, 4) Dek overexpression duration was 52 weeks compared to four weeks, 5) exogenous Dek was unmodified and localized to the nucleusas compared to FLAG-tagged exogenous Dek protein localized predominantly to the cytoplasm, and 6) FVB/N mice were used compared to C57BL/6 mice. Interestingly, C57BL/6 and FVB/N harbor variations in immune phenotype, raising the intriguing possibility that immune surveillance and/or evasion account at least in part for the differing tumor phenotypes in mice with ubiquitous versus epithelial cell targeted Dek overexpression. Preliminary studies in the esophageal tumors in the current model did not reveal a significant CD3 positive T cell infiltrate by immunohistochemistry. Additional studies are needed to definitely determine the role of inflammatory cells in Dek dependent tumorigenesis, however, the lack of a prominent T-cell infiltrate suggests that differences in tumor growth in the two models cannot be simply explained by tumor infiltrating T cells acquiring an exhausted T-cell phenotype. The availability of these distinct complementary mouse models now provide a valuable system to identify cell specific functions that drive Dek induced carcinogenesis. Distinct effects of global versus tissue-specific Dek expression might reflect interesting cell-type specific functions of Dek in the tumor microenvironment including immune cells, or systemic effects on epidermal proliferation and tumor growth. The complexity of DEK functions in vivo is exemplified in studies of non-vertebrate organisms For instance, in Arabidopsis, DEK3 overexpression decreases germination efficiency under high salinity conditions, and conversely, plants deficient in DEK3 germinated significantly better compared to wild-type plants suggesting DEK3 levels are crucial for stress tolerance \[[@pgen.1007227.ref087]\]. The overexpression of human DEK in the Drosophila eye caused a rough-eye phenotype due to caspase-9 and 3-mediated apoptosis suggesting that DEK overexpression caused (rather than diminished) apoptosis \[[@pgen.1007227.ref088]\]. These non-vertebrate eukaryote model systems highlight the need for balanced DEK expression and its versatile functions *in vivo*. In the *Bi-L-Dek_K5-tTA* mouse model, Dek overexpression at the message and protein level was approximately 2--4 fold over that of endogenous Dek. This relatively modest level is in agreement with other published studies suggesting DEK expression levels are tightly regulated \[[@pgen.1007227.ref001], [@pgen.1007227.ref016], [@pgen.1007227.ref058], [@pgen.1007227.ref077]--[@pgen.1007227.ref079]\]. Achieving strong overexpression of DEK *in vitro* in our hands has been notoriously difficult, potentially due to toxicity and cell death, e. g. in the above Drosophila study \[[@pgen.1007227.ref088]\]. Importantly, a modest level of DEK overexpression in epithelial cells has been linked to oncogenic phenotypes *in vitro*. These DEK dependent oncogenic activities include enhanced cancer stem cell growth, colony formation, cellular invasion, mitotic abnormalities, and metabolic de-regulation, providing evidence that subtle increases in DEK protein expression are sufficient to elicit significant cellular consequences \[[@pgen.1007227.ref016], [@pgen.1007227.ref077]--[@pgen.1007227.ref079]\]. In human ESCC, HNSCC, breast, bladder, colorectal, hepatocellular, and non-small cell lung carcinoma, DEK protein levels were increased in tumor versus adjacent normal tissue, and the extent of overexpression was variable. Per cell DEK protein detection in various tumor types can range from intense to weak staining by IHC, and overexpression by western blot analysis can range from 2--30 fold \[[@pgen.1007227.ref023], [@pgen.1007227.ref025], [@pgen.1007227.ref026], [@pgen.1007227.ref031], [@pgen.1007227.ref034], [@pgen.1007227.ref039], [@pgen.1007227.ref042], [@pgen.1007227.ref077], [@pgen.1007227.ref080], [@pgen.1007227.ref081]\]. Overall, this patient data suggests that high levels of DEK can be tolerated by some human tumor cells, and that even modest DEK expression is associated with cancer growth and/or maintenance. In conclusion, *Bi-L-Dek_K5-tTA* mice subjected to 4NQO harbor trends toward increased cellular proliferation in the tongue and esophagus (**[Fig 4B--4D](#pgen.1007227.g004){ref-type="fig"}**) and a significantly increased incidence of gross esophageal tumors (**[Fig 5C](#pgen.1007227.g005){ref-type="fig"}**). Tongue tumors were not detected in these same mice. Importantly, control *Bi-L-Dek_K5-tTA* mice on dox nonetheless developed microscopic ESCC tumors, thus suggesting that Dek overexpression does not stimulate tumor initiation, but promotes tumor growth in the esophagus. This is in alignment with previously published Dek loss of function data from HNSCC-prone K14E7 transgenic mice wherein keratinocyte proliferation and tumor growth, but not the presence of microtumors, were diminished in the absence of Dek \[[@pgen.1007227.ref039]\]. While an abundance of data has suggested that DEK promotes tumor growth in the presence of oncogenic stimuli, the above experiments do not unequivocally rule out a role for Dek in tumor initiation. Overall larger tumors in the Dek overexpressing mice may be due to increased growth of tumors once initiated, or due to premature initiation and thus extended time for growth. In either case, Dek overexpression significantly increased the incidence of gross tumors and over 40% of *Bi-L-Dek_K5-tTA* mice died prior to the 45 week end point, while all mice in the dox treated control group survived. A plethora of Dek knockdown experiments have shown the importance of DEK expression for cancer cell growth and survival \[[@pgen.1007227.ref010], [@pgen.1007227.ref016], [@pgen.1007227.ref039], [@pgen.1007227.ref058], [@pgen.1007227.ref077]\]. These data, in conjunction with evidence that transformed keratinocytes are more sensitive to DEK loss when compared to their normal or differentiated counterparts \[[@pgen.1007227.ref016]\], make DEK an attractive therapeutic target. Furthermore, Dek knockout mice are healthy and fertile, suggesting potential feasibility and relative safety for the targeting of DEK in cancer. However, no DEK inhibitors exist commercially nor have been published. Thus, the inducible targeting of Dek in *Bi-L-Dek* mice harboring ESCC tumors should now be an attractive model to interrogate the requirement of continued Dek expression for cancer maintenance and progression. Taken together, we have generated and validated a new mouse model of esophageal transformation using an inducible *Bi-L-Dek* transgene which is now available for broader studies of Dek in health and disease of the intact organism. Materials and methods {#sec008} ===================== Generation of *Bi-L-Dek* transgenic mice {#sec009} ---------------------------------------- Murine Dek (mDek) DNA sequences were excised from the previously published R780 retroviral vector, using the restriction enzymes Sal I and Not I \[[@pgen.1007227.ref016], [@pgen.1007227.ref071]\], and cloned into the pBi-L plasmid (Clontech, Mountainview, CA Catalog No. 631005; GenBank Accession No.: U89934.) cleaved with the same restriction enzymes. The resulting *pBi-L-Dek* construct harbors the bi-directional Pbi-1 promoter which is responsive to the tTA regulatory protein in this Tet-Off system. The Tet-responsive element (TRE) consists of seven copies of the 42-bp tet operator sequence (tetO), and is located between two minimal CMV promoters that lack the CMV enhancer. Gene expression is silent in the absence of the tTA bound to tetO sequences and is silenced with the addition of doxycycline. The p*Bi-L-Dek* transgene sequences were liberated using the restriction enzymes AatII and AselI. A 5247bp (*Bi-L-Dek*) DNA sequence was purified and microinjected into the pronucleus of a fertilized egg and inserted into a pseudo-pregnant mouse to produce *Bi-L-Dek* founders. Transgene transmission was validated, and pups from the F1 generation were mated with *K5-tTA* mice. Resulting F2 *Bi-L-Dek_K5-tTA* mice were further characterized. Four *Bi-L-Dek* founders were generated. One founder line never produced offspring. Another founder died before producing a pup that harbored the transgene. Of the two remaining lines, both overexpressed Dek but founder \#317 was a better breeder. The murine Dek sequence that was cloned into the pBi-L-Tet vector is: 5'-ATGTCGGCGGCGGCGGCCCCCGCTGCGGAGGGAGAGGACGCCCCCGTGCCGCCC TCATCCGAGAAGGAACCCGAGATGCCGGGTCCCAGGGAAGAGAGTGAGGAGGAGGAGGAGGATGACGAAGACGATGATGAAGAGGACGAGGAGGAAGAAAAAGAAAAGAGTCTTATCGTGGAAGGCAAGAGAGAGAAGAAGAAAGTAGAGAGACTGACGATGCAAGTGTCTTCCTTACAGAGAGAGCCATTTACAGTGACACAAGGGAAGGGTCAGAAACTTTGTGAAATTGAAAGGATACATTTCTTTCTGAGTAAGAAAAAACCAGATGAACTTAGAAATCTACACAAACTGCTTTACAACAGGCCGGGCACAGTGTCCTCGTTGAAGAAGAACGTGGGTCAGTTCAGTGGCTTTCCATTCGAAAAAGGCAGTACCCAGTATAAAAAGAAGGAAGAAATGTTGAAAAAGTTTCGAAATGCCATGTTAAAGAGCATCTGTGAGGTTCTTGATTTAGAGAGGTCAGGCGTGAACAGCGAACTCGTGAAGAGGATCTTGAACTTCTTAATGCATCCAAAGCCTTCTGGCAAACCATTACCAAAGTCCAAAAAATCTTCCAGCAAAGGTAGTAAAAAGGAACGGAACAGTTCTGGAACAACAAGGAAGTCAAAGCAAACTAAATGCCCTGAAATTCTGTCAGATGAGTCTAGTAGTGATGAAGATGAGAAGAAAAATAAGGAAGAGTCTTCGGAAGATGAAGAGAAAGAAAGTGAAGAGGAGCAACCACCAAAAAAGACATCTAAAAAAGAAAAAGCAAAACAGAAAGCTACTGCTAAAAGTAAAAAATCTGTGAAGAGTGCTAATGTTAAGAAGGCAGACAGCAGTACCACCAAGAAGAATCAAAAAAGTTCCAAAAAAGAGTCTGAATCCGAAGACAGTTCTGATGATGAACCCTTAATTAAAAAATTGAAAAAGCCACCTACAGATGAAGAGCTAAAGGAAACAGTGAAGAAATTACTGGCTGATGCTAACTTGGAAGAAGTCACAATGAAGCAGATTTGCAAAGAGGTATATGAAAATTATCCTGCTTATGATTTGACTGAGAGGAAAGATTTCATTAAAACAACTGTAAAAGAGCTAATTTCTTGA-3' Genetic mouse models {#sec010} -------------------- *K5-tTA* mice were obtained internally at CCHMC and have previously been published \[[@pgen.1007227.ref090]\]. Dek knockout mice (*Dek-/-*) have previously been published \[[@pgen.1007227.ref016]\]. *Dlx5/6-tTA* mice were obtained internally at CCHMC and were generated in Dr. Kenneth Campbell's lab by Lisa Ehrman. *Dlx5/6-tTA* mice have been analyzed for tTA expression, and will be fully described and characterized in a separate publication. *Dlx5/6* tTA expression is similar to Cre expression in the reported *Dlx5/6-*Cre-IRES-EGFP *(CIE)* transgenic mouse model \[[@pgen.1007227.ref082]\]. *E2A-Cre* mice were obtained from Jackson Laboratory and are strain number 003724. The Cre transgene is under the control of the adenovirus EIIa promoter, which targets expression of Cre recombinase to the early mouse embryo. This model is useful for deletions, in the germ line, of *loxP*-flanked genes. *E2A-Cre* mice were bred to *Rosa-LNL-tTA* transgenic mice. These mice were also obtained from Jackson laboratories and are strain number 008600. Rosa-LNL-tTA mice contain a loxP-flanked nonsense sequence inhibiting expression of tTA that is removed once exposed to E2A controlled Cre. Genotyping {#sec011} ---------- Ear clips were digested with 25mM NaOH in 0.2mM EDTA at a pH of 12 and incubated at 95°C for 20 minutes. The reaction was neutralized with 40mM Tris-HCl. For PCR analysis, one ul of the digest with DNA was added to JumpStart Taq Ready Mix from Invitrogen (Carlsbad, CA, product \# P2893) using the manufacturer's specifications. Transgenes were detected with the following primers: *Bi-L-Dek*: Forward: GAAATGTCCGTTCGGTTGGCAGAAGC; Reverse: CCAAAACCGTGATGGAATGGAACAACA. *K5-tTA*: Forward: GCTGCTTAATGAGGTCGG Reverse: CTCTGCACCTTGGTGATC. *Bi-L-Dek* primers that do not detect endogenous Dek (exogenous Dek cDNA primers) Forward: CAGTGACACAAGGGAAGGGTCAGA Reverse: AGCCACTGAACTGACCCACGT. *Bi-L-Dek* copy number determination by qPCR {#sec012} -------------------------------------------- Genomic DNA was isolated from the tails of mice from successive generations of offspring from founder 317. A minimum of two mice were used per generation and analyzed as replicates. The DNA concentration was adjusted to 20ng/ul in each case, and 60ng of DNA was used for qPCR per sample and performed in duplicate. Primers were used to quantify the beta actin gene and a region in exon 6 of the Dek gene. This region is present in Dek+/+ mice but absent in *Dek-/-* mice thus allowing for a negative control. The following sequences were used: Beta actin forward: GATATCGCTGCGCTGGTCGTC Beta actin reverse: ACCATCACACCCTGGTGCCTAG Dek Exon 6 forward: AGGTCAGGCGTGAACAGCGA Dek Exon 6 reverse: TGCCAGAAGGCTTTGGATGCATTA The critical threshold (CT) values for Dek exon 6 primers were normalized to actin, and quantified relative to Dek wild type mice using the delta delta CT method. Values were multiplied by two to account for the two endogenous Dek alleles in WT mice and the number of *Bi-L-Dek* transgene insertions was determined. Error bars represent multiple mice from the same generation. *In vivo* Imaging Systems (IVIS) {#sec013} -------------------------------- Mice were injected with 15ng of luciferin per gram in body weight, and allowed to metabolize the luciferin for five minutes prior to sedation with isoflurane. Mice were imaged in the Perkin Elmer IVIS Spectrum CT, Waltham, Massaschusetts, USA. For *ex vivo* IVIS, mice were allowed to metabolize luciferin for eight minutes following luciferin injection, and then sacrificed with CO2. The mice were then dissected and tissues placed in PBS containing 300ug/mL of luciferin, kept on ice, and protected from light before immediate analysis by IVIS. Mouse keratinocyte culture {#sec014} -------------------------- For validation of p*Bi-L-Dek* expression, the plasmid was transfected into previously isolated and cultured *K5-tTA* expressing murine keratinocytes \[[@pgen.1007227.ref091]\]. Cells were collected for Dek protein expression by western blot analysis. *K5-tTA* keratinocytes were grown in E-media supplemented with 0.05 mM Ca2 and 15% serum as previously published \[[@pgen.1007227.ref092]\]. Keratinocytes were isolated from *Bi-L-Dek_K5-tTA* mice and single transgenic littermate controls using a previously published protocol with modifications \[[@pgen.1007227.ref093]\]. Briefly, pups were euthanized within 48 hours of birth, rinsed in 70% ethanol, and placed in PBS. Flank skin was removed, and placed dermis side down in 1 mL of dispase (Dispase Gibco/Invitrogen, Calsbad, CA, USA, product\# 17105--041) and 1 mL of DMEM (1:1 mixture) in a 35mm plate, and incubated overnight at 4° Celsius. The epidermis was removed and placed in 1 mL of accutase (Sigma, St. Louis, MO, USA, product \# A6964) for 20 minutes with agitation to release the keratinocytes. Cells were collected and centrifuged, then plated on irradiated MEFs and overlaid with CnT07 media (CellnTec, Bern, Switzerland). Cells were used for experiments in passage 0 or 1. Immunofluorescence microscopy {#sec015} ----------------------------- Keratinocytes were plated onto 100 mg/ml poly-D-lysine coated coverslips, and fixed with 2% paraformaldehyde for 30 minutes. Coverslips were incubated in 0.1% Triton X-100 for three minutes, blocked with 5% normal goat serum, and incubated with primary antibody for one hour at 37°C. Antibody dilutions were as follows: DEK-antibody (Cusabio, Baltimore, MD, USA) 1:300 dilution; keratin 5 antibody (Acris, San Diego, CA, USA) 1:500; and sealed with a coverslip using Vectashield with DAPI (Vector Laboratories, Burlingame, CA). ImageJ (National Institutes of Health, Bethesda, Maryland, USA) \[[@pgen.1007227.ref089]\] was used to quantify Dek staining. Dek immunofluorescences (IF) images were converted to 8-bit images, followed by the identification of the location of cells with the nucleus counter ImageJ plugin. Each cell was visually validated and added to the regions of interest (ROI). The mean grayscale intensity was measured in these ROIs. Quantification was from successive images to encompass the entire coverslip of keratinocytes isolated from each genotype with or without dox treatment. Western blot analyses {#sec016} --------------------- Tissues were lysed using mortar and pestle, resuspended in RIPA buffer (1% Triton, 1% deoxycholate, 0.1% SDS, 0.16M NaCl, 10 mmol/L Tris pH 7.4, and 5 mmol/L EDTA), supplemented with a protease inhibitor cocktail (Pharmingen, San Diego, CA, USA), and analyzed as described previously (48). Primary antibodies used for DEK were as follows: DEK (1:1000; BD Biosciences, San Diego, CA, USA), pan-actin (1:20,000; a gift from James Lessard). Membranes were exposed to enhanced chemiluminescence reagents (Perkin Elmer, Boston, MA, USA) and imaged using the BioRad Chemidoc (Hercules, CA, USA). 4NQO induction of HNSCC development *in vivo* {#sec017} --------------------------------------------- All mice were maintained in a hemizygous state for the *Bi-L-Dek* and *K5-tTA* transgenes. All *Bi-L-Dek* mice were F3 and F4 generations from founder 317. *Bi-L-Dek* mice were bred to *K5-tTA* mice and bi-transgenic offspring were given 4NQO water for 16 weeks at a dose of 10mg/ml starting at six weeks of age. Mice on doxycycline were continuously fed dox chow from the start of 4NQO treatment until sacrifice. After 16 weeks on 4NQO, mice were returned to normal water until sacrifice at week 45 or when determined excessively morbid by veterinary services thus warranting sacrifice. At the time of sacrifice, tumors were resected and counted, localization was noted, and tumors were measured by calipers. Tumor volume was measured by (length x width x depth). All statistical analyses were performed in GraphPad Prism. The survival curve was analyzed using the log-rank (Mantel-Cox) test. Tumor incidence was determined significant/non-significant using the Chi Square (and Fisher's exact) test. Histological analyses and immunohistochemistry {#sec018} ---------------------------------------------- Mouse tumors and tissues were fixed in 4% paraformaldehyde, embedded in paraffin, sectioned at 5 μm thickness, and fixed onto slides. Routine H&E stained sections were analyzed for histopathology.[^13^](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4160430/#R13) The area of microscopic tumors was determined by multiplying the widest part of the tumor by the longest part that was observed in the sections. Paraffin sections were deparaffinized in xylene and rehydrated for antigen retrieval in sodium citrate. Sections were then treated with the Mouse on Mouse peroxidase immunostaining kit (Vector Labs, Burlingame, CA, USA). Sections were stained with diaminobenzidine (DAB) and counterstained with Nuclear Fast Red (Poly Scientific, Bay Shore, NY, USA) and mounted with Permount (Fisher Scientific, Pittsburgh, PA, USA). Images were captured at the indicated magnifications and antibodies used are noted in each case. Antibody dilutions were used as follows: BrdU (1:100, Invitrogen, Calsbad, CA, USA), and DEK (1:200, BD Biosciences, San Jose, CA, USA; or 1:300, Proteintech Group, Chicago, IL, USA; or 1:50, Cusabio, Baltimore, MD, USA). BrdU quantification {#sec019} ------------------- 10x or 20x magnified images of BrdU stained tongue or esophagus were analyzed for BrdU positive cells using ImageJ (National Institutes of Health, Bethesda, Maryland, USA). In ImageJ, the bottom portion of the basal cell layer of the stratified squamous epithelium was traced using the freehand tool and measured in the indicated tissue. The distance was converted into millimeters using scale bars based on magnification to determine BrdU positive cells per millimeter of epithelium. Statistical analysis was performed using GraphPad Prism with t-tests and the two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli. Luciferase experiments {#sec020} ---------------------- Luciferase assays were performed using the Dual-Luciferase Reporter Assay System from Promega and following manufacturer specifications. Ethics statement {#sec021} ---------------- All animal work was conducted according to Cincinnati Children\'s Hospital Medical Center Institutional Animal Care and Use Committee guidelines under protocol number \#2017--0004. To ameliorate animal suffering mice were euthanized with carbon dioxide when moribund as determined by veterinary services. Supporting information {#sec022} ====================== ###### Generation of tetracycline responsive *Dek* transgenic mice. (**A**) Murine Dek cDNA was cloned into the pBi-L-tet plasmid (Clontech, Mountain View, CA, USA) as described in the Materials and Methods. Therein, Dek and luciferase expression are under control of a tetracycline response element (TRE). Restriction enzyme sites for AatII and AseI were used to excise the transgenic construct. (**B**) Western blot analysis for Dek expression in mouse keratinocytes isolated from a *K5-tTA* mouse (84) and transfected with the p*Bi-L-Dek* plasmid. Dek protein expression was repressed with 1ug/ml of dox in the media. (**C**) Luciferase assay with keratinocytes from B, treated with 0, 0.33 or 1.5 ug/ml of dox show dose dependent luciferase repression. (**D**) Agarose gel electrophoresis of the p*Bi-L-Dek* plasmid linearized with restriction enzymes AseI and AatII and the resulting 5247bp *Bi-L-Dek* band. (**E**) The 5247bp *Bi-L-Dek* band was isolated from the gel, purified, and used for micronuclear injection into the pronucleus of FVB/N fertilized eggs to generate *Bi-L-Dek* founders. (**F**) Four founders were generated as confirmed by genotyping with primers that detect luciferase and exogenous Dek cDNA sequences. Founder \#317 was used for subsequent experiments. (EPS) ###### Click here for additional data file. ###### Detection of Dek mRNA expression from the *Bi-L-Dek* transgene. RT-qPCR for quantification of relative Dek transcript levels in *Dek-/-*, *Dek+/+*, and *Dek-/-*\_*Bi-L-Dek_K5-tTA* (black bar) flank tissue (n = 1 mouse per genotype). Primers detect a cDNA region present in endogenous and *Bi-L-Dek* mice that is absent in the *Dek-/-* mice (Dek exon 6 primer set in methods and materials). (EPS) ###### Click here for additional data file. ###### *Bi-L-Dek* transgene expression targeted to the brain. (**A**) *Dlx5/6-tTA* mice were crossed with the *Bi-L-Dek* mice. The *Dlx5/6* enhancer sequences drive tTA expression in neurons that originate in the ventral forebrain (**B**) Genotyping confirmed transgene transmission to offspring near expected ratios. Lanes 1--6 represent offspring of the cross from A. Lane 1 represents a bi-transgenic mouse that was used for Dek expression studies in C. (**C**) IHC for Dek in the brains of pups at embryonic day 18 shows Dek overexpression in the *Dlx5/6-tTA_Bi-L-Dek* mouse cortex and striatum. Insets are magnified images of the cortex and striatum where Dek is overexpressed (Dek antibody: Proteintech Group, Chicago, IL, USA). (EPS) ###### Click here for additional data file. ###### Doxycycline responsive, whole body *Bi-L-Dek* transgene expression. (**A**) To generate mice with whole body Dek overexpression, we bred *Bi-L-Dek* mice to *Rosa-tTA* mice. To generate *Rosa-tTA* mice, *E2A-Cre* mice were bred to Rosa-LNL-tTA mice. *E2A-Cre* mice carry a Cre transgene under the control of the adenovirus EIIa promoter that targets expression of Cre recombinase to the early mouse embryo. *Rosa-LNL-tTA* mice harbor a *tTA* transgene in the Rosa 26 locus that is preceded by a stop codon flanked by *loxP*- sequences (LNL) which inhibits tTA translation. The *E2A-Cre* mice were used for germ line deletion of the *LNL* resulting in *Rosa-tTA* mice with global expression of tTA. (**B**) Genotyping of the offspring shows excision of the *LNL* and transmission of the *Rosa-tTA* and *Bi-L-Dek* transgenes. Mouse 1 and 3 have both *Bi-L-Dek* and *Rosa-tTA* transgenes as indicated by red arrows. (**C**) *Bi-L-Dek_Rosa-tTA* mice 1 and 3 along with a single transgenic littermate were subjected to IVIS at six weeks of age for luciferase detection. (**D**) Mice were placed on dox for two weeks and subjected to IVIS again to show repression of luciferase expression from the *Bi-L-Dek* transgene in *Bi-L-Dek_Rosa-tTA* bi-transgenic mice. (EPS) ###### Click here for additional data file. ###### DEK is overexpressed in esophageal cancer. \(A\) TCGA data analyzed for DEK expression in esophageal squamous cell carcinomas (ESCC) versus normal tissue and (B) in adenoma compared to squamous cell carcinoma (SCC). Analysis was performed using RNAseq (V2) and clinical (histological subtype) data for esophageal carcinoma (ESCA) and was downloaded from GDAC Firehose. (EPS) ###### Click here for additional data file. We thank Dr. Paul F. Lambert and the transgenic animal core at CCHMC for help with the design of the *Bi-L-Dek* mice. We thank Dr. Kenneth Campbell at CCHMC for the *Dlx5/6-tA* mice and Diana Nardini for her help in validating the *Dlx5/6-tTA_Bi-L-Dek* model. We appreciate instructive discussions and critical feedback by Dr. Lisa Privette Vinnedge. We thank Dr. James Lessard for the monoclonal actin antiserum, and Dr. Trisha Wise-Draper and Sarah Marrochello for their technical help and thoughtful discussions. [^1]: The authors have declared that no competing interests exist.

Introduction {#s1} ============ Acute myeloid leukaemia (AML), one of the most common and deadliest forms of proliferative neoplasms, is established through a stepwise acquisition of genetic and epigenetic alterations that result in the malignant transformation of haematopoietic progenitor cells ([@bib15]; [@bib21]). Often, AML arises through the collaboration between mutations affecting transcription factors (e.g., CEBPA, PU.1, and RUNX1) and signalling proteins (such as FLT3, RAS, and KIT) that lead to an aberrant proliferation capacity coupled with a disruption of terminal myeloid differentiation ([@bib38]; [@bib30]). C/EBPα, a leucine zipper transcription factor with a known tumour suppressor function, has been demonstrated to play an important role in granulocytic development and in the maintenance of haematopoietic stem cell homeostasis ([@bib27], [@bib26]; [@bib47]; [@bib17]; [@bib43]; [@bib45]). C/EBPα is translated as two major isoforms, namely a full-length 42-kD form (p42) and a truncated 30-kD protein (p30) that arises from a downstream translational initiation codon ([@bib19]). Mutations in the *CEBPA* gene are frequently associated with leukaemia, being found in 8--14% of all de novo AML with normal karyotype ([@bib22]; [@bib18]; [@bib33]) and typically involve both alleles. C/EBPα-mutant proteins are classified into two major groups: (i) C-terminal insertions or deletions within the basic region leucine zipper DNA-binding domain; and (ii) N-terminal mutations that lead to the complete ablation of p42 while retaining normal p30 function ([@bib24]; [@bib18]; [@bib8]). Most patients carrying *CEBPA* mutations harbour one allele with an N-terminal mutation and one with a C-terminal mutation, with homozygosity for N- or C-terminal mutations being less common ([@bib10]; [@bib23]). Furthermore, several reports have demonstrated that biallelic mutations of *CEBPA* are associated with a favourable outcome, when not found in association with FLT3-activating mutations ([@bib28]; [@bib7]). Efforts aimed at understanding how mutations or oncoproteins may cooperate in driving the leukaemogenesis have pointed to cooperation between C/EBPα and other transcription factors, such as RUNX1, MYB, and PU.1. We have previously demonstrated the functional cooperation of Myb and C/EBPα in the regulation of the *Flt3* gene in both haematopoietic and leukaemia stem cells ([@bib41], [@bib40]). Our studies indicated that Myb and C/EBPα act cooperatively through their combined activity on promoter and intronic elements in the *Flt3* gene ([@bib41]). Furthermore, we reported a strong linear correlation between expression of the two transcription factors and *FLT3* RNA levels in human CN-AML, adding to an increasing body of evidence that points to MYB being a crucial component of leukaemia maintenance and oncogene addiction ([@bib13]; [@bib50]; [@bib6]). Our findings on the cooperation of Myb and C/EBPα in *Flt3* gene regulation prompted us to investigate the global extent of this cooperation in leukaemia and to determine how manipulation of Myb expression might impact on the maintenance of C/EBPα-driven leukaemia. To address this, we performed genetic manipulation studies in murine haematopoietic progenitor cell lines harbouring either wild-type C/EBPα or the most frequently occurring combinations of biallelic CEBPA mutations, that is N^ter^/N^ter^ or N^ter^/C^ter^ to determine the biological and molecular consequences of reduced Myb activity on the leukaemia driven by those mutations. Here, we show that reducing Myb activity can override the differentiation barrier, although the dependency on *Myb* expression generally observed in leukaemia is minimal in the presence of CEBPA biallelic N-terminal mutations. Materials and Methods {#s2} ===================== Cell lines {#s3} ---------- Cells were cultured in RPMI medium supplemented with 10% fetal bovine serum, 50U/ml penicillin, 50 μg/ml streptomycin, and 2 mM [l]{.smallcaps}-glutamine. The culture of FMH9 cells ([@bib41]) were supplemented with 50 ng/ml stem cell factor, 5 ng/ml GM-CSF, 5 ng/ml IL-3 (IL3), and 5 ng IL-6 (IL6), whereas KL cells (GV, JF, and FG, unpublished) and LL cells ([@bib12]; [@bib32]) required 2 ng/ml IL3. Both KL and LL have been established by serial replating of E14.5 foetal liver cells obtained from mice homozygous for the Lp30 allele ([@bib16]) or carrying both Lp30 and K313KK alleles ([@bib3]). Briefly, the KL and LL cell lines were obtained by performing six rounds of replating in M3434 semisolid medium (Stem Cell Technologies Inc) followed by an initial liquid culture for 4 wk in the presence of stem cell factor (50 ng/ml), IL3 (2 ng/ml), and IL6 (2 ng/ml). After this initial period, the cells were transferred into the culture medium described above. All cytokines were purchased from Peprotech EC. Transfection experiments, cell viability, proliferation, apoptosis, and differentiation assays {#s4} ---------------------------------------------------------------------------------------------- In total, 5 × 10^6^ FMH9, KL, or LL cells were electroporated with 300 mM of *Myb* siRNA (s70212, Ambion; Life Technologies) or a scrambled negative control siRNA (4390843 Silencer Select Negative Control \#1; Life Technologies) using an Amaxa 4D-nucleofector with solution SF Cell Line (V4XC-2024; Lonza) and program EO-100 for FMH9 cells or solution P3 Primary Cells (V4XP-3024; Lonza) and program DS-120 for KL and LL cells. After transfection, the cells were plated at a density of 10^6^ cells/ml and viable cells counted and passaged at a ratio of 1:2 every 24 h for 4 consecutive days. Cell cycle analysis was performed by labelling transfected cells (48 h post-nucleofection) with 10 μM BrdU for 1 h. Cells were co-stained with 7-AAD (A9400-1mg; Sigma-Aldrich) and BrdU using the BrdU flow kit (8811-6600; BD Bioscience) according to the manufacturer's instructions, as previously described ([@bib2]). Apoptosis analysis was performed using the Annexin V kit (eBioscience) as previously described ([@bib40]). The percentage of apoptotic cells was obtained by performing live cells gating. Proliferation analysis was performed using CellTrace carboxyfluorescein succinimidyl ester (CFSE) Cell Proliferation Kit (C34554; Thermofisher Scientific). Assessment of differentiation following *Myb* knockdown was achieved by flow cytometry/immunofluorescence staining of the cells with anti-CD11b PE-Cy7 (25-0112-81; eBioscience) anti-Gr-1 APC (14-5921-82; eBioscience), anti-CD135 PE (12-1351-81; eBioscience), and anti-CD117 PE-Cy5 (15-1171-82; eBioscience). Acquisition and analysis of flow cytometric data were performed using Cyan ADP with either Summit 4.4 software (Beckman Coulter) or FlowJo software (FlowJo, LLC). Quantitative reverse transcriptase polymerase chain reaction (RT-PCR) analysis {#s5} ------------------------------------------------------------------------------ 10^6^ cells from each line were harvested 24 h post-transfection. RNA was extracted using RNeasy Mini kit (QIAGEN), and first-strand cDNA synthesis was performed using standard protocols. Quantitative PCR reactions were performed using predesigned Taqman gene expression assays as previously described ([@bib40]). Statistical analysis {#s6} -------------------- Statistical significance was determined by performing *t* test for pairwise comparison, and the *P*-values are indicated where appropriate. Analysis of *MYB* expression in human patient array data presented in [Fig 1A](#fig1){ref-type="fig"} was performed using non-parametric Kruskal--Wallis test. All statistical analyses were performed using GraphPad Prism 7 (GraphPad Software Inc). ![*Myb* expression is required for the proliferation of CEBPA biallelic mutant cell lines.\ **(A)** Scatter plot depicting the abundance of *MYB* transcript in subgroups of patients from the [@bib39] dataset, characterised by the molecular abnormalities indicated in the graph. Statistical significance presented in this plot has been calculated using non-parametric the Kruskal--Wallis test. **(B)** Bar plot representing *Myb* mRNA quantification by quantitative RT-PCR in FMH9, KL, and LL cell lines, normalised against *B2m* house-keeping gene results. Statistical analysis was performed using *t* test (\*\**P* \< 0.01 and \**P* \< 0.05). **(C)** Quantitative RT-PCR of *Myb* transcript abundance in FMH9, KL, and LL cells 24 h post-transfection with *Myb* siRNA. Expression is normalised to *B2m* and standardised to the control samples. Error bars represent the SEM and numbers are plotted as mean ± SEM. Each plot is representative of six independent experiments (\*\*\**P* \< 0.001 and \**P* \< 0.05). **(D)** Bar plot indicating the cell viability and relative proliferation of FMH9, KL, and LL cells after *Myb* siRNA transfection relative to the corresponding negative control (\*\*\**P* \< 0.001 and \*\**P* \< 0.01, \**P* \< 0.05). This plot represents an average of six independent experiments. **(E)** Flow cytometric analysis of cellular proliferation by CFSE incorporation in FMH9, KL, and LL cells after *Myb* siRNA transfection relative to the corresponding negative control. Continuous lines indicate cells transfected with the negative control siRNA, whereas dashed lines indicate si*Myb*-transfected cells. The statistical analysis was performed using *t* test on the geometric means of fluorescence intensity for each time point comparing siNEG versus si*Myb*-treated cells, as indicated by the colour-matched bar on top of every peak (\*\**P* \< 0.01 and \**P* \< 0.05). Each histogram is representative of six independent experiments. **(F)** Flow cytometric analysis of the cell cycle in FMH9, KL, and LL cells was performed by staining with 7-AAD at 48 h after *Myb* siRNA transfection and compared with the negative control. Percentages of cells in G0/G1 are indicated in each histogram. Each plot is indicative of six independent experiments. **(G)** Representative bar plot showing apoptosis analysis performed by Annexin V staining 72 h post *Myb* siRNA transfection in FMH9, KL, and LL cells (\*\*\**P* \< 0.001 and \*\**P* \< 0.01). Each bar plot represents an average of six independent experiments.](LSA-2018-00207_Fig1){#fig1} Bioinformatic analyses {#s7} ---------------------- Details of the methodologies used to analyse and compare RNA-seq, microarray, and ChIP-seq data, including Gene Ontology (GO) and Gene Set Enrichment Analysis (GSEA) comparisons, can be found in the Supplementary Information. RNA-sequencing {#s8} -------------- For RNA-Seq, libraries were prepared using the Illumina TruSeq Stranded kit according to the manufacturer's instructions. Sequencing was performed at the Institute of Medical Biochemistry, University of Veterinary Medicine, Vienna, Austria, and in Genomics Birmingham, University of Birmingham, Birmingham, UK, on Illumina HiSeq 2500 and NextSeq 500 sequencers, respectively. Data availability {#s9} ----------------- RNA-Seq data generated in this study are available at the Gene Expression Omnibus under series [GSE119348](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE119348). Results {#s10} ======= High *MYB* expression is associated with *CEBPA* mutations in AML {#s11} ----------------------------------------------------------------- Previous studies have provided evidence for functional cooperation between C/EBPα and MYB in activating the expression of key genes for both haematopoietic and leukaemia stem cell functions in mouse and human ([@bib41], [@bib40], [@bib42]). Using publicly available AML patient profiling arrays ([@bib39]) and focussing on specific molecular abnormalities found in cytogenetically normal AML (CN-AML), we observed that *MYB* expression levels are highest in those patients carrying biallelic *CEBPA* mutations ([Fig 1A](#fig1){ref-type="fig"}). Given the lack of a suitable cellular system to investigate the relationship between MYB and C/EBPα in human AML, we used murine cell lines modelling the spectrum of *CEBPA* mutations. Bereshchenko and coworkers demonstrated that mutations in the C/EBPα protein efficiently drive leukaemia in vivo and that the combination of N- and C-terminal mutations were the most highly leukaemogenic, whereas biallelic C-terminal mutation resulted in the longest latency ([@bib16]; [@bib3]). We generated cell lines by performing serial replating in semisolid medium of E14.5 foetal liver cells carrying knock-in modifications mimicking either the N-terminal LP30 (L allele) or the C-terminal K313KK (K allele) *CEBPA* mutations ([@bib12]). LL cells were used to represent AML harbouring biallelic N-terminal *CEBPA* mutations, whereas cells carrying one K mutation and one L mutation (KL cells) provided a model of leukaemia with both N- and C-terminal *CEBPA* mutations. AML with wild-type C/EBPα expression was modelled using a previously characterised myelomonocytic leukaemia cell line, namely, FMH9 ([@bib41]), which was established by ectopic expression of HoxA9 and Meis1 in bone marrow haematopoietic progenitor cells. The phenotype of these cells lines was investigated by determining the surface expression of several myeloid markers (CD11b, Gr1, Kit, and Flt3) ([Fig S1A](#figS1){ref-type="fig"}). Importantly, mRNA quantification indicated that both FMH9 and LL cells displayed similar *Myb* mRNA levels, whereas KL cells exhibited a higher level of *Myb* expression ([Fig 1B](#fig1){ref-type="fig"}), thus being in agreement with the observations obtained from the patient array data. {ref-type="fig"}. Bar plot depicting the siNEG/si*Myb* proliferation ratio comparison between FMH9 control and CEBPA-mutant cell lines. These data represent an average of six independent experiments. Statistical analysis was performed using *t* test (\*\*\**P* \< 0.001, \*\**P* \< 0.01). **(C)** Related to [Fig 1E](#fig1){ref-type="fig"}. Bar plots representing the average of CFSE geometric mean intensities (as indicated by the bar on top of every peak in [Fig 1E](#fig1){ref-type="fig"}) of FMH9, KL, and LL cells and their relative changes upon si*My*b-mediated knockdown. Filled bars represent siNEG-treated cells, whereas empty bars indicated si*Myb*-treated cells. Geometric mean intensities have been calculated using FlowJo software and have been used to perform statistical analysis using the *t* test (\*\**P* \< 0.01, \**P* \< 0.05). These data represent an average of six independent experiments.](LSA-2018-00207_FigS1){#figS1} Manipulation of Myb expression does not reverse the differentiation block in cells carrying biallelic N-terminal *CEBPA* mutations {#s12} ---------------------------------------------------------------------------------------------------------------------------------- To investigate the requirement for Myb in the maintenance of CEBPA-driven leukaemia, we performed siRNA-mediated knockdown of *Myb*. Cells were transfected with siRNAs targeting either *Myb* or a scrambled negative control and were harvested after 24 h to determine the efficiency of knockdown. This analysis revealed a decrease in *Myb* transcripts by 60--80% in all cell lines ([Fig 1C](#fig1){ref-type="fig"}). To determine the biological consequences of *Myb* knockdown cells were cultured for up to 96 h and cell numbers determined daily. *Myb* down-regulation induced growth retardation in the three cell lines, regardless of the *CEBPA* mutational status; albeit those cells showing a similar pattern, the growth defect observed in FMH9 cells was significantly more pronounced in comparison with the *CEBPA* mutant cells lines ([Figs 1D](#fig1){ref-type="fig"} and [S1B](#figS1){ref-type="fig"}). Flow cytometric analysis of CSFE dye dilution revealed that *Myb* knockdown induced a significant proliferation defect in FMH9 cells, whereas both KL and LL cells were unaffected ([Figs 1E](#fig1){ref-type="fig"} and [S1C](#figS1){ref-type="fig"}). The pattern of the proliferation defect observed in FMH9 cells 72 h post *Myb* down-regulation appeared to be bimodal, probably because of cells being induced to differentiate or due to the transient effect of the siRNA-mediated knockdown. By combined staining with 7-AAD and BRDU 48 h post-transfection, we observed that *Myb* knockdown led to a significant increase in the proportion of cells in the G0/G1 phase with a concomitant decrease in both S and G2/M phases in FMH9, whereas KL and LL cells showed no alteration in their capacity to progress through the cell cycle ([Figs 1F](#fig1){ref-type="fig"} and [S2A, and B](#figS2){ref-type="fig"}). This analysis also revealed an increase in the percentage of cells with less than 2n DNA content in *CEBPA*-mutant cell lines only, this being indicative of cells undergoing apoptosis/necrosis ([Fig S2C](#figS2){ref-type="fig"}). To confirm this observation, we performed Annexin V staining. This revealed a significant increase in the extent of apoptotic/necrotic cells in both KL and LL cells, whereas no increase was observed in FMH9 cells ([Figs 1G](#fig1){ref-type="fig"} and [S3A](#figS3){ref-type="fig"}). {ref-type="fig"}.\ **(A)** Phenotypic two-dimensional dot-plots representing the percentage of FMH9, KL, and LL cells in the G0/G1, S, and G2/M phases of the cell cycles and their changes in response to *Myb* siRNA-mediated down-regulation. The percentage of each population is indicated in every histogram. **(B)** Bar plots depicting the changes in the percentages of FMH9, KL, and LL cells in the G0/G1, S, and G2/M phases of the cell cycle as shown in the top panel. These data represent an average of six independent experiments. Statistical analysis was performed using *t* test (\*\*\**P* \< 0.001, \*\**P* \< 0.01). **(C)** Representative histograms showing cell cycle profile changes in response to *My*b siRNA-mediated knockdown in FMH9, KL, and LL. The percentage of cells in G0/G1 and apoptotic/necrotic cells is indicated in each histogram. This panel is indicative of six independent experiments.](LSA-2018-00207_FigS2){#figS2} {ref-type="fig"}.\ **(A)** Two-dimensional dot-plots representing the percentage of Annexin V^+^ cells in FMH9, KL, and LL cells in response to *Myb* siRNA-mediated down-regulation. Each dot plot is representative of six independent experiments. **(B)** Bar plots showing quantitative PCR analysis of *Bcl2* and *Bim* expression in FMH9, KL, and LL cells in response to *Myb* knockdown in four independent experiments. Expression is normalised to *β2m* and standardised to the siNEG-treated samples for every cell line. Error bars represent the standard error or the mean. Statistical analysis was performed using *t* test (\**P* \< 0.05).](LSA-2018-00207_FigS3){#figS3} Because it is accepted that homozygous *CEBPA* mutations lead to a block in myeloid lineage commitment, we investigated how *Myb* knockdown affects the differentiation capacity of cells in the presence of either wild-type or biallelic mutant C/EBPα. At 96 h, both FMH9 and KL cells exhibited a clear induction of myeloid differentiation as indicated by increased expression of Gr-1 and CD11b ([Fig 2](#fig2){ref-type="fig"}). However, this phenomenon was not observed in LL cells, which is intriguing because it suggests that leukaemia cells carrying biallelic N-terminal *CEBPA* mutations have a reduced dependency on *Myb* expression in respect to differentiation control. {#fig2} Molecular consequences of *Myb* manipulation in cell lines carrying either wild-type or mutant CEBPA {#s13} ---------------------------------------------------------------------------------------------------- Myb was previously reported to suppress myeloid commitment and to promote self-renewal in haematopoietic progenitors and leukaemia cells ([@bib20]; [@bib48]). We wished to explore how the interplay between Myb and wild-type or mutant C/EBPα influences the transcriptome, so we performed RNA-seq analysis following *Myb* knockdown in the context of wild-type or mutant C/EBPα. In line with previous reports, inspection of the RNA-seq datasets for FMH9, KL, and LL cells showed the expected patterns for typical myeloid genes that are known to be Myb targets, such as *Gfi1, Itgam/CD11b*, *Gr-1/Ly6d,* and *Ccnd2* ([Fig S4A and B](#figS4){ref-type="fig"}). {ref-type="fig"}.\ **(A)** University of California at Santa Cruz (UCSC) genome browser screenshots of RNA-Seq performed in FMH9, KL, and LL cells with si*Myb* and siNEG treatment at differentially regulated genes. Profiles scaled to 1% GAPDH. **(B)** Absolute (top) and siNEG-relative (bottom) FPKM quantification of expression levels of genes are shown.](LSA-2018-00207_FigS4){#figS4} Analysis of the RNA-seq data by global correlation clustering of fragments per kilobase of transcript per million mapped reads (FPKM) in steady-state conditions (i.e., after control siRNA treatment only) revealed higher similarity between KL and LL cells, with FMH9 cells clustering on their own, in line with previous reports on biallelic CEBPA mutants ([@bib44]; [@bib36]) ([Fig 3A](#fig3){ref-type="fig"}). Differential gene expression analysis comparing control and si*Myb*-transfected FMH9, KL, and LL cells revealed 790, 1217, and 40 genes being down-regulated, whereas 1364, 1668, and 329 genes were up-regulated ([Fig S5A](#figS5){ref-type="fig"}). This demonstrates that, whereas FMH9 and KL cells display a large number of genes responsive to *Myb* knockdown, LL cells exhibited only minor transcriptomic responses. Intersection of the gene expression changes showed generally a greater overlap of up-regulated genes than down-regulated genes ([Fig 3B](#fig3){ref-type="fig"}), whereas hierarchical clustering revealed a higher degree of similarity between gene expression changes in FMH9 and KL cells over changes observed in LL cells ([Fig 3C](#fig3){ref-type="fig"}). These findings confirm our hypothesis that the LL-mutant leukaemia phenotype is less dependent on Myb. {#fig3} {ref-type="fig"}.\ **(A)** Boxplot showing FPKM gene expression levels of genes differentially regulated in FMH9, KL, and LL cells following *Myb* knockdown. **(B)** Heat maps showing gene expression fold changes at individual genes from GO categories depicted in [Fig 3D](#fig3){ref-type="fig"}. The detailed list of genes from each pathway/group is provided in Table S2.](LSA-2018-00207_FigS5){#figS5} We next investigated whether *Myb* knockdown results in differential regulation of relevant gene ontologies. We examined negative regulation of G1 to S phase transition ([@bib1]), leukaemia stem cell state ([@bib9]), response to AML treatment ([@bib5]), and terminal myeloid differentiation, computing average log~2~ si*Myb*/control fold changes for each ontologies (Table S1). We found that FMH9 and KL cells, but not LL cells, exhibited down-regulation of the leukaemia stem cell gene expression programme following *Myb* knockdown ([Figs 3D](#fig3){ref-type="fig"} and [S5B](#figS5){ref-type="fig"}). Crucially, this analysis also showed that specifically for FMH9 and KL cells, si*Myb* treatment resulted in up-regulation of genes that are also up-regulated during the treatment of AML. Table S1 Gene sets used for GO analysis in this study. Table S2 List of genes signatures from GSEA analysis presented in Fig S4. To confirm the observations obtained from the RNA-seq, we performed quantitative RT-PCR analysis of selected key genes. Significantly, *Cebpa* mRNA was down-regulated upon *Myb* knockdown in both FMH9 and KL cells, but not in LL cells. This suggests that the functional cross-regulation between Myb and C/EBPα could be lost in the presence of biallelic N-terminal *CEBPA* mutations. This analysis confirmed specific regulation by Myb of differentiation-related genes in FMH9 and KL cells but not in LL cells (*Gfi1* and *Sbno2*), changes in the expression of genes leading to a negative impact on proliferation in FMH9 cells (*Ccnd2* and *Sema4d*), and down- and up-regulation of genes related to increased apoptosis seen in KL and LL cells (*Prune2* and *Dusp1*) ([Figs 3E](#fig3){ref-type="fig"} and [S4B](#figS4){ref-type="fig"}). Moreover, inspection of known Myb target genes revealed a significant repression of *Bcl2* ([@bib37]; [@bib31]), a known anti-apoptotic regulator, in KL cells; concomitantly, we observed the up-regulation of a pro-apoptotic gene normally anti-correlated with *Myb* in AML, namely, *Bcl2l11* (*Bim*) ([@bib14]), in LL cells. This is in agreement with the strong induction of apoptosis observed in these cells upon *Myb* knockdown ([Fig S3A and B](#figS3){ref-type="fig"}). Overall, the transcriptome changes following *Myb* knockdown in cell lines with different CEBPA mutational status are consistent with the corresponding phenotypic changes and further demonstrate that cells with the biallelic LL C/EBPα configuration lack a major dependence on Myb. Coincident binding of C/EBPα p42 and Myb correlates with gene repression by Myb, whereas genes positively controlled by Myb tend to bind C/EBPα p30 {#s14} --------------------------------------------------------------------------------------------------------------------------------------------------- To shed light on the possible interplay between the mutational status of C/EBPα and the Myb-dependent regulation of C/EBPα target genes, we set out to investigate the chromatin binding properties and transcriptional effects of C/EBPα p30, C/EBPα p42, and K313KK-mutant C/EBPα isoforms. The N-terminal--mutant L allele ([@bib16]) leads to the expression of the p30 isoform only, whereas the C/EBPα K313KK--mutant allele gives rise to a C-terminal mutant that disables the DNA-binding domain, resulting in a block in differentiation ([@bib3]). Because the binding dynamics of C/EBPα in double-mutant cells would be technically difficult to characterise, we used previously published chromatin immunoprecipitation sequencing (ChIP-Seq) data from a single isoform transfection model ([@bib12]). This study used HA-tagged *Cebpa* constructs transfected into the FDCP1 cell line ([@bib4]) and an immortalised IL3-dependent murine myeloid cell line that approximates to the wild-type C/EBPα leukaemia line FMH9 that we used for the *Myb* knockdown studies. This latter study concurrently provided gene expression microarray analysis of mock- and C/EBPα-transfected cells. As expected, there were significant increases in *Cebpa* transcript abundance following overexpression of C/EBPα p30, C/EBPα p42, and C/EBPα K313KK as compared with mock transfection ([Fig S6A](#figS6){ref-type="fig"}, compare with [Fig S6B](#figS6){ref-type="fig"} in FMH9, KL, and LL cells). Analysis of ChIP-seq data revealed 17,452, 22,873, and 68,432 peaks for C/EBPα p30-, C/EBPα p42-, and C/EBPα K313KK--transfected cells, respectively, which were mostly located in intergenic and intronic regions ([Fig S6C and D](#figS6){ref-type="fig"}). However, visual inspection of the C/EBPα K313KK dataset revealed low signal in K313KK peaks, hinting that those are not specific, consistent with the loss of binding due to the K313KK mutation in the DNA-binding domain ([@bib12]; [@bib3]). We next characterised specific p30 and p42 peaks by ranking tag counts around merged summits by p30/p42 fold change and identified 3,585 p42-specific, 19,949 shared, and 4,421 p30-specific peaks ([Fig 4A](#fig4){ref-type="fig"}). Although we did not use the C/EBPα K313KK dataset as a direct base for comparison because of the deleterious effect of the K allele on DNA binding, and thus low signal to noise ratio, its binding pattern was mostly located in shared sites. Crucially, by retrieving tag counts for Myb ChIP-Seq datasets, we observed that Myb binding largely parallels that of p42 binding, both in FDCP1 and in an MLL-AF9/NrasG12D murine AML cell line ([@bib29]), although there was an overlap with some regions that predominantly bind p30 ([Fig 4A](#fig4){ref-type="fig"}). GO analysis of the p42-specific peaks revealed signalling pathways involved in haematopoietic homeostasis and pro-apoptotic genes ([Fig 4B](#fig4){ref-type="fig"}). Conversely, p30-specific peaks were enriched in pluripotency genes, consistent with the leukaemia stem cell signature seen amongst the up-regulated genes following *Myb* knockdown in LL cells ([Figs 3D](#fig3){ref-type="fig"} and [4B](#fig4){ref-type="fig"}, and [S5B](#figS5){ref-type="fig"}). {ref-type="fig"}.\ **(A)** *Cebpa* transcript abundances in FDCP1 cells transfected with mock, p30, and p42 C/EBPα isoforms measured by RNA-Seq, relative to *B2m*. **(B)** *Cebpa* transcript abundances in FMH9, KL, and LL cells measured by quantitative RT-PCR, relative to *B2m*. This bar plot represents an average of six independent experiments. Statistical analysis was performed using *t* test (\*\*\**P* \< 0.001, \*\**P* \< 0.01, \**P* \< 0.05). **(C)** UCSC genome browser screenshot of C/EBPα p30, p42, and K313KK ChIP-Seq datasets in the FDCP1 cell line at the *Runx1* gene locus. **(D)** Stacked percentage bar plots showing genomic annotation of C/EBPα p30, p42, and K313KK ChIP-seq peaks in the FDCP1 cell line.](LSA-2018-00207_FigS6){#figS6} {#fig4} To characterise the consequences of C/EBPα binding on gene expression, we performed GSEA using microarray datasets from FDCP1 cells expressing p30, p42, and K313KK, ranking by log~2~ fold change against mock transfection. We selected gene sets corresponding to the closest genes of the top 1,000 peaks for cognate ChIP-Seq datasets (p30, p42, and K313KK) in accordance with the constraints of GSEA. We observed significant correlations between p42-induced gene activation and p42 binding, as well as between p30-induced gene repression and p30 binding ([Fig 4C](#fig4){ref-type="fig"}). However, C/EBPα K313KK binding was not correlated with changes in gene expression. We next asked how C/EBPα isoform binding correlates with genes whose expression is altered by *Myb* knockdown. We plotted gene expression fold changes caused by *Myb* knockdown in FMH9, KL, and LL cells against the ChIP-seq data ranking of p30/p42 binding. Two broad conclusions arise from this analysis: first, Myb-repressed genes largely bound p42 in the presence or absence of p30; second, genes that are positively regulated by Myb are more predominant amongst the group of genes that preferentially bind p30 ([Fig 4D and E](#fig4){ref-type="fig"}). Conversely, p30 binding was linked with down-regulation of gene expression upon *Myb* knockdown. We next investigated the sequence content of p30- and p42-specific sites by performing motif discovery analysis in the sequences corresponding to these peaks. This revealed that C/EBP, AP-1, Ets, Myb, and Runx motifs were highly enriched in p42-specific peaks ([Fig S7A](#figS7){ref-type="fig"}). However, the C/EBP motif was not enriched in the p30-specific peaks. Instead, these peaks were enriched in CTCF, Nrf, and Ets motifs. To confirm these trends, we plotted motif matches amongst increasing p30/p42 fold change as above. Strikingly, C/EBP motifs seemed to be restricted to p42-specific and shared sites ([Fig S7B](#figS7){ref-type="fig"}). Myb, Runx, AP-1, and Ets motifs also followed this trend. Conversely, CTCF motifs were highly enriched in p30-specific sites and to some extent in shared sites corresponding to lower p42 binding. Nrf, Sp1, and CREB motifs also seemed to follow this trend, but not Elf motifs. To verify binding of cognate factors to these motifs as well as active and inactive transcriptional hallmarks, we made use of publicly available AML ChIP-Seq datasets ([@bib46]; [@bib29]). Our analysis revealed that C/EBPβ, an essential transcription factor in normal myeloid development ([@bib11]), which is able to bind C/EBP motifs as well, also co-localised with p42 ([Fig S7C](#figS7){ref-type="fig"}). Furthermore, by retrieving tag counts for the transcription activation hallmark p300, we could also show that p42-specific and shared sites, but not p30-specific sites, correspond to putative enhancer sites, consistent with our GSEA analyses for this isoform ([Figs 4C](#fig4){ref-type="fig"} and [S7D](#figS7){ref-type="fig"}). {ref-type="fig"}. **(A, C)** RN2 ChIP-Seq tag counts for transcription factors known to bind motifs from (A) as well p300 sorted as in (C) and [Fig 4A](#fig4){ref-type="fig"}. **(D)** Heat map showing MEL ChIP-Seq tag counts for transcription factors known to bind motifs from [Fig S5A](#figS5){ref-type="fig"}, as well as co-activators and co-repressors sorted by p30/p42 C/EBPα tag count fold change as in [Fig 4A](#fig4){ref-type="fig"}.](LSA-2018-00207_FigS7){#figS7} Discussion {#s15} ========== In the present study, we have investigated the requirement for the transcription factor Myb in the maintenance of CN-AML driven by different combinations of *CEBPA* mutations in comparison with leukaemia characterised by the expression of wild-type C/EBPα. We show for the first time that the dependency on Myb is affected by the mutational status of C/EBPα. Compared with cells expressing wild-type C/EBPα, which show a proliferation and differentiation response to enforced reduction in Myb levels, leukaemia driven by biallelic *CEBPA* mutations exhibits distinct phenotypic responses that are reflected in changes in gene expression. Furthermore, leukaemia with biallelic N^ter^/N^ter^ *CEBPA* mutations shows a reduced dependency on Myb, whereas C^ter^/N^ter^ mutant--driven AML cells are as reliant on Myb as those expressing wild-type C/EBPα but exhibit a quite distinct pattern of phenotype and gene expression changes upon *Myb* knockdown. Here, we show that knockdown of Myb in leukaemia cells harbouring either wild-type or C^ter^/N^ter^--mutant C/EBPα reverses the abnormal myeloid phenotype normally observed in leukaemia, whereas AML cells carrying biallelic N^ter^/N^ter^ mutations exhibit persistence of the undifferentiated phenotype. These behaviours are reflected in distinct changes in gene expression. Hence, *Myb* knockdown in leukaemia cells with wild-type or C^ter^/N^ter^--mutant C/EBPα resulted in the loss of a leukaemia stem cell signature and the up-regulation of a gene expression pattern generally observed in patients that are responding to therapeutic treatment. In line with previous reports ([@bib49]), our analysis demonstrated *Gfi1* and *Sbno2*, both of which encode transcription factors that can be related to the loss of differentiation block, to be positive and negative targets of Myb activity, respectively. The reduced dependency of N^ter^/N^ter^--mutant C/EBPα--driven AML cells on Myb is also paralleled by very little change in the number of genes affected by *Myb* knockdown, including no effect on *Gfi1* or *Sbno2*. Interestingly, Myb targets the expression of the *Cebpa* gene itself in the wild-type and C^ter^/N^ter^ C/EBPα contexts, suggesting positive feedback that is not seen in C/EBPα N^ter^/N^ter^ cells. This result might provide a hint why LL leukaemia displays a different response to *My*b manipulation. Analysis of cell cycle and apoptosis also revealed how knockdown of Myb can lead to quite a distinct phenotype; for instance, although no induction in apoptosis/necrosis was observed in C/EBPα wild-type cells, both mutant cell lines displayed a large increase in the percentage of Annexin V^+^ cells and the appearance of a sub2n population, which is indicative of cells undergoing cell death. However, it is possible that other biological pathways could be affected that lead to such a different response, with the cells perhaps being forced to engage in non-apoptotic fates, such as necroptosis, autophagic cell death, or pyroptosis ([@bib35]); these possibilities remain to be further elucidated. Considering the different involvement of p30 and p42 in the three leukaemia scenarios we have investigated, the fact that there are p42-only-- and p30-only--bound genes indicates that a distinct response should be expected when Myb is reduced, especially when only p30 is present. Most p42- or p42+p30--binding genes also bind Myb, but interestingly, a significant number of p30-only target genes probably do not bind Myb. The genes that bound predominantly p42 fall into GO groups, including those associated with myeloid homeostasis and differentiation, consistent with the finding that *Myb* knockdown restores myeloid differentiation. Conversely, but in agreement with a persisting leukaemia phenotype not being affected by Myb manipulation, genes bound by p30 include pluripotency genes. This is in line with a previous report that mature cells can tap into stem cell regulatory networks when experiencing mutational hits in key differentiation factors ([@bib34]). The vast majority of genes affected by *Myb* knockdown in the context of N^ter^/N^ter^ biallelic--mutant C/EBPα are de-repressed and overlap with genes similarly affected in the context of C^ter^/N^ter^. That genes such as *Dusp1* are not responsive to Myb changes in the context of wild-type C/EBPα presumably means that p42, being fully competent to dimerise and bind DNA is less dependent on Myb. Mapping these gene expression changes onto the profile of p42/p30 binding revealed that they can be expected normally to bind both p42 and p30. The large number of genes affected by *Myb* knockdown that distinguished the response of C^ter^/N^ter^ from N^ter^/N^ter^ are both down-regulated (1,188 genes) and up-regulated (1,423 genes), the majority of the former corresponding to genes that are preferentially bound by p30. In contrast, the up-regulated genes, normally repressed by Myb as seen in the N^ter^/N^ter^ situation, tend to be genes that exhibit greater binding by p42 when it is present. Intriguingly, in the N^ter^/N^ter^ situation, in which most of the genes affected by *Myb* knockdown reflect loss of Myb-dependent repression, these genes also fall into the category of preferential binding to p42. Because no p42 is present in the N^ter^/N^ter^ leukaemia, this must mean that such genes can be regulated by Myb without a prerequisite for cooperation with C/EBPα p42 or alternatively that another C/EBP family protein such as C/EBPβ can act in place of C/EBPα. In this circumstance, the profile of C/EBPβ binding to genes in the context of myeloid leukaemia cells parallels that of C/EBPα p42. The consequence of this is that most of the genes bound by C/EBPα p42 that are Myb dependent would not exhibit such redundancy of C/EBP protein requirement. By analysing the occurrence of transcription factor consensus binding motifs in the sequence content of the C/EBPα ChIP-seq peaks, we observed that the regulatory network maintaining the different leukaemia statuses involves different possible sets of transcription factor binding. p42-specific peaks are enriched in C/EBP, Ap1, Myb, Ets, and Runx motifs, whereas the p30-specific peaks contain mostly CTCF, Nrf, and Ets motifs. Importantly, we also observed that p42-specific peaks that are also bound by Myb are highly enriched in the transcription activation hallmark p300. This is to be expected as p300 has been demonstrated to be one of the most important cofactors for Myb and the inhibition of the Myb/p300 interaction is crucial for the maintenance of the leukaemia state ([@bib25]). The same analysis performed for the p30-specific peaks revealed those sites may be enriched for CTCF in cells that do not express CEBPα p42 ([Fig S7D](#figS7){ref-type="fig"}). In conclusion, we have shown that the nature of mutations in one transcription factor that drives leukaemia can dictate how another leukaemia-associated oncoprotein affects the maintenance of the leukaemia phenotype. The precise nature of the interactions between C/EBPα or its mutated variants and the Myb protein on specific genes that dictate the leukaemia phenotype remain to be elucidated, and it would be fruitful to assess in more detail the relevance of this interaction in human leukaemia patients harbouring those mutations. Added complexity in the involvement of C/EBP proteins in determining the Myb dependency of a given leukaemia most likely goes beyond solely the balance of the C/EBPα isoforms, especially given the possibility that co-expressed factors such as C/EBPβ might compete for binding to C/EBP motifs and heterodimerise with C/EBPα. Ultimately, our findings call for a larger study to determine how manipulating MYB would impact on the maintenance of both murine and human leukaemia driven by other genetic lesions. Supplementary Material ====================== ###### Reviewer comments The work was supported by a Bloodwise Programme Grant (12010) held by J Frampton, S Dumon, and P Garcia and through funding provided by the College of Medical and Dental Sciences of the University of Birmingham. This project has received funding from the European Research Council under the European Union's Horizon 2020 research and innovation programme (grant agreement no. 636855/StG to F Grebien) Authors Contributions {#s17} ===================== G Volpe: conceptualisation, data curation, investigation, and writing---original draft.P Cauchy: data curation, formal analysis, and writing---original draft.DS Walton: investigation.C Ward: formal analysis.D Blakemore: investigation.R Bayley: investigation.ML Clarke: investigation.L Schmidt: investigation.C Nerlov: resources.P Garcia: funding acquisition and writing---review and editing.S Dumon: conceptualisation, supervision, and funding acquisition.F Grebien: conceptualisation, formal analysis, investigation, and writing---review and editing.J Frampton: conceptualisation, supervision, funding acquisition, project administration, and writing---review and editing. Conflict of Interest Statement {#s18} ============================== The authors declare no conflicts of interest. [^1]: Giacomo Volpe and Pierre Cauchy are joint first authors [^2]: Stéphanie Dumon, Florian Grebien, and Jon Frampton are joint senior authors

Background ========== In many low- and middle-income countries, the private commercial sector plays an important role in the provision of malaria treatment \[[@B1]-[@B6]\]. A study on the market for anti-malarial drugs in six developing countries found that private providers were responsible for around 40% of all anti-malarial sales volumes in Zambia, 55% in Uganda, 71% in Cambodia, and 75% in both Benin and the Democratic Republic of Congo and 98% in Nigeria \[[@B2]\]. Their popularity is commonly attributed to convenience as they tend to operate closer to homes \[[@B7]-[@B10]\], and availability and reliability of drug stocks compared to public health providers \[[@B8],[@B9],[@B11]-[@B14]\]. Private providers vary substantially within and across countries and can include hospitals, clinics, pharmacies, drug shops, grocery stores, village shops, market stalls and mobile providers. Given the importance of private commercial outlets, there has been increased interest in analysing their role in treatment provision and how this can be improved. A key aspect of this is the measurement of anti-malarial and diagnostic sales volumes. Sales volumes data can be used to calculate market shares and provide information on the relative importance of public and private sectors and how this varies across countries. These data can also be used to estimate the share of recommended first-line drugs or banned drugs, such as oral artemisinin monotherapy, sold in the market \[[@B1],[@B2],[@B15]\]. Data on sales of rapid diagnostic test kits for malaria can indicate the low frequency of parasitological-based diagnosis of malaria fevers \[[@B1]\]. Anti-malarial sales volumes have been measured to evaluate the effect of major drug subsidy programmes, such as the Affordable Medicine Facility-malaria (AMFm) for which the change in the market share of quality assured artemisinin combination therapy (ACT) was one of four key success metrics, alongside availability, price and use \[[@B15]\]. Finally, market share data can be used for assessing the nature of competition in the market for malaria treatment and its relationship with key retail market outcomes, notably anti-malarial prices and price mark-ups. For example, research on the market for malaria treatment in Tanzania and Cambodia has shown that market concentration measured using the Hirshman-Herfindahl index (HHI) (the sum of squared market shares of each firm in the market) is related to the extent to which providers can influence the price of anti-malarial drugs sold in the market \[[@B16],[@B17]\]. A number of methods for measuring sales volumes have been identified, namely reviewing providers' sales records, asking providers to recall their sales volumes over a given period, conducting exit interviews with customers, and retail audits. Retail audits involve visiting a panel of outlets to collect stock information at regular intervals; at each visit, fieldworkers measure the stocks of an entire product category and ask the shopkeeper about any volumes added and/or disposed of during the visit interval. The volume of sales for each shop during the period is then estimated by subtracting the stock at the end of the period from the stock at the initial visit, corrected by any additions/disposals during the period. These data collection methods each have methodological challenges for collecting commercial sales volume data at private businesses in developing country settings. Providers' sales records may be non-existent, incomplete and/or outdated. Private commercial providers may be reluctant to share their sales records as they may fear that these could be disclosed to drug regulation bodies, revenue authorities or competitors. Asking providers to recall their volumes is a convenient and popular method used in many surveys \[[@B18]\], but it may be prone to recall bias. Providers may also be unwilling to record or recall the sales of products that they are not authorized to handle. The retail audit approach could be perceived to provide more accurate responses as it does not rely on respondents' ability to remember their sales volumes. However, as it requires at least two visits to the same outlets within a given period it is likely to be more costly and logistically complex than relying on records or provider recall. Exit interviews may be another approach to address this problem, although the presence of interviewers may bias sales patterns. Shopkeepers may also not allow interviewers to stand outside their shops or consumers may be reluctant to share information about their purchases or they may be in a rush leaving the shop. Overall, the available approaches tend to be better suited for estimating the sales of licensed rather than unlicensed outlets \[[@B19]\] and sales of registered rather than unregistered products. Whilst the challenges of these different methods have been identified, what is lacking is evidence on how sales volume estimates collected through different methods in the same context compare. Furthermore, the available literature concentrates mainly on retail medicine providers, yet drug retailers are the last link in a chain of suppliers, including several layers of wholesalers. Much less attention has been paid to this market segment \[[@B20]\] and guidance on how to study wholesalers, notably on how best to collect their sales volume data, is scarce \[[@B17],[@B18]\]. Using sales volume data collected at retail and wholesale outlets through two different methods, namely provider recall and retail audits, this paper provides evidence on the degree of agreement between the two methods in Cambodia. The study was undertaken with retail and wholesale providers as part of the ACTwatch project \[[@B18]\], which measured sales volumes using the recall method. Retail audits were selected as the comparative method because they had been used for measuring retail anti-malarial sales volumes in previous studies \[[@B16],[@B21]\]. Written records were not included in the study due to their rarity among less formal providers. Exit interviews were not included because anti-malarials represented a small share of wholesalers' total business \[[@B17]\], implying that interviewers may have had to wait many days outside a shop before identifying a wholesale anti-malarial customer. At the time of the study, there were three categories of licensed medicine outlets in Cambodia, including pharmacies that were managed by a pharmacist, and dépots managed by an assistant pharmacist (dépot A) or a retired public health staff member (dépot B) with a minimum qualification of nurse or midwife \[[@B17]\]. Pharmacies were authorized to engage in both wholesale and retail activities whilst dépots were authorized to retail only \[[@B17]\]. Licensed providers were authorized to sell registered pharmaceutical drugs, hygienic and cosmetic products with preventive and curative properties, and dental, laboratory and medical equipment. At the time of the study, there were no clear regulations on the sales of malaria diagnostics, such as rapid diagnostics tests (RDTs). Around 520 pharmacies and 695 dépots were estimated to operate in Cambodia, supplemented by many other medicine sellers that operated illegally, including unlicensed pharmacies and drug shops that sold medicines, cosmetics and household goods; private clinics (sometime referred to as cabinets or clinical pharmacies) that sold medicines and also provided outpatient and/or inpatient clinical services; mobile providers who travelled to patients' homes to provide clinical services, and at times offered outpatient and/or inpatient care at fixed outlets; and, grocery and village shops that sold medicines alongside food, soft drinks and other consumer goods \[[@B1],[@B22]\]. Methods ======= Sales volume data for anti-malarials and RDTs were collected through the recall (RC) and the retail audit (RA) methods, together referred to as the sales level surveys (SLS) in retail and wholesale commercial anti-malarial providers. The aim of the SLS was to explore whether RC and RA for measuring sales volumes agreed sufficiently that they can be used interchangeably. The relative strengths and weaknesses of each method were also analysed from an implementation perspective using qualitative methods. The sales level surveys ----------------------- RA consisted of visiting each sampled outlet two times with a two-week time interval between each visit. At the first visit, referred to as the sales level survey 1 (SLS1), data on quantities stocked of each product were collected. At the second visit (SLS2), data on quantities stocked, quantities delivered between first (SLS1) and second (SLS2) visits, and quantities thrown away/transferred to other shops or sent back to wholesalers or confiscated were collected for each product in stock, including products in stock at either or both visits. Quantities stocked were physically counted where possible or providers were asked to state the quantities in stock. To collect data on quantities delivered and disposed of, providers were asked to check any available written records or sales receipts and in the absence of records, to recall these quantities. RC consisted of asking retailers and wholesalers to recall the quantities sold during the two-week time interval between SLS1 and SLS2. It was implemented at the start of SLS2 before collecting stock data in order to minimize bias as recall data may have been influenced by the process of counting stocks for the RA. A time interval of two weeks between the two visits at each outlet was chosen based on the existing literature in which a two-week time interval was considered reasonable for capturing wholesale deliveries \[[@B16],[@B21]\]. A recall period of two weeks has also generally been used for collecting data such as fever episodes in household surveys \[[@B23]-[@B25]\]. The SLS sampling strategy drew on data collected during the ACTwatch retail outlet and supply chain surveys, which are described in detail elsewhere \[[@B18]\]. Briefly, for the ACTwatch retail outlet survey, a sample of 38 administrative clusters (health centre areas with catchment populations of 10--15,000 inhabitants) were selected with probability proportional to size from all 255 malaria endemic clusters in Cambodia \[[@B1]\]. Then, a census of all public and private outlets in these 38 clusters was completed and a list of those stocking anti-malarial drugs was created \[[@B1]\]. At each retail outlet, data were collected on the two most important wholesale supply sources for anti-malarial drugs. All anti-malarial wholesale supply sources mentioned by retailers were visited during the ACTwatch supply chain study. Data were collected on the two most important wholesale supply sources. This process was then repeated until the top of the chain was reached. For the SLS, retail and wholesale outlets were purposively sampled from the list of anti-malarial retailers and wholesalers surveyed during the ACTwatch surveys. The geographical location of each commercial outlet and the number of outlets stocking anti-malarials in each location at the time of the outlet survey were used to select areas in which all outlets could be visited two times with a two-week time interval (in order to conduct the RA component). A total of 107 retailers and 67 wholesale outlets were sampled. Wholesale and retail outlets not found, not stocking anti-malarials or not available at the time of the SLS were not replaced. At retail level, the SLS was conducted as a standalone survey few months after the ACTwatch retail outlet survey. At wholesale level, the SLS took place during the ACTwatch supply chain survey. The questions relating to SLS1 (questions about quantities stocked) were asked after the supply chain survey questionnaire was administered, whilst SLS2 (questions about recall sales volumes, quantities stocked and quantities received and disposed of) was conducted two weeks later as a standalone survey. All data collection tools were translated from English to Khmer and piloted before the start of data collection. A team of two interviewers entered each business, informed shopkeepers about the study objective and obtained consent. Interviews were conducted in Khmer, with the person most involved in the management of the business. Interviews were conducted in the premises, with breaks each time a customer arrived. Interviewers then asked whether they could return after two weeks and if so they arranged an appointment, and returned on that date. All types of anti-malarial drugs in all dosage forms and packaging types, and RDTs were surveyed. For anti-malarials, data were collected in terms of both full packs and loose tablets (ie, those kept in containers/tins). Stock data for anti-malarials stored in half-full containers were estimated based on the height of the tablets in the pot measured using a ruler and the number of tablets in a full pot. RDT data were collected in terms of single RDT units. For each anti-malarial observation, volume estimates were converted into adult equivalent treatment doses (AETDs) \[[@B1]\]. One AETD was defined as the amount of the drug needed for a full adult course of treatment based on guidelines from the World Health Organization (WHO) where available, or else from peer-reviewed literature or manufacturers. Anti-malarials missing data required to calculate AETDs (eg, drug strength) were excluded from the sales volume estimation \[[@B1]\]. RA estimates were calculated as: (total quantities stocked at SLS1) + (quantities delivered between SLS1 and SLS2) -- (quantities disposed of between SLS1 and SLS2) -- (total quantities stocked at SLS2). Negative RA estimates indicating data collection errors during the SLS and anti-malarial/RDT observations without both RA and RC estimates were excluded from the analysis. In outlets with sales data for more than one type of anti-malarial/RDT the sum of all RC estimates and of all RA estimates was calculated in order to obtain for each outlet single total sales volume estimates with each method. The level of agreement between the two methods was explored following the Bland-Altman approach \[[@B26],[@B27]\]. The first step was to calculate, for each outlet, the difference between RA and RC sales volume estimates for outlet ~i~. Formally: $$RA_{i}‒RC_{i}$$ where RA~i~ and RC~i~ are sales volumes estimated through the two different methods at outlet ~i~. The second step was to estimate the "bias of the measurement" between the two methods, which is the mean of the differences between the two different methods (b1), and its standard deviation (SD) (b2). Formally: $$\overline{{RA}‒{RC}} = \frac{1}{n}{\sum{{}_{i}^{n}\left( {RA_{i}‒RC_{i}} \right)}}$$ $${SD} = \sqrt{\frac{1}{n‒1}{\sum\left( {X_{i}‒\overline{X}} \right)^{2}}}$$ where $x_{i} = \left( {RA_{i}‒RC_{i}} \right)$ is the difference between RA and RC in outlet~i~$\overline{x} = \overline{{RA}‒{RC}}$ the mean of the differences between RA and RC across all outlets and n the total number of outlets with a pair of RA and RC estimates. Differences between sales volume estimates were plotted on a histogram (not shown) to verify that they were approximately normally distributed. The third step was to investigate for each outlet whether there was an association between the total volume sold and the bias (ie, the mean of the difference, b1). This is because for the bias to be a meaningful estimate of the level of agreement between the two different methods, it should be constant throughout the range of measurements \[[@B26],[@B27]\]. In the absence of a recognized gold standard method for measuring sales volumes, an outlet's "true" total sales volume was proxied as the mean of RC and RA estimates. Formally: $$\frac{RA_{i} + RC_{i}}{2}$$ The association between total volume sold (c) and measurement bias (b1) was explored graphically using a scatter plot of the differences against total volume sold and confirmed statistically using a correlation coefficient obtained through the STATA command *baplot*\[[@B28]\]. The fourth step was to calculate the interval within which 95% of paired estimates were expected to lie, referred to as the upper and lower limits of agreement (LoA) between the two methods \[[@B26]\]. Formally: $${LoA} = \overline{{RA}‒{RC}} \pm 1.96{SD}$$ Qualitative methods ------------------- The quantitative analysis was supplemented by qualitative data on information about the implementation process of RC and RA. Qualitative data were drawn from fieldworkers 'diaries, which had been completed at the end of each outlet visit. In each diary, fieldworkers described and compared their experiences in collecting data across RC and RA, products, dosage forms and packaging types. They also recorded observations of shopkeepers' behaviour during data collection. Semi-formal group discussions were also organized during the course of the fieldwork to clarify diary entries. These discussions provided a forum for fieldworkers to elaborate on particular topics, share arduous experiences, discuss their views and trade funny stories. Group discussions also had the advantage of creating interactions between fieldworkers, which prompted others to remember their own experiences \[[@B29]\]. Group discussions were facilitated in English and/or Khmer by the corresponding author with the assistance of a trained Cambodian research assistant and recorded using written notes. Five group discussions were conducted at the mid and end of data collection with each of the three fieldworker teams involved in the SLS. Fieldwork diaries kept in Khmer were translated into English by a trained research assistant. These data were analysed using a simple thematic content approach through which recurrent themes under each of the topics discussed were listed and compared. Ethics considerations --------------------- The study received ethics clearance from the Cambodian National Ethics Committee for Health Research (no. 041 NECHR) and ethics review committee of the LSHTM (no. 5466). Informed consent from each interviewed shopkeeper was obtained at SLS1 to cover both SLS visits. For diaries and group discussions, fieldworkers' participation as research subjects was explained during the recruitment process and consent received orally from each fieldworker recruited. Results ======= Quantitative results: Bland-Altman approach ------------------------------------------- Of the 67 wholesalers and 107 retailers initially sampled, 58 and 62% participated in the SLS, respectively. Reasons for non-participation at both wholesale and retail outlets at SLS1 included outlets not found, not open at the time of visit or not stocking anti-malarial drugs, whilst at SLS2 the main reason for non-participation was provider refusal. The SLS wholesale sample was similar to that surveyed during the nationally representative ACTwatch supply chain survey: outlets had a median of two workers (IQR 2--2), had been in operation for 10 years (IQR 4--13) and around 70% employed a member of staff with health qualifications, with nurse/midwife being the most commonly reported qualification type \[[@B17]\]. The SLS sample of retailers included pharmacies/clinical pharmacies (13%), drug shops (21%), mobile providers (20%), grocery stores (26%) and village shops (20%). Retailers shared similar characteristics with those of the commercial outlets interviewed during the ACTwatch outlet survey: staff with health qualifications were more commonly found at pharmacies (85%), drug shops (76%) and mobile providers (70%) than at grocery and village shops (13 and 12%, respectively), and the most commonly reported health qualifications were nurses/midwives. A median of two people (IQR 1--2) worked at the sampled outlets and shops had been in operation for a median of eight years (IQR 2--15) \[[@B17]\]. Surveyed anti-malarials were found in tablet and injectable forms only. Tablets were commonly stocked in packs, and injectables in individual ampoules. Tablets kept in opened tins/containers were rare and found at retail outlets only. At wholesale outlets, 104 different anti-malarial products were surveyed. Sales volumes were collected for 76 anti-malarial products through RC and for 82 through RA (Table [1](#T1){ref-type="table"}). For RDTs, 34 different products were surveyed and sales volumes were collected for 26 products through the RC and for 29 through the RA. The main reasons for non-response included wholesalers' refusal to recall their sales volumes for the RC, and for the RA wholesalers' refusal to let interviewers record stock data (Table [1](#T1){ref-type="table"}). Wholesale sales volumes were estimated through both RC and RA for 62 anti-malarials and 23 RDTs. ###### Data collected on wholesale and retail sales volumes using recall and retail audit methods   **Number of products surveyed (%)**^**1**^ -------------------------------------------------------- -------------------------------------------- ------------ ------------- ------------- Total products surveyed 104 (100%) 34 (100%) 143 (100%) 42 (100%) Recall method (RC)     Sales volume data collected 76 (73.1%) 26 (76.5%) 130 (91.0%) 41 (97.6%) \- Not remembered 17 (16.3%) 7 (21.6%) 3 (2.8%) 1 (2.4%) \- Refused 11 (10.6%) 1 (2.9%) \- \- \- Missing^2^ \- \- 10 (7.0%) \- Retail Audit method (RA)     Sales volumes data calculated (excluding negatives) 82 (78.8%) 29 (85.3%) 115 (80.4%) 35 (83.3%) Sales volumes data calculated (including negatives)^3^ 94 (90.4%) 31 (91.2%) 121 (84.6%) 39 (92.9%) Stock data collected 96 (92.3%) 31 (91.2%) 121 (84.6%) 39 (92.9%) \- Refused 8 (7.7%) 3 (8.8%) 12 (8.4%) 3 (7.1%) \- Missing^2^ \- \- 10 (7.0%) \- Received quantities collected 103 (99.0%) 33 (97.1%) 133 (93.0%) 42 (100.0%) \- Refused 1 (1.0%) 1 (2.9%) \- \- \- Missing^2^ \- \- 10 (7.0%)   Disposed quantities collected 101 (97.1%) 33 (97.1%) 133 (93.0%) 42 (100.0%) \- Refused 3 (2.9%) 1 (2.9%) \- \- \- Missing^2^ \- \- 10 (7.0%)   ^1^At the second visit of RA during which RC was implemented. ^2^missing strength data impeded calculation of data in terms of adult equivalent treatment doses. ^3^negative sales volume estimates were obtained when calculating (quantities in stock at 1st visit + quantities received in-between the 2 visits -- quantities at 2nd visit -- quantities disposed in-between the 2 visits), e.g. quantities stocked at second visit were higher than quantities stocked at first visit although shopkeepers did not report any quantities received. These negative estimates were excluded from the analysis. At retail outlets, 143 anti-malarial products were surveyed. Sales volume data were collected through the RC for 130 anti-malarial products and through the RA for 115 anti-malarial products (Table [1](#T1){ref-type="table"}). For RDTs, 42 different products were surveyed and sales volumes were collected for 41 through the RC and for 35 through the RA (Table [1](#T1){ref-type="table"}). Retail sales volumes were estimated through both RC and RA for 113 anti-malarials and 33 RDTs. RC and RA sales volume estimates were obtained for 34 wholesale outlets when considering anti-malarials and 23 wholesale outlets when considering RDTs. Similarly, estimates were obtained for 58 retail outlets when considering anti-malarials and 33 outlets when considering RDTs. At one retail outlet, the total volume sold was surprisingly high and well above other retailers' total sales volumes (outlier outlet total sales volume estimated at 129 AETDs compared to volumes at all other retail outlets ranging from 0 to 20 AETDs). This outlying observation obscured the interpretation of results so it was excluded from the main analysis, which was run on paired anti-malarial estimates available for 57 retail outlets. Figure [1](#F1){ref-type="fig"} presents the scatter plots showing on the y-axis the between-method differences, and on the x-axis the mean of the sales volume obtained by the two methods for the outlet. The dashed blue line drawn at y = 0 represents the line of equality between the volumes measured by the two methods. The mean of the between-method differences is represented by the red line and the LoA between the two methods are represented by the two dashed red lines. {#F1} Figure [1](#F1){ref-type="fig"}a shows no evidence of correlation between the between-method differences and the size of the volumes sold for wholesale anti-malarial sales. This was confirmed by a coefficient of correlation *r* = −0.04 (p = 0.83). The mean difference between RA and RC estimates for anti-malarials was four AETDs indicating that RA provided on average significantly higher estimates than RC (95% CI 0.6-7.2). The LoA indicated that for 95% of paired estimates the between-method difference (RA minus RC) would lie between plus 23 AETDs (95% CI 16.0-28.0) and minus 15 AETDs (95% CI −20.4- -9.0). For RDT sales volumes at wholesale outlets, Figure [1](#F1){ref-type="fig"}c shows no evidence of correlation between the between-method differences and volumes sold (*r* = 0.04, p = 0.86) and no significant difference between RC and RA estimates (95% CI −6.0-4.0). The LoA were from plus 22 to minus 21 tests. At retail outlets, there was some indication from Figures [1](#F1){ref-type="fig"}b and [1](#F1){ref-type="fig"}d that the between-method differences were positively correlated with volumes sold (for anti-malarials *r* =0.49, p \< 0.001; RDTs *r* =0.38, p = 0.03). When including the outlet with the outlying anti-malarial sales volume the evidence of a correlation between the between-method differences and sales volumes was also statistically significant and stronger (*r* = 0.823, p \< 0.001; mean difference was two AEDTs (95% CI 0.23-4.0) and LoA were from plus 16 AEDTs to minus 12 AETDs). The mean difference cannot therefore be considered a meaningful estimate of the level of agreement between the two different methods. However, Figures [1](#F1){ref-type="fig"}b and [1](#F1){ref-type="fig"}d seem to indicate that RA tended to provide higher estimates than RC at higher levels of sales volumes (around above five AETDs for anti-malarials and 10 units for RDT). Qualitative results: fieldworkers' experiences and perceptions -------------------------------------------------------------- Data collectors found the RC to be a more convenient approach than RA for collecting sales volume data, notably at retail outlets where shopkeepers said they rarely had customers for malaria treatment. Data collectors also mentioned that retailers seemed to be more comfortable remembering sales volumes of RDTs than anti-malarials and that this was because performing a malaria test was a more memorable and discrete event than selling anti-malarials. However, data collectors often questioned the accuracy of RC sales volume data collected both at retail and wholesale outlets. *"She said: "it might be like this, \[or\] it may be like that"".* (Fieldwork diary \#1 about RC implemented at a retail outlet). *"She might have misreported her sales volumes, because I saw five empty boxes of anti-malarials near her".* (Fieldwork diary \#5 about RC implemented at a wholesale outlet). Data collectors indicated that when shopkeepers could not remember their sales volumes, this was because they often handled other consumer goods, including toiletries or groceries, which were their main selling items. Another reason was that more than one person worked at the shop, making it difficult for respondents to provide accurate estimates. Fieldworkers also said that they perceived wholesalers to be less capable of remembering their sales volumes because they generally handled a wider range of drugs and sold larger volumes. Data collectors reported that during the RA counting stocks was relatively easy and quick because of the small range of anti-malarials and RDTs available at each outlet. They also reported that counting RDTs tended to be easier than anti-malarials, especially when anti-malarials were kept in opened tins. For example, one interviewer explained that in one shop the tin was not transparent, preventing him from using a ruler so that he had to count each tablet left in the tin. Also, at times, interviewers reported they had estimated more pills in the tin at the second than at the first visit although shopkeepers said that no new tin had been opened. Interviewers said that when they collected data on quantities received and disposed of shopkeepers remembered generally very easily because the reported quantities were generally small and often null. They also reported that shopkeepers were generally surprised to be asked about disposed quantities because they said that they never throw products away nor send these back to suppliers. However, data collectors reported important challenges around the implementation of RA. First, they indicated that both wholesalers and retailers refused at times to let interviewers physically count the quantities in stock, with this challenge occurring more commonly at wholesale than retail outlets. *"They did not allow us to count and they did not want to count for us at all \[...\] they said they didn't want to spend time with us \[...\] they said that it \[the survey\] was useless and wasting their time"* (Fieldwork diary \#40 about RA implemented at a wholesale outlet). *"She claimed that I asked the same question at first visit. She said that I should write the same amount as at first visit".* (Fieldwork diary \#2 about RA implemented at a wholesale outlet). Second, data collectors reported that in many cases shopkeepers preferred to estimate their stock from memory, rather than have these counted. Fieldworkers added that this situation was again more common amongst wholesalers who often refused to let interviewers open the cupboards where they kept the drugs. This was corroborated by the SLS quantitative data, which showed that at SLS1 stocks of anti-malarial drugs were physically counted for around 51% of all anti-malarial products surveyed at wholesale outlets compared to 97% at retail outlets. In outlets where stocks could not be physically counted, quantities stocked were stated by wholesalers. In some cases, data collectors explained that the quantities stocked were estimated by memory due to factors beyond the control of shopkeepers. For example, one wholesaler was said to be refurbishing his shop at first visit so that it was not possible to proceed to the stock count. In other cases, one wholesaler and one retailer did not stock all drugs at the shop premises but at their home so the stock count could not be performed. As during RC, data collectors questioned the accuracy of the data they had recorded during RA. Data collectors said that in some shops they counted higher quantities at SLS2 than at SLS1, although no new supplies were reportedly received. During a group discussion, a data collector explained that in one wholesale outlet, the shopkeeper had prepared an order at first visit (so quantities were not counted as 'stocked') but that a few days later the customer had cancelled the order and the shopkeeper had put the drugs back on the shelves but forgot to consider it as a new quantity received. Last but not least, fieldworkers reported being worn out by the implementation of RA, because of respondents' attitudes. *"She blamed me about what the questions asked"* (Fieldworker \#1 during a group discussion). *"I could hear that she whispered 'what the hell they come again' "* (Fieldworker \#1 during a group discussion). Discussion ========== This study compared two methods for measuring commercial outlet sales volumes, the recall and retail audit methods. Before discussing the results, some limitations should be noted. First, the samples were relatively small, notably at the wholesale level where sales volume observations were available for 34 outlets in the case of anti-malarial drugs and for 23 outlets in the case of RDTs. A second limitation is that whilst negative RA estimates clearly indicated data collection errors, and were excluded from the analysis, positive outliers may also have been errors but could not be easily identified. Third, RC estimates that were compared to RA estimates were collected at SLS2, which may have contributed to improving their accuracy as shopkeepers who expected a second interview may have paid more attention to anti-malarial/RDT sales or may have found it easier to identify the recall period because the SLS1 was a memorable event. At wholesale outlets, the analysis did not allow to conclude that on average the two methods 'agreed'. The mean difference in anti-malarial sales volume estimates between the two methods was significant and large (four AETDs, 95% CI 0.2-7.2), equivalent to about 30% of mean total sales volumes, and 66% of median total sales volumes. The limits of agreement, which provide an indication of the difference between the measurements at individual outlets, confirmed that estimates obtained through RA and RC methods were often quite different. Overall, for one third of wholesale outlets, the difference between RA and RC estimates represented as much as 50% of the total sales volumes being measured. For RDTs, the analysis of sales volumes showed that the two methods could, on average, be used interchangeably for estimating average sales: the mean difference was small (mean 1 test, 95% CI −6.0 to 4.0), equivalent to about 10% of both mean and median total sales volumes, and overall not statistically significant. However, the study lacked power to estimate the difference with sufficient precision, and the result may be a consequence of the small sample. Overall, RDT sales volume measures varied greatly at individual outlets, with differences between RA and RC of more than 50% of the volumes measured at more than 20% of wholesalers. At retail outlets, results were more difficult to interpret: bias and limits of agreement were not constant throughout the range of measurements and the between-method differences were positively correlated with volumes, with increasing differences with larger volumes being measured. Several reasons may explain the between-method differences. At wholesale level, fieldworkers reported that shopkeepers had difficulty remembering anti-malarial sales volumes as they generally stocked a wide range of other products. It is also possible that wholesalers underestimated their sales volumes during recall for fear of disclosure to competitors or regulatory authorities. As indicated in the background section, outlets with a pharmacy license were authorized to wholesale. However, at the time of the study, less than 40% of wholesalers reported holding a pharmacy license allowing them to wholesale medicines \[[@B17]\]. At retail level, the percentage of retailers holding a license allowing them to retail was somewhat higher at 66% \[[@B17]\]. Fieldworkers also experienced some challenges when implementing the RA method, during which it was not always possible to count the quantities stocked. At wholesale outlets, the SLS was implemented at the end of the supply chain survey questionnaire during which wholesalers were also asked about their business characteristics and practices and this may have created fatigue and/or anxiety amongst both respondents and fieldworkers leading to data collection errors. RA estimates might have been affected in some cases by "recall" bias for stock data, and if wholesalers had under-reported their sales volumes through RC they may well have misreported their stocks during RA. At retail outlets, fieldworkers did not report retailers to have had difficulties remembering their sales volumes but there were concerns about the accuracy of the RC data. Furthermore, the higher estimates produced by RA than RC as sales volumes increased could indicate that retailers may have had more difficulty accurately recalling larger sales volumes than smaller ones. If so, the more accurate measures produced by RA in higher sales volume contexts may justify the additional resources required for implementing this method. Based on project expenditure records, the implementation of RA that required two visits to the same outlet was, as expected, around twice the total cost of RC, amounting to US\$8,369, equivalent to US\$48 per outlet visit or US\$96 per outlet for the two RA visits combined. However, results show that the RA method is also prone to many data collection errors, and is more complex, time-consuming and invasive from the perspective of some commercial providers; these limitations are likely to jeopardize the quality of RA data collected. Stock-outs may have affected the results, with RC estimates being systematically lower than RA as observed at wholesale and retail levels. However, stock-outs may not be an important source of bias in measurement of sales volumes, as generally if a certain drug was not in stock at the time of visit one would expect its sales volumes to be zero. It is possible that sales volumes measured by the RC method may be underestimated if a product was sold during the recall period but was no longer in stock and, therefore, not asked about. Similarly for the RA method, it might be possible to have failed to record sales for a product that was stocked in between the two visits but not during either visit, though this is unlikely. The finding that the choice of method is likely to dramatically affect the size of the volumes measured has important implications if outlet-specific volumes are needed, for example in the context of an intervention rewarding individual sellers as a function of volume sold. In a study of the market for malaria treatment, this would also have implications if the objective is to measure market size in terms of volumes purchased. The feasibility and acceptability of different methods is also likely to vary across countries. Providers' willingness to recall their sales volumes and more generally participate in a medicine outlet survey may be very variable across different sociocultural contexts. For instance, during the ACTwatch supply chain survey, 5% of wholesalers in Cambodia refused to participate whilst in Zambia and Nigeria the refusal rate was 2% and 19% respectively \[[@B19],[@B30]\]. More specifically, whilst anti-malarials represented a small share of wholesale and retail providers' businesses in Cambodia, larger and highly variable volumes were handled in other ACTwatch countries. For example, the median anti-malarial volumes sold by wholesalers during the week preceding the supply chain survey ranged from 244 AETDs (IQR 26--1,104) in Benin, 689 AETDs (IQR 125--1,933) in Zambia and 1,346 AETDs (IQR 364--4,728) in Nigeria, whilst it was 0 AETD (IQR 0--0.4) in Cambodia. The suitability of different methods for measuring sales volumes is therefore likely to vary across contexts, including the type of outlets (e g, retail *vs* wholesale) and country under study (e g, malaria incidence). Conclusion ========== This study is the first that empirically compares two methods for measuring sales volumes at different anti-malarial outlet types, including both retail and wholesale outlets. It showed that at wholesale outlets the two methods did not agree for most measurements, except for RDT volumes, although the sample size was small for the latter. At retail outlets, the between-method difference was not constant throughout the range of measurement, which made the interpretation of results more difficult. Qualitative research indicated that both RA and RC methods have implementation challenges and demonstrated that the choice of empirical methods in a research project is likely to have important implications for the quality of data to be collected. Whilst the analysis did not provide firm conclusions on which method is more likely to provide more accurate sales estimates, it demonstrated that in Cambodia where sales volumes are relatively small, the RC method appeared to have key advantages: retailers were perceived to easily remember their sales volumes, wholesalers were perceived to find the method less invasive, and fieldworkers found it more convenient. The RC method was also the cheapest to implement. However, as mentioned above, sales volumes in Cambodia are low compared to other malaria-endemic countries and future research should aim to repeat such comparative analysis in contexts where anti-malarial and RDT sales volumes are larger. Competing interests =================== The authors declare that they have no competing interest. Authors' contributions ====================== EP designed the study, led the data collection, conducted the data analysis and wrote the manuscript. IK provided advice during the development of the study design and statistical data analysis. CG provided intellectual guidance on all aspects of the study, including study design, data collection and analysis. IK, BP, ST, CG, KH and KO provided comments and inputs to the manuscript and approved its final version. All co-authors are members of the ACTwatch Study, from which this study draws in terms of methods for sampling retailers and wholesalers, and the design of data collection tools for collecting sales volume data using the recall method. All authors read and approved the final manuscript. Acknowledgements ================ This study was conducted with the financial support of the ACTwatch project funded by the Bill and Melinda Gates Foundation (\#058992) and by the UK Medical Research Council through a PhD studentship awarded to the corresponding author (\# 201621). The ACTwatch project is a project of Population Services International (PSI), conducted in collaboration with the London School of Hygiene and Tropical Medicine (LSHTM). We would like to thank the ACTwatch Study Group, notably ACTwatch PSI Central team members who contributed to the ACTwatch outlet survey methods and questionnaire design, as well as the PSI-Cambodia county office, and Henrietta Allen in particular, for operational assistance with the study. We thank Rik Bosman and Prashant Yadav for their contribution to the development of the ACTwatch supply chain survey methods. EP, IK, BP, ST, CG and KH are members of the LSHTM Malaria Centre.

Introduction {#s1} ============ Does exercise enhance cognitive functioning in human beings? Meta-analyses have provided support for the beneficial effect of exercise on cognitive performance with effect sizes (*g*) ranging from 0.097 for acute exercise (Chang et al., [@B11]) to 0.158 for chronic exercise (Smith et al., [@B66]). Additionally, some authors have reported on several underlying mechanisms by considering evidence from behavioral and psychophysiological studies (for a review, see Hillman et al., [@B27]). These arguments seem to offer convincing evidence that exercise results in cognitive performance enhancement. The present study takes a critical perspective on this conclusion by assessing methodological characteristics of relevant evidence. The most relevant evidence comes from exercise-cognition randomized controlled trials (RCT). First, these RCTs are considered clinical trials. According to World Health Organization ([@B74], para. 3) and the International Committee of Medical Journal Editors (Laine et al., [@B39], p. 275), a clinical trial "is any research study that prospectively assigns human participants or groups of humans to one or more health-related interventions to evaluate the effects on health outcomes." Second, RCT is generally regarded as the best design for testing causal relationship because it makes group equivalence likely on all covariates (Freedman et al., [@B23]; Torgerson, [@B68]). Several Exercise-cognition RCTs\' findings support the causal relationship between exercise and cognition. For example, Chang et al. ([@B11]) reported a larger effect size from RCTs (*d* = 0.19) compared to those from either quasi-experimental or observational designs (*d* = −0.02 and *d* = −0.14, respectively). These results have led some authors to conclude that exercise benefits cognition in a population ranging from children to older adults. Although such message is exciting, as Rubin ([@B57]) cautioned, the relevance of evidence to answering research questions is not solely determined by the choice of research design but many other factors. Guided by this message, we examined exercise-cognition RCTs published in the past 20 years for potential methodological shortcomings. Why are errors possible ----------------------- When analyzing pretest-posttest data from RCTs, researchers typically apply two group-comparison strategies to draw causal inferences: analysis of covariance and gain score analysis (Vickers and Altman, [@B71]; Van Breukelen, [@B69]). *Analysis of Covariance* (ANCOVA)[^1^](#fn0001){ref-type="fn"} refers to the approach where posttest scores are compared between groups, adjusting for baseline scores (as covariates in the linear model). Assuming baseline group equivalence, *Analysis of Partial Variance* is a parallel of this strategy (Cohen et al., [@B13]). The alternative approach, *Gain Score Analysis* (GSA), considers the gain score (i.e., posttest minus pretest) as the criterion for group comparison. Forms of GSA include repeated-measures analysis of variance (RM ANOVA), gain score *t*-test, and ANOVA of gain score, among others. Researchers\' choice between ANCOVA and GSA often leads to disparate conclusions, an inconsistency historically termed "Lord\'s Paradox" (Lord, [@B46]). Lord\'s paradox generated a lasting research effort and a consensus was reached among methodologists. The consensus is that, as long as baseline group equivalence is likely by randomization (such as in a RCT design), investigators should choose ANCOVA in drawing causal conclusions, because ANCOVA has a higher testing power and unbiased effect estimate compared to GSA (Cronbach and Furby, [@B14]; Huck and McLean, [@B32]; Holland and Rubin, [@B29]; Miller and Chapman, [@B49]; Senn, [@B64]; Van Breukelen, [@B69]). However, when baseline group equivalence is unlikely (such as in a quasi-experimental design), none of the statistical procedures enables to "control for" such a flaw, and thus no causal inferences should be attempted (Campbell and Stanley, [@B9]; Lord, [@B46]; Cronbach and Furby, [@B14]; Meehl, [@B48]; Senn, [@B64]; Van Breukelen, [@B69]). To reiterate previous points with an analogy, perfect dishes ("causal inferences") come from fresh raw food ("baseline group equivalence") and skillful cooking ("ANCOVA"), whereas no perfect dishes can be made from non-fresh food ("baseline group non-equivalence") irrespective of how skillful the cook is. Given Lord\'s paradox conclusion, strong evidence for causal inferences can be obtained only if (a) baseline group equivalence is likely, and (b) pretest-posttest data are analyzed using ANCOVA. In practice, researchers never know with certainty that a given RCT has baseline group equivalence, but they can ascertain baseline group non-equivalence when group baseline measures show statistical differences. Assuming that baseline group equivalence is achieved by identifying no baseline group differences on any baseline measures (which is a likely portrait of a given RCT, at least on baseline measures statistically tested), researchers should choose ANCOVA over GSA when comparing groups. One advantage of ANCOVA over GSA is an increased power. Originally, ANCOVA was not developed to "control" for anything but to enhance the testing power of independent variables (Miller and Chapman, [@B49]). For instance, assuming identical within-group variance between pretest and posttest, Van Breukelen ([@B69]) quantified that ANCOVA requires only 75% of the sample size of ANOVA of gain score (i.e., one form of GSA) to detect the same effect when the pretest-posttest correlation is 0.50. The other advantage of ANCOVA over GSA has to do with effect estimate accuracy. Specifically, ANCOVA produces the unbiased effect estimate, whereas GSA can generate under- or over- estimated effect size depending on the situation of baseline group imbalance (Vickers and Altman, [@B71]). Baseline group imbalance is the descriptive difference between groups on baseline measures. If an exercise-cognition RCT has only two groups (i.e., one control and one exercise group), the control group and the exercise group have an equal chance to perform better than the other descriptively on a cognitive task at baseline. The interpretation of "better" is task specific. For instance, a shorter reaction time (RT) is better in simple reaction time tasks (e.g., Stroop Color), whereas a larger value is better in time-limited memory tasks (e.g., Digit Symbol). If the control group has baseline superiority (*control-BS*) by having, for instance, a shorter RT than that of the exercise group on the Stroop Color task, the adoption of GSA will lead to an over-estimate of exercise\'s benefits on cognition. Conversely, baseline exercise group superiority (*exercise-BS*) will generate an under-estimated effect with the GSA method (Vickers and Altman, [@B71]). Baseline measures are usually negatively correlated with gain scores (Cronbach and Furby, [@B14]; Knapp and Schafer, [@B36]), a phenomenon known as "regression to the mean" (Galton, [@B24]; Bland and Altman, [@B4]). In such instances, the bias due to GSA\'s failure to account for baseline group imbalance can be larger. As a consequence, the Type I error (i.e., false positive) from control-BS and Type II error (i.e., false negative) from exercise-BS are likely to happen when using GSA. For example, Bland and Altman ([@B5]) reported that comparing a baseline with a follow-up separately in each group by using *t*-test (i.e., one form of GSA) could raise the actual alpha level to be as high as 0.50 when comparing two groups and 0.75 when comparing three groups, depending on the power of a specific test. To make things worse, Bland and Altman\'s results were based on one outcome measure. When an exercise-cognition RCT assesses the effect of exercise on multiple cognitive measures (which is often the case), the practice of having a presumable false positive threshold (e.g., α = 0.05) could turn meaningless. How to test for possible errors ------------------------------- Rather than assessing the effect of exercise on cognition by considering potential moderators, a procedure common to meta-analytic studies, the focus of the present study was to determine whether exercise-cognition RCTs published in the past 20 years (1996--2015) involve false positives or false negatives due to GSA application in pretest-posttest data analysis. We provided a simple test to achieve this goal. Because group assignment was random, one would expect an equal chance for control-BS and exercise-BS on a certain cognitive measure. In other words, across all RCTs in our review, we expect half RCTs to show control-BS and the other half to have exercise-BS. In terms of a probability distribution, if we assume that *X* represents the number of RCTs showing control-BS, we would expect the probability of observing *X*, P (*X*), to follow a binomial distribution: $$\begin{array}{l} {\left. P{(X)} \right.\sim\text{Binomial}{({n,k})}} \\ \end{array}$$ where *n* represents the total number of RCTs examined and *k* symbolizes the expected probability (*k* = 0.5) of getting control-BS in a given exercise-cognition RCT[^2^](#fn0002){ref-type="fn"}. Similarly, if researchers select randomly between GSA and ANCOVA, we should expect the group comparison strategy to follow the same binomial distribution with the only difference being that *X* is representing the number of RCTs employing GSA. In order to detect possible false positive and/or negative errors among exercise-cognition RCTs using GSA, we must check for independence between baseline group imbalance (i.e., control-BS vs. exercise-BS) an statistical significance test result (i.e., significant vs. non-significant). If baseline group imbalance were independent to statistical significance test result, we would expect *X*, representing the number of RCTs using GSA that showed control-BS, to continue following the binomial distribution when conditioned on statistical test result. Assuming that *Y* stands for the statistical test result that has two possible outcomes (i.e., significant or non-significant), we will have the following conditional binomial distribution: $$\begin{array}{l} {\left. \text{P}{(\left. X \middle| Y \right.)}\ \right.\sim\text{~Binomial}{(\left. n \middle| Y,\ k \right.)}} \\ \end{array}$$ where *n* is the total number of RCTs using GSA method and *k* still takes the value of 0.5. To summarize, we had three hypotheses in the present study. First, we hypothesized that, among all the RCTs, half of them should demonstrate control-BS and the other half should show exercise-BS due to randomization. Second, we hypothesized that researchers, as a group, selected between GSA and ANCOVA without preference, and therefore half of the RCTs should employ GSA and the other half should use ANCOVA as a group-comparison strategy. Lastly, we hypothesized that, when GSA-RCTs are counted separately based on whether they are positive (i.e., include at least one significant finding) or negative (i.e., include no significant findings), more control-BS (than exercise-BS) GSA-RCTs should be found in positive GSA-RCTs, whereas more exercise-BS (than control-BS) GSA-RCTs should be found in negative GSA-RCTs. Methods {#s2} ======= Literature search and inclusion criteria ---------------------------------------- The second author (J.-C. L.) conducted a literature search in April and May 2015 using SPORTDiscus, Web of Science, and Google Scholar databases. The search strategy utilized the following key words within full documents: (*exercise* OR *physical activity*) AND (*cognition* OR *cognitive performance*) AND *randomized controlled trial*. A manual search of reference list from key studies (e.g., meta-analysis) was also performed. The first author (S. L.) screened studies by title and abstract, then by full documentation. Trial authors were contacted when required information was missing. In total, 38 RCTs were considered for coding. However, five articles were excluded because they were missing information and corresponding authors were unable to respond to our request by July 1, 2015. The final set of studies consisted of 33 exercise-cognition RCTs. The following inclusion criteria were applied to the exercise-cognition RCTs: (a) studies were published between January1996 and May 2015, (b) randomization is evident at the individual level, (c) the design included pre- and post-intervention measures on cognitive tasks such as perception, intelligence, academic achievement, memory, executive function, and cognitive impairment, (d) exercise intervention focused on aerobic, resistance training, or a combination of both, (e) studies included a passive control (e.g., waiting list), an active control (that can have a cognitive, physical, or social focus), or a combination of both (see Scherder et al., [@B61]), and (f) group differences were tested on cognitive measures. If multiple exercise intensities were used within an RCT, we regarded the group receiving the highest intensity as the exercise group and compared it to the control group. For example, if an RCT has two exercise groups (e.g., participants exercising at 60 and 70% of their VO~2max~) and a reading control group, the group exercising at 70% VO~2max~ was selected as the treatment group and was compared to the control group. In addition, if the two exercise groups differed in exercise modality (i.e., aerobic training and resistance training), we compared each of these exercise groups to the control group, respectively, and the results were coded under a given RCT. Furthermore, if multiple interventions were included and at least one of the groups received an intervention focusing on elements other than exercise (e.g., cognitive training), only the exercise group was considered as a treatment group and was compared to the control group. Finally, if multiple follow-up measurements were available after the intervention period, we chose the immediate post-intervention measurement as the post-test measure. Details of the literature search and study selection were shown in a flowchart (Figure [1](#F1){ref-type="fig"}). {#F1} Coding and reliability ---------------------- The first two authors discussed and settled coding variables to be included in the coding sheet. One author (S. L.) independently coded all the studies. The coded variables focused on the information relevant to the focus of the study, which is to check potential Type I and Type II errors in exercise-cognition RCTs. Therefore, for every cognitive task, we coded the targeted cognitive process (e.g., executive functioning), baseline group imbalance (control-BS vs. exercise-BS), and statistical test result (significant vs. non-significant). Other key methodological information were also coded including (a) group-comparison strategy in pretest-posttest data analysis (ANCOVA vs. GSA), (b) the form of control (passive vs. active), (c) the presence or absence of randomization procedure, (d) testing baseline group equivalence on cognitive measure(s), (e) the use of blinding procedures (i.e., single-, double-, or triple-blind), (f) explicit inclusion of intention-to-treat (ITT) analysis, (g) presence of *a priori* power analysis, (h) total participant number and number of groups (enabling participant number per group to be calculated), and (i) the presence or absence of pre-registering the trial. Table [1](#T1){ref-type="table"} displays the coded information for each study included. ###### **Study coding sequenced by group comparison strategy and study positivity**. **Authors and Year** **Grp. (T/C)** **Sig**. **Anal**. **Control** **Random** **Test Base**. **Blind** **ITT** **Power** ***N* (Grp. \#)** **Prereg**. ------------------------------------------ ---------------- ---------- ----------- ------------- ------------ ---------------- ----------- --------- ----------- ------------------- ------------- Williamson et al., [@B73] C/C N ANCOVA A-Cog. N N Single N Y 102(2) Y Scherder et al., [@B61] E/E Y ANCOVA Both N Y Single N N 43(3) N Lautenschlager et al., [@B41] E/E Y ANCOVA A-Cog. Y Y Single Y Y 170(2) Y Liu-Ambrose et al., [@B44] C/C Y ANCOVA A-Phy. Y N Single Y Y 155(3) Y Davis et al., [@B15] E/E Y ANCOVA P N N Single Y Y 171(2) Y Nagamatsu et al., [@B52] E/E Y ANCOVA A-Phy. N N Single N N 86(3) Y Okumiya et al., [@B55] E/E N GSA P N Y Single N N 42(2) N Lemmink and Visscher, [@B42] E/E N GSA A-Cog. N N N N N 16(2) N Foley et al., [@B22] E/E N GSA A-Phy. N Y N Y N 20(2) N Krogh et al., [@B38] E/E N GSA A-Phy. Y N Single Y N 165(3) Y Kimura et al., [@B35] E/E N GSA A-Cog. N Y Single N N 171(2) N Varela et al., [@B70] C/C N GSA A-Mix N N Single Y N 68(3) N Ruscheweyh et al., [@B60] C/C N GSA P N N Single N N 62(3) N Linde and Alfermann, [@B43] E/E N GSA P Y Y Single Y N 70(4) N Ruiz et al., [@B59] E/E N GSA A-Mix N Y Single Y N 40(2) N Williams and Lord, [@B72] E/E Y GSA P N Y N N N 187(2) N Emery et al., [@B18] C/C Y GSA P Y N N N N 79(2) N Erickson et al., [@B20] E/E Y GSA A-Phy. N N Single N N 120(2) N Bakken et al., [@B2] C/C Y GSA P N N N N N 15(2) N Kramer et al., [@B37] C/C Y GSA A-Phy. N N N N N 124(2) N Fabre et al., [@B21] C/C Y GSA A-Soc. N Y N N N 32(4) N Netz et al., [@B53] C/C Y GSA A-Cog. N Y Single N N 59(3) N Busse et al., [@B8] C/C Y GSA P N N N N N 31(2) N Chang and Etnier, [@B10] C/C Y GSA A-Cog. N N N N N 41(2) N Barella et al., [@B3] E/C Y GSA A-Soc. N N N N N 40(2) N Muscari et al., [@B51] C/C Y GSA A-Cog. N Y Single Y Y 120(2) N Ellemberg and St-Louis-Deschênes, [@B17] N/N Y GSA A-Cog. N N N N N 72(2) N Kamijo et al., [@B34] C/C Y GSA P N N N N N 43(2) N Chang et al., [@B12] C/C Y GSA A-Cog. N Y N N Y 42(2) N Hopkins et al., [@B31] C/C Y GSA P N N N N N 75(4) N Maki et al., [@B47] E/E Y GSA A-Cog. N Y N Y N 150(2) N Liu-Ambrose et al., [@B45] C/C Y GSA A-Phy. Y N Single Y Y 155(3) Y Hillman et al., [@B28] N/C Y GSA P Y N Single Y Y 221(2) Y *Year, Year of publication; Grp, (T/C), Baseline group imbalance (total count/conditional count); Sig., Study positivity (at least one significant test result identified by corresponding RCT); Anal., Group comparison strategy in pretest-posttest data analysis; Control, Form of control group; Random, Described random allocation procedures; Test Base, Tested baseline group equivalence on cognitive measures; Blind, Blinding procedures reported; ITT, Explicitly mentioned following intention-to-treat principle; Power, Performed a priori power analysis; N (Grp.), Total sample size (number of groups); Prereg., Pre-registered the trial. Liu-Ambrose et al. ([@B45]) reported data dependence with Liu-Ambrose et al. ([@B44]); E, Exercise-BS; C, Control-BS; Y, Yes; N, No; GSA, Gain score analysis; ANCOVA, Analysis of covariance; A-Cog., Active control with a cognitive focus; A-Phy., Active control with a physical focus; A-Soc., Active control with a social focus; A-Mix, Active control with more than one focus (e.g., cognitive and social); P, Passive control, Both, A control group consisting both actively and passively controlled participants; Single, Single blinding procedure (i.e., cognitive task assessors)*. Eleven articles (33.3% of total) were randomly selected and separately coded to produce inter-coder reliability. A research assistant blinded to the study purposes completed the coding. Inter-rater reliability was calculated using Cohen\'s *Kappa* coefficient for each coding variable (Table [2](#T2){ref-type="table"}). Following Landis and Koch\'s ([@B40]) recommendations, we considered *Kappa* values between 0.61 and 0.80 as substantial and above 0.80 as very good. All the coded variables in the present study showed very good reliability. Coding discrepancies were resolved by re-visiting studies and discussion. ###### **Kappa coefficients for coding variables**. **Coding Variable** **Kappa** ------------------------------------------------------------ ----------- Cognitive task 1.00 Baseline group imbalance (Control vs. Exercise) 0.92 Group difference results (significant vs. non-significant) 1.00 Group comparison strategy (GSA vs. ANCOVA) 0.85 Form of control 1.00 Description of randomization 1.00 Baseline group equivalence test on cognitive measures 1.00 Description of blinding 0.80 Intention-to-treat principle (ITT) 1.00 *A priori* power analysis 1.00 Total participant number and number of groups 1.00 Trial pre-registration 1.00 RCT count and statistical analysis ---------------------------------- We categorized and counted all the RCTs regarding their group-comparison strategy and baseline group imbalance. For group-comparison strategy, we categorized a given RCT into GSA-RCT if it used *gain scores* as the criterion in comparing groups. We classified an RCT as ANCOVA-RCT if the outcome variable was the post-test score while controlling for baseline score as covariate, or if analysis of partial variance was used. Although we coded baseline group imbalance for every cognitive task within an RCT, we later counted the number of RCT regarding their baseline group imbalance favorableness (control-BS vs. exercise-BS). This ensured an equal weight for every RCT given their varying number of cognitive measures. For example, one RCT reported 42 cognitive measures but several RCTs reported only one cognitive measure. In this case, the 42-task RCT would be over-weighted if the count were made at the task level. We applied the "dominance rule" in judging whether a given RCT favors control-BS or exercise-BS. For example, if an RCT used four cognitive measures, we coded it as favoring control-BS if three of the four measures had better performing control group at baseline. Due to within-study measurement dependence, multiple cognitive measures tended to show homogeneous results with respect to baseline group imbalance. Among 33 RCTs, we applied the dominance rule to 14 RCTs. Two RCTs showed equal number of cognitive measures between control-BS and exercise-BS, and thus were dropped from the final count on baseline group imbalance. We also made "conditional count" among GSA-RCTs. First, all the RCTs were screened for GSA employment. Then, GSA-RCTs were categorized as either positive (i.e., having at least one significant finding) or negative (i.e., having no significant findings). The "conditional count" process was very similar to the previous count except that a RCT\'s baseline group imbalance was decided only on those cognitive measures fitting the positive/negative category. Specifically, if a GSA-RCT had at least one significant result (i.e., positive study), its baseline group imbalance was determined on all significant cognitive measures. If a GSA-RCT had no significant results (i.e., negative study), all its cognitive measures were included to determine its baseline group imbalance. These decisions were made for two reasons. First, some positive RCTs employed only one cognitive task (which reached statistical significance). Second, we could bias the negative RCT count regarding baseline group imbalance if we retained the non-significant measures from positive RCTs and recycled them in the negative RCT count. During the "conditional count," we applied the dominance rule to only one GSA-RCT because it included one cognitive measure supporting control-BS and one cognitive measure with description-wise equal baseline between the control and exercise group; and thus it was counted as control-BS. In addition, one positive GSA-RCT reported a control-BS on one cognitive measure and exercise-BS on the other cognitive measure. This RCT was subsequently classified as neutral and was dropped from the final conditional count. We used the R version 3.2.0 (R Core Team, [@B56]) to estimate the probability of obtaining those counts based on continuity-corrected binomial distributions. Whereas the first two hypotheses had two-sided tests, the third hypothesis had one-sided test. The alpha level was set at 0.05. Results {#s3} ======= Table [3](#T3){ref-type="table"} summarizes results pertaining to the first two hypotheses. The first hypothesis assumed that the occurrence of control-BS and exercise-BS are equally likely. Among all the RCTs (*n* = 31), we observed that 16 RCTs resulted in a control-BS and 15 RCTs in an exercise-BS (two RCTs were dropped in the count because they showed no clear favorableness between control-BS and exercise-BS). The probability of detecting this result met our expectation, $\hat{k}$ = 0.52, *p* = 0.99, with a 95% CI of (0.33, 0.69). The second hypothesis assumed that the incidence of GSA and ANCOVA as a group comparison strategy are equal among RCTs. The count revealed 27 GSA-RCTs and 6 ANCOVA-RCTs. The test of such occurrence reached significance, $\hat{k}$ = 0.82, *p* \< 0.001, with a 95% CI of (0.64, 0.92). Therefore, we rejected the second hypothesis and concluded that researchers predominantly used GSA over ANCOVA in analyzing pretest-posttest data. ###### **The probability of observed RCT counts regarding baseline group imbalance and group comparison strategy**. **Group (*****N*** = 31**)** **Strategy (*****N*** = 33**)** ----------------------------------------- ------------------------------ --------------------------------- ---- --- RCT Count 16 15 27 6 $\overset{\hat{}}{\text{k}}$ (95% C.I.) 0.52 (0.33, 0.69) 0.82 (0.64, 0.92) *p* 0.99 \<0.001 *Group, Baseline group imbalance; Control, Control-BS; Exercise, Exercise-BS; Strategy, Group-comparison strategy used in pretest-posttest data analysis; GSA, Gain score analysis; ANCOVA, Analysis of covariance*. Table [4](#T4){ref-type="table"} displays results for the third hypothesis, which tested independence between baseline group imbalance and statistical significance test result among GSA-RCTs. Among positive GSA-RCTs (*n* = 17), 14 resulted in a control-BS and three in exercise-BS. This pattern reached significant level, $\hat{k}$ = 0.82, *p* = 0.006, with a 95% CI of (0.60, 1.00). Among the negative GSA-RCTs (*n* = 9), two studies had a control-BS and seven had exercise-BS. This observation was not significant, $\hat{k}$ = 0.22, *p* = 0.09, with a 95% CI of (0.00, 0.55). Thus, baseline group imbalance was related to statistical test in that more control-BS GSA-RCTs (which had over-estimated effect sizes) than exercise-BS GSA-RCTs resulted in significant results. ###### **The probability of observed conditional count on GSA-RCTs regarding baseline group imbalance**. **Positive (*****n*** = 17**)** **Negative (*****n*** = 9**)** ----------------------------------------- --------------------------------- -------------------------------- --- --- RCT Count 14 3 2 7 $\overset{\hat{}}{\text{k}}$ (95% C.I.) 0.82 (0.60, 1.00) 0.22 (0.00,0.55) *p* 0.006 0.09 *Positive, GSA-RCTs identifying at least one significant finding; Negative, GSA-RCTs identifying no significant findings; Control, Control-BS; Exercise = Exercise-BS*. Discussion {#s4} ========== The objective of the present study was to determine whether exercise-cognition RCTs published in the past 20 years (1996--2015) include false positives or false negatives due to the ignorance of Lord\'s paradox (i.e., performing GSA in analyzing pretest-posttest data). Overall, several findings emerged from this study. First, baseline group superiority was found to be randomly determined among all the RCTs, with an equal probability of control-BS and exercise-BS. Second, GSA was the more popular group comparison strategy (27 RCTs) compared to ANCOVA (6 RCTs). Lastly, evidence suggested that positive GSA-RCTs were likely to include false positive errors because 82% (14 out of 17 studies) of them tested on over-estimated effect sizes. However, no clear evidence supported false negative errors among negative GSA-RCTs although a descriptive consistency was revealed. Given findings that GSA is prevalent and misleading, it is necessary to re-emphasize the adoption of ANCOVA in pretest-posttest data analysis. The employment of ANCOVA could eliminate the biased effect estimate due to baseline group imbalance and increase testing power, thus reducing inferential errors. However, choosing ANCOVA as group comparison strategy is only half the story because ANCOVA enhances causal inferences only when group equivalence is likely. The other half, baseline group equivalence, depends on multiple factors during the experimental process. Some important factors are discussed next. Randomization procedures ------------------------ One factor influencing group equivalence is randomization procedure. According to Schulz ([@B62]), randomization consists of two stages: generation of unpredictable assignment sequence and concealment of that sequence until group allocation occurs. The first stage is related to the reliability of the randomizing tool (e.g., computer algorithm), and is often mistakenly identified as randomization itself. Consequently, sequence-concealment often receives insufficient attention, which introduces bias that emerges from the predictability of participant allocation. Ideally, the information on participant allocation should be revealed "as late as possible." As an example, Newell ([@B54]) reported an anecdotal story of a surgeon who tosses a sterilized coin after a patient\'s abdomen was opened to decide which "treatment" he should perform. Although a little extreme, it highlights the importance of concealing participants\' allocation information from experimenters. Table [1](#T1){ref-type="table"} shows that only 7 out of 33 RCTs described randomization tools and even fewer RCTs described sequence-concealment procedures. In a couple of occasions, the randomization was done with imbalanced assignment ratio (e.g., 2:1 in assigning participants to exercise and control group, respectively) and no justifications were offered. Therefore, it is encouraged to report the randomization tool and to describe procedures for concealing the randomization sequence. In cases of imbalanced group assignment ratios, justifications are required. Baseline check -------------- Prior to intervention, researchers must examine group equivalence on baseline measures. To foster such an examination, the CONSORT (Consolidated Standards of Reporting Trials) statement (Schulz et al., [@B63]) suggests reporting baseline data of demographic and clinical characteristics for each group. Concerning the CONSORT statement and the difficulty in conducting double-blind trials in exercise-cognition area, we recommend researchers to examine baseline group equivalence using both significance tests and subjective judgments. Baseline significance tests can alert researchers to factors interfering with randomization (e.g., no double-blinding); even when no significant group differences are identified at baseline, researchers must still review descriptive group imbalance on its size and prognostic strength (Altman, [@B1]). If meaningful group differences are found on any of the baseline measures (regardless of test significance), researchers could take different approaches in solving the problem, depending on how many baseline measures showed group differences. For instance, researchers can block participants when only few baseline measures (i.e., one or two) showed group differences in baseline check, or can re-randomize participants when more baseline variables exhibited group differences (Rubin, [@B58]). Single-blinding and differential expectation -------------------------------------------- Blinding procedure also affects group equivalence. When participants were assigned to either exercise or control group, it was challenging (if not impossible) to blind them to their respective interventions. In the present review, 18 out of the 33 RCTs reported blinding procedures and all of them were "single-blinded" (i.e., cognitive task assessors were blinded to participants\' group assignment). No RCTs reported blinding participants to their group assignments. This raises the concern that participants may show differential expectations due to open group assignment. Such a possibility is consistent with the idea of "unmatched task" for the control group in the literature dealing with the effect of exercise on cognition (Brisswalter et al., [@B7]). The concern of differential expectation can also be evidenced by the diversity of control conditions in Table [1](#T1){ref-type="table"}. This diversity reveals little agreement among researchers in speculating an active control for exercise intervention. To help select and/or design a good control, we recommend an empirical solution. That is, researchers should measure differential expectation. Although, preliminary effort has been made to survey differential group expectations prior to intervention (e.g., Stothart et al., [@B67]), we echoed Boot et al. ([@B6]) in suggesting future research to consider testing differential expectation either during or after the intervention period. The optimal active control of exercise intervention must equate expectations on all these periods. Intention-to-treat principle ---------------------------- Intention-to-Treat (ITT) is a widely accepted principle in analyzing clinical trials. ITT prevents group non-equivalence due to participant dropout (e.g., differential attrition) by including all the randomized participants in data analysis based on their intended treatment assignment (Gillings and Koch, [@B26]). The ideal situation for ITT would be having complete data for all the randomized participants (Hollis and Campbell, [@B30]). However, attrition is typically inevitable for clinical trials. In order to include participants with incomplete data into the analysis, missing values need to be handled. Some missing value imputation methods are available. For example, methods based on multiple imputation or maximum likelihood are generally recommended, but special considerations must be given to specific situations (Enders, [@B19]). However, no statistical methods can perfectly fix experimental flaws. When applying ITT, it is necessary to develop protocols (e.g., excluding likely exercise-intolerant participants before randomization) to ensure that participant adherence rate is roughly 80% or higher (Gillings and Koch, [@B26]; Montori and Guyatt, [@B50]). Regardless of adherence rate for a given RCT, a sensitivity test should always be performed to compare the ITT analysis results (as primary outcome) with the complete-case analysis results (Gillings and Koch, [@B26]). Compatible result of the sensitivity test precludes the concern of differential attrition, whereas incompatibility suggests this threat to internal validity. In short, future investigations are advised to include protocols that maximize adherence rate, to follow ITT principle, and to perform sensitivity analysis. Two other important elements of clinical trials are discussed next, although they do not affect group equivalence directly. Power ----- Despite that no clear evidence of false negative errors was observed in the present study, it was still important to make sure that each RCT has sufficient power so that false negative errors could be minimized. Among all the RCTs included, only eight of 33 RCTs reported performing an *a priori* power analysis. Depending on the inputted parameters, the sample sizes varied among these RCTs. However, the average group size among the RCTs with *a priori* power analysis was about 65 participants, whereas the average group size for those not performing an a priori power analysis was about 32 participants[^3^](#fn0003){ref-type="fn"}. It seems that a substantial proportion of exercise-cognition RCTs was underpowered, and thus could lead to false negative errors. It might be argued that 23 out of 33 included RCTs had at least one significant result, and thus false negative errors should not be a concern. However, 23 out of 33 RCTs having at least one positive result is not an evidence of sufficient power. First, we showed that false positive errors are likely to be included in those 17 positive GSA-RCTs, and by extension in the 23 positive RCTs. Second, as highlighted by Rubin ([@B57]), a poorly implemented experiment can maintain many errors and ultimately be irrelevant to testing the research question. An experiment should follow optimal procedures (including *a priori* power analysis) for its conclusions to appropriately address research questions. Researcher degrees of freedom and trial pre-registration -------------------------------------------------------- Although researchers are following the best paradigm including fixed set of practices, they still make decisions on quite some circumstances. These decision-calling circumstances are regarded as the *researcher degrees of freedom* (Simmons et al., [@B65]). It includes, among others, types of measure used in data collection, group-comparison strategies employed for data analysis, and type of data reported. When considering the researcher degrees of freedom with publication bias, an increased likelihood of Type I error would follow. For example, Gelman and Loken ([@B25]) argued that data analysis strategies could be unwittingly conditioned on data patterns, which allow for false positive findings. To restrict researcher degrees of freedom by increasing clinical trial transparency, the International Committee of Medical Journal Editors (ICMJE) declared a trial\'s pre-registration as a condition for publishing in its 11 member journals in 2004 (De Angelis et al., [@B16]). ICMJE only recognizes registries meeting several criteria, including being free to public access, electronically searchable, open to all registrants, run by not-for-profit organization, as well as able to ensure validity of registration data by offering a mechanism. For example, [www.clinicaltrials.gov](http://www.clinicaltrials.gov) maintained by the U.S. National Institute of Health is a qualified registry, even though many other registries have become available since 2004 (Humphreys et al., [@B33]) maintained by the U.S. National Institute of Health is a qualified registry, even though many other registries have become available since 2004 (Humphreys et al., [@B33]). It is by revealing critical trial information before participant enrollment that trial pre-registration combats researcher degrees of freedom. By pre-registering trials, researchers can still make changes afterwards as long as they offer good justifications. Although pre-registration has been the rule in clinical trial publication for almost 10 years (Laine et al., [@B39]), it is not true among exercise-cognition RCTs because only 8 out of 27 studies published in 2005 and later had trial pre-registration (Table [1](#T1){ref-type="table"}). Therefore, we recommend future exercise-cognition RCTs to follow ICMJE\'s guidelines and make trial pre-registrations before enrolling participants. Limitations ----------- Several limitations in the present study are worth pointing out. First, we only focused on group comparison strategies in analyzing pretest-posttest data in exercise-cognition RCTs because it generates good evidence to evaluate the claim that exercise benefits cognition, and it is a design shared by all the exercise-cognition RCTs. Second, although ANCOVA should be used in analyzing pretest-posttest data in RCTs given group equivalence, it should be noted that ANCOVA was developed under several statistical assumptions, among which the assumption of homogeneity of regression slopes should receive particular attention (Miller and Chapman, [@B49]). However, these assumptions should not be used as an excuse to choose GSA against ANCOVA because GSA shares the same set of assumptions and because of ANCOVA\'s robustness and flexibility under assumption violation (Huck and McLean, [@B32]). Lastly, the counting process may have introduced bias in our conclusions, especially for the conditional count. We made the counts at trial level rather than at task level, and thus applied the "dominance rule" in order to maintain equal weight among exercise-cognition RCTs. Even though a better approach may be possible, evidence supported our decision. For example, we applied the "dominance rule" only to a minority of collected RCTs and the marginal count met the exact expectation from a probability point of view. Among the 33 RCTs, only two RCTs switched the group regarding baseline superiority between the marginal count and the conditional count. Conclusion {#s5} ========== Although exercise-cognition RCTs showed randomness of baseline group imbalance, RCTs adopting GSA as group comparison strategy were likely to have false positive errors and thus weakened the overall exercise-benefit-cognition claim. Future research will benefit from employing ANCOVA in analyzing pretest-posttest data while maintaining baseline group equivalence. Several suggestions have been offered to maintain baseline group equivalence in future research. It is likely that the results of current study are not limited to the effect of exercise on cognition and could potentially be extended to RCTs in other domains. Author contributions {#s6} ==================== Conceived and designed the study: SL, JL. Searched publications: JL. Screened publications, coded data, and analyzed results: SL. Calculated inter-rater reliability: JL. Contributed to the writing of this manuscript: SL, JL, GT. Conflict of interest statement ------------------------------ The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The authors would like to thank Dr. Yu-Kai Chang and Dr. Walter R. Boot for their review of the initial draft of this paper. ^1^In this paper, the key distinction between ANCOVA and GSA is how researchers use the baseline measure. Although researchers can choose variables (e.g., age) as covariates in testing group difference on gain scores, these analyses are not what we mean by ANCOVA here. ^2^We chose *k* instead of *p* to avoid confusion later when reporting the probability of our hypothesis testing. ^3^This information was calculated based on the "*N* (Grp.)" column of Table [1](#T1){ref-type="table"}. ^\*^References marked with an asterisk indicate studies included in Table [1](#T1){ref-type="table"}. [^1]: Edited by: Jason C. Immekus, University of Louisville, USA [^2]: Reviewed by: Evgueni Borokhovski, Concordia University, Canada; Daniel Saverio John Costa, University of Sydney, Australia [^3]: This article was submitted to Quantitative Psychology and Measurement, a section of the journal Frontiers in Psychology

Related literature {#sec1} ================== For the structure at 200 K, see: Seidel *et al.* (2010[@bb10]). For the 2-chloro­phenol solvate of cyclic tetra­meric ZnTPyP, see: Lipstman & Goldberg (2010[@bb8]). For a review article on structural motifs in coordination polymers of the 5,10,15,20-tetra­4-pyrid­ylporphyrin ligand, see: DeVries & Choe (2009[@bb5]). For the supra­molecular chemistry of ZnTPyP in the solid-state, see: Lipstman & Goldberg (2010[@bb8]); Seidel *et al.* (2010[@bb10]) and references cited therein. For a description of the IμS microfocus X-ray source used in the present study, see: Graf (2008[@bb7]); Schulz *et al.* (2009[@bb9]). For *PLATON* / *SQUEEZE*, see: van der Sluis & Spek (1990[@bb12]); Spek (2009[@bb13]). For a description of the program *COOT*, see: Emsley *et al.* (2010[@bb6]). Experimental {#sec2} ============ {#sec2.1} ### Crystal data {#sec2.1.1} \[Zn~4~(C~40~H~24~N~8~)~4~\]·8C~3~H~7~NO·3H~2~O*M* *~r~* = 3366.98Tetragonal,*a* = 23.6897 (5) Å*c* = 14.9876 (7) Å*V* = 8411.1 (5) Å^3^*Z* = 2Cu *K*α radiationμ = 1.24 mm^−1^*T* = 100 K0.16 × 0.04 × 0.02 mm ### Data collection {#sec2.1.2} Bruker X8 PROSPECTOR diffractometerAbsorption correction: multi-scan (*SADABS*; Bruker, 2008[@bb3]) *T* ~min~ = 0.827, *T* ~max~ = 0.97644415 measured reflections7723 independent reflections6768 reflections with *I* \> 2σ(*I*)*R* ~int~ = 0.018 ### Refinement {#sec2.1.3} *R*\[*F* ^2^ \> 2σ(*F* ^2^)\] = 0.042*wR*(*F* ^2^) = 0.108*S* = 1.047723 reflections442 parametersH-atom parameters constrainedΔρ~max~ = 0.59 e Å^−3^Δρ~min~ = −0.42 e Å^−3^ {#d5e709} Data collection: *APEX2* (Bruker, 2008[@bb3]); cell refinement: *SAINT* (Bruker, 2010[@bb4]); data reduction: *SAINT*; program(s) used to solve structure: *SHELXS97* (Sheldrick, 2008[@bb11]); program(s) used to refine structure: *SHELXL97* (Sheldrick, 2008[@bb11]); molecular graphics: *DIAMOND* (Brandenburg, 2010[@bb2]); software used to prepare material for publication: *enCIFer* (Allen *et al.*, 2004[@bb1]). Supplementary Material ====================== Crystal structure: contains datablocks global, I. DOI: [10.1107/S1600536811002054/bv2170sup1.cif](http://dx.doi.org/10.1107/S1600536811002054/bv2170sup1.cif) Structure factors: contains datablocks I. DOI: [10.1107/S1600536811002054/bv2170Isup2.hkl](http://dx.doi.org/10.1107/S1600536811002054/bv2170Isup2.hkl) Additional supplementary materials: [crystallographic information](http://scripts.iucr.org/cgi-bin/sendsupfiles?bv2170&file=bv2170sup0.html&mime=text/html); [3D view](http://scripts.iucr.org/cgi-bin/sendcif?bv2170sup1&Qmime=cif); [checkCIF report](http://scripts.iucr.org/cgi-bin/paper?bv2170&checkcif=yes) Supplementary data and figures for this paper are available from the IUCr electronic archives (Reference: [BV2170](http://scripts.iucr.org/cgi-bin/sendsup?bv2170)). The Deutsche Forschungsgemeinschaft (DFG) is acknowledged for financial support. RWS is grateful to Professor William S. Sheldrick and Professor Christian W. Lehmann for generous support. Comment ======= 5,10,15,20-Tetra(4-pyridyl)porphyrin has been widely used as ligand for the construction of coordination polymers (DeVries & Choe, 2009). We and others have reported on the solid-state supramolecular chemistry of the self-complementary \[5,10,15,20-tetra(4-pyridyl)porphyrinato\]zinc(II) (ZnTPyP) building block (Lipstman & Goldberg, 2010; Seidel *et al.*, 2010 and references cited therein). Recently, we reported the title structure of \[ZnTPyP\]~4~. The small dark red plate-shaped crystals of the title compound were subjected to diffraction experiments using a Bruker AXS X8 PROSPECTOR diffractometer equipped with an INCOATEC microfocus X-ray source (IµS) for Cu radiation (Graf, 2008). Such microfocus X-ray sources use multilayer mirrors to focus the X-ray beam onto the crystal and, therefore, lead to a significant reduction of the background and an increase in diffracted intensities. It has already been demonstrated that the Mo IµS gives data of significantly higher quality than a 2 kW Mo fine focus sealed tube, when small crystals are examined (Schulz *et al.*, 2009). The data collection presented here, using the Cu IµS, resulted in intensity data of surprisingly good quality and, hence, indicated a re-refinement of the crystal structure. The crystals investigated in the original work were significantly larger than those examined in the present study and split on cooling to 100 K. For this reason, the data were collected at 200 K with a Cu rotating anode system at that time. Using small crystals has the advantage that these are less likely to split on flash cooling. The molecular structure of \[ZnTPyP\]~4~ is depicted in Fig. 1. The asymmetric unit contains one ZnTPyP unit (Fig 2.) and the *S*~4~ symmetric tetramer is generated by crystallographic fourfold rotoinversion symmetry. One peripheral pyridyl group binds to the central Zn atom of an adjacent symmetry related ZnTPyP unit. Zn1 is pentacoordinated and is displaced from the N~4~ mean plane by 0.3196 (9) Å. The coordination geometry parameters about Zn1 are given in Table 1. The three remaining pyridyl groups are non-coordinating. Even at 100 K, the pyridyl groups attached to C5 and C15 show elongated ellipsoids, which cause a checkCIF B level alert (Spek, 2009) due to large *U*~eq~(max)/*U*~eq~(min) ratio. This reveals that the disorder is rather of static than dynamic nature. Attempts were made to describe the electron density of the pyridyl ring attached to C15 (Fig. 3) by a split model. However, the refinement results could not be improved thereby. Thus, both pyridyl rings were finally described with large displacement parameters. In the crystal, the \[ZnTPyP\]~4~ entities are stacked into columns located at *x* = 1/4, *y* = 1/4 and *x* = 3/4, *y* = 3/4 (Fig 4). The stacking propagates *via* C~β~---H···N~py~ interactions (see Table 2) by translational symmetry in the *c* axis direction. Within a column, the distance between the centroids of the pyridyl rings attached to C5 and C15^iii^ is 4.0714 (1) Å. Adjacent columns of \[ZnTPyP\]~4~ are arranged with an offset of *c*/2 (*ca* 7.49 Å). Interstitial channels are formed parallel to the *c* axis direction centred at *x* = 1/4, *y* = 3/4 and *x* = 3/4, *y* = 1/4 (Fig 5). The potential solvent accessible void estimated with *PLATON* / *SOLV* (Spek, 2009) is 33.2% of the unit cell volume. On cooling to 100 K, the *a* lattice vector is shortened by approximately 0.27 Å in comparison to the tetragonal unit cell at 200 K (*a* = 23.958 (2) Å), whereas the length of *c* lattice vector remains relatively unaffected (*c* = 15.0646 (16) Å at 200 K; Seidel *et al.*, 2010). Despite intensive efforts, the disordered solvent molecules filling the voids within the columns of \[ZnTPyP\]~4~ and the interstitial channels could not be modeled reasonably with the data collected at 100 K. Nevertheless, residual electron density was visible in a difference Fourier synthesis calculated for the solvent regions (Fig. 6) with phases based on the model using *COOT* (Emsley *et al.*, 2010). For the visualization of the surface of the (difference) electron density using a three-dimensional mesh, the electron densities should be read into *COOT* in terms of structure factors. To obtain a structure factor (.fcf) file containg the informations necessary for the calculation of electron density maps and suitable for *COOT*, the LIST 6 instruction of *SHELXL-97* was used. The atomic model of the framework was read into *COOT* by means of the *SHELXL-97*. res file. The visual inspection of the difference electron density map indicates that four molecules of dimethylformamide (DMF) plus one water molecule are located within the voids in the columns approximately centred at (1/4,1/4,0), whereas another four molecules of DMF and two water molecules are clustered around the 4~2~ screw axes running through the interstitial channels parallel to the *c* axis direction. The compound can, therefore, probably best be described as \[ZnTPyP\]~4~. 8 DMF. 3 H~2~O. The compound was originally formulated as being a pure DMF solvate (Seidel *et al.*, 2010). To improve the fit of the model to the data and, hence, the precision of the main part of the structure, the contributions of the disordered solvent molecules were removed from the diffraction data with *PLATON* / *SQUEEZE* (van der Sluis & Spek, 1990; Spek, 2009). *SQUEEZE* estimated the electron counts in the voids within the columns and interstitial channels of \[ZnTPyP\]~4~ to be 182 and 207, respectively. These values are relatively close to those based on the proposed chemical formula (178 and 196). Experimental {#experimental} ============ Small dark red plate-shaped crystals of the title compound were obtained similarly as reported previously (Seidel *et al.*, 2010); 12 mg of ZnTPyP (Aldrich) and 11 mg of \[Pd(NO~3~)~2~(en)\] (en = 1,2-diaminoethane) were placed in an ampoule and 4 ml of DMF were added. The ampoule was sealed and placed in a heater. The sample was heated to 150 °C in 24 h and held for five days at this temperature. Subsequently, the sample was cooled down to room temperature in 100 h. Noteworthy, the crystals of the title compound were accompanied by crystals of the triclinic phase, containing a polymeric one-dimensional ladder structure of ZnTPyP, as observed previously (Seidel *et al.*, 2010). Refinement {#refinement} ========== For the final refinement, the contributions of severely disordered DMF and water molecules of crystallization were removed from the diffraction data with *PLATON* / *SQUEEZE* (van der Sluis & Spek, 1990; Spek, 2009), see comment. H atoms were placed at geometrically calculated positions and refined with constrained C---H bond length of 0.95 Å and *U*~iso~(H) = 1.2 *U*~eq~(C) allowing them to ride on the parent C atom. Figures ======= {#Fap1} ![Displacement ellipsoid plot of one repeat unit of cyclic \[ZnTPyP\]4 drawn at 50% probability. H atoms are omitted for clarity. Symmetry code: (i) y, -x + 1/2, -z + 1/2.](e-67-0m236-fig2){#Fap2} {#Fap3} ![Stacking of the \[ZnTPyP\]4 entities viewed along the a axis direction. H atoms are omitted for clarity. Cβ---H···Npy interactions are represented by dashed lines.](e-67-0m236-fig4){#Fap4} {#Fap5} {#Fap6} Crystal data {#tablewrapcrystaldatalong} ============ ------------------------------------------------- -------------------------------------- \[Zn~4~(C~40~H~24~N~8~)~4~\]·8C~3~H~7~NO·3H~2~O *D*~x~ = 1.329 Mg m^−3^ *M~r~* = 3366.98 Cu *K*α radiation, λ = 1.54178 Å Tetragonal, *P*4~2~/*n* Cell parameters from 130 reflections Hall symbol: -P 4bc θ = 3.5--31.5° *a* = 23.6897 (5) Å µ = 1.24 mm^−1^ *c* = 14.9876 (7) Å *T* = 100 K *V* = 8411.1 (5) Å^3^ Plate, dark red *Z* = 2 0.16 × 0.04 × 0.02 mm *F*(000) = 3500 ------------------------------------------------- -------------------------------------- Data collection {#tablewrapdatacollectionlong} =============== ------------------------------------------------------------ -------------------------------------- Bruker X8 PROSPECTOR goniometer diffractometer 7723 independent reflections Radiation source: Incoatec IµS microfocus X-ray source 6768 reflections with *I* \> 2σ(*I*) Incoatec Quazar Multilayer Mirror *R*~int~ = 0.018 Detector resolution: 8.33 pixels mm^-1^ θ~max~ = 69.2°, θ~min~ = 2.6° ω scans *h* = −28→25 Absorption correction: multi-scan (*SADABS*; Bruker, 2008) *k* = −24→28 *T*~min~ = 0.827, *T*~max~ = 0.976 *l* = −17→14 44415 measured reflections ------------------------------------------------------------ -------------------------------------- Refinement {#tablewraprefinementdatalong} ========== ------------------------------------- ------------------------------------------------------------------------------------------------- Refinement on *F*^2^ Primary atom site location: structure-invariant direct methods Least-squares matrix: full Secondary atom site location: difference Fourier map *R*\[*F*^2^ \> 2σ(*F*^2^)\] = 0.042 Hydrogen site location: inferred from neighbouring sites *wR*(*F*^2^) = 0.108 H-atom parameters constrained *S* = 1.04 *w* = 1/\[σ^2^(*F*~o~^2^) + (0.0472*P*)^2^ + 5.0454*P*\] where *P* = (*F*~o~^2^ + 2*F*~c~^2^)/3 7723 reflections (Δ/σ)~max~ \< 0.001 442 parameters Δρ~max~ = 0.59 e Å^−3^ 0 restraints Δρ~min~ = −0.42 e Å^−3^ ------------------------------------- ------------------------------------------------------------------------------------------------- Special details {#specialdetails} =============== ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Geometry. All e.s.d.\'s (except the e.s.d. in the dihedral angle between two l.s. planes) are estimated using the full covariance matrix. The cell e.s.d.\'s are taken into account individually in the estimation of e.s.d.\'s in distances, angles and torsion angles; correlations between e.s.d.\'s in cell parameters are only used when they are defined by crystal symmetry. An approximate (isotropic) treatment of cell e.s.d.\'s is used for estimating e.s.d.\'s involving l.s. planes. Refinement. Refinement of *F*^2^ against ALL reflections. The weighted *R*-factor *wR* and goodness of fit *S* are based on *F*^2^, conventional *R*-factors *R* are based on *F*, with *F* set to zero for negative *F*^2^. The threshold expression of *F*^2^ \> σ(*F*^2^) is used only for calculating *R*-factors(gt) *etc*. and is not relevant to the choice of reflections for refinement. *R*-factors based on *F*^2^ are statistically about twice as large as those based on *F*, and *R*- factors based on ALL data will be even larger. ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Fractional atomic coordinates and isotropic or equivalent isotropic displacement parameters (Å^2^) {#tablewrapcoords} ================================================================================================== ------ --------------- --------------- --------------- -------------------- -- *x* *y* *z* *U*~iso~\*/*U*~eq~ Zn1 0.350673 (11) 0.520064 (11) 0.185590 (16) 0.03823 (9) N21 0.38601 (8) 0.54998 (7) 0.06850 (10) 0.0433 (4) N22 0.27781 (7) 0.50134 (7) 0.11541 (10) 0.0404 (4) N23 0.41242 (7) 0.56428 (7) 0.25451 (10) 0.0404 (4) N24 0.30440 (7) 0.51392 (7) 0.30222 (10) 0.0377 (4) C1 0.46288 (9) 0.58336 (8) 0.22081 (13) 0.0420 (4) C2 0.50033 (10) 0.59793 (10) 0.29314 (14) 0.0532 (6) H2 0.5379 0.6115 0.2880 0.064\* C3 0.47175 (10) 0.58861 (10) 0.36945 (15) 0.0534 (6) H3 0.4853 0.5949 0.4283 0.064\* C4 0.41675 (9) 0.56739 (9) 0.34544 (13) 0.0428 (5) C5 0.37497 (9) 0.55138 (9) 0.40720 (13) 0.0442 (5) C6 0.32280 (9) 0.52641 (9) 0.38657 (12) 0.0411 (4) C7 0.28126 (9) 0.50864 (9) 0.45108 (13) 0.0451 (5) H7 0.2835 0.5130 0.5140 0.054\* C8 0.23852 (9) 0.48459 (9) 0.40491 (13) 0.0422 (4) H8 0.2052 0.4684 0.4292 0.051\* C9 0.25323 (8) 0.48825 (8) 0.31165 (12) 0.0362 (4) C10 0.21846 (8) 0.46994 (8) 0.24113 (12) 0.0360 (4) C11 0.22974 (8) 0.47834 (8) 0.14991 (12) 0.0376 (4) C12 0.19052 (9) 0.46543 (9) 0.07858 (13) 0.0460 (5) H12 0.1543 0.4486 0.0845 0.055\* C13 0.21544 (10) 0.48202 (10) 0.00230 (14) 0.0526 (6) H13 0.1998 0.4794 −0.0559 0.063\* C14 0.26979 (9) 0.50435 (10) 0.02480 (13) 0.0477 (5) C15 0.30923 (11) 0.52569 (10) −0.03612 (14) 0.0545 (6) C16 0.36281 (10) 0.54737 (10) −0.01508 (13) 0.0515 (5) C17 0.40164 (12) 0.57099 (11) −0.07905 (15) 0.0636 (7) H17 0.3959 0.5743 −0.1416 0.076\* C18 0.44735 (11) 0.58736 (10) −0.03328 (15) 0.0581 (6) H18 0.4801 0.6046 −0.0575 0.070\* C19 0.43790 (9) 0.57411 (9) 0.05937 (13) 0.0455 (5) C20 0.47569 (9) 0.58754 (8) 0.12914 (13) 0.0429 (5) N51 0.40921 (10) 0.58630 (15) 0.68340 (15) 0.0812 (8) C52 0.38597 (16) 0.62318 (17) 0.6300 (2) 0.0945 (11) H52 0.3762 0.6589 0.6542 0.113\* C53 0.37466 (15) 0.61389 (13) 0.54115 (18) 0.0816 (9) H53 0.3581 0.6429 0.5060 0.098\* C54 0.38735 (9) 0.56279 (11) 0.50349 (14) 0.0513 (5) C55 0.41194 (13) 0.52451 (15) 0.55825 (18) 0.0792 (8) H55 0.4222 0.4884 0.5360 0.095\* C56 0.42214 (14) 0.53814 (18) 0.6469 (2) 0.0885 (10) H56 0.4397 0.5105 0.6834 0.106\* N101 0.06117 (6) 0.38923 (7) 0.30189 (10) 0.0355 (3) C102 0.06574 (8) 0.44515 (8) 0.29549 (13) 0.0403 (4) H102 0.0327 0.4673 0.3038 0.048\* C103 0.11567 (8) 0.47254 (8) 0.27747 (13) 0.0406 (4) H103 0.1167 0.5126 0.2740 0.049\* C104 0.16441 (8) 0.44146 (8) 0.26451 (11) 0.0339 (4) C105 0.16034 (9) 0.38345 (9) 0.27246 (17) 0.0506 (5) H105 0.1928 0.3604 0.2652 0.061\* C106 0.10845 (9) 0.35925 (9) 0.29112 (16) 0.0490 (5) H106 0.1064 0.3194 0.2965 0.059\* N151 0.2660 (2) 0.5329 (2) −0.3142 (2) 0.1294 (17) C152 0.2458 (3) 0.5685 (2) −0.2569 (3) 0.155 (2) H152 0.2211 0.5972 −0.2780 0.185\* C153 0.2581 (2) 0.56707 (18) −0.1666 (2) 0.1290 (18) H153 0.2421 0.5945 −0.1278 0.155\* C154 0.29319 (12) 0.52625 (14) −0.13291 (16) 0.0729 (8) C155 0.31337 (15) 0.4886 (2) −0.19209 (18) 0.1009 (13) H155 0.3375 0.4591 −0.1726 0.121\* C156 0.29872 (18) 0.4931 (2) −0.2832 (2) 0.1174 (17) H156 0.3133 0.4659 −0.3237 0.141\* N201 0.63736 (10) 0.65387 (10) 0.05085 (15) 0.0695 (6) C202 0.60235 (13) 0.68222 (12) 0.1038 (2) 0.0742 (8) H202 0.6140 0.7184 0.1241 0.089\* C203 0.55016 (12) 0.66270 (10) 0.13114 (18) 0.0635 (7) H203 0.5271 0.6852 0.1689 0.076\* C204 0.53176 (10) 0.61002 (9) 0.10312 (14) 0.0474 (5) C205 0.56798 (10) 0.58046 (10) 0.04725 (15) 0.0548 (6) H205 0.5577 0.5442 0.0255 0.066\* C206 0.61911 (11) 0.60416 (12) 0.02354 (17) 0.0645 (7) H206 0.6429 0.5831 −0.0153 0.077\* ------ --------------- --------------- --------------- -------------------- -- Atomic displacement parameters (Å^2^) {#tablewrapadps} ===================================== ------ -------------- -------------- -------------- --------------- -------------- --------------- *U*^11^ *U*^22^ *U*^33^ *U*^12^ *U*^13^ *U*^23^ Zn1 0.04778 (16) 0.04374 (16) 0.02318 (14) −0.00756 (11) 0.00168 (10) −0.00080 (10) N21 0.0579 (10) 0.0455 (9) 0.0266 (8) −0.0116 (8) 0.0027 (7) 0.0007 (7) N22 0.0485 (9) 0.0476 (9) 0.0251 (8) −0.0020 (7) −0.0005 (7) 0.0025 (7) N23 0.0531 (10) 0.0414 (9) 0.0266 (8) −0.0084 (7) 0.0028 (7) −0.0033 (6) N24 0.0438 (9) 0.0446 (9) 0.0247 (8) −0.0010 (7) 0.0004 (6) −0.0035 (6) C1 0.0521 (12) 0.0395 (10) 0.0342 (10) −0.0129 (9) 0.0034 (8) −0.0050 (8) C2 0.0581 (14) 0.0608 (14) 0.0408 (12) −0.0204 (11) 0.0039 (10) −0.0108 (10) C3 0.0599 (14) 0.0649 (14) 0.0354 (11) −0.0209 (11) −0.0001 (10) −0.0111 (10) C4 0.0522 (12) 0.0473 (11) 0.0290 (10) −0.0094 (9) 0.0010 (8) −0.0072 (8) C5 0.0520 (12) 0.0531 (12) 0.0275 (10) −0.0066 (9) −0.0019 (8) −0.0055 (8) C6 0.0479 (11) 0.0494 (11) 0.0259 (10) 0.0000 (9) 0.0024 (8) −0.0038 (8) C7 0.0494 (12) 0.0603 (13) 0.0256 (10) −0.0036 (10) 0.0010 (8) −0.0026 (9) C8 0.0458 (11) 0.0528 (12) 0.0279 (10) 0.0012 (9) 0.0035 (8) −0.0002 (8) C9 0.0410 (10) 0.0403 (10) 0.0274 (9) 0.0045 (8) 0.0019 (7) −0.0024 (7) C10 0.0421 (10) 0.0373 (10) 0.0286 (9) 0.0047 (8) 0.0014 (7) −0.0020 (7) C11 0.0428 (10) 0.0414 (10) 0.0287 (10) 0.0024 (8) −0.0018 (8) −0.0010 (8) C12 0.0457 (11) 0.0623 (13) 0.0300 (10) −0.0037 (10) −0.0036 (8) −0.0017 (9) C13 0.0565 (13) 0.0725 (15) 0.0289 (11) −0.0093 (11) −0.0068 (9) 0.0036 (10) C14 0.0566 (13) 0.0600 (13) 0.0264 (10) −0.0071 (10) −0.0044 (9) 0.0036 (9) C15 0.0687 (15) 0.0670 (15) 0.0279 (11) −0.0171 (12) −0.0035 (10) 0.0076 (9) C16 0.0697 (15) 0.0574 (13) 0.0273 (10) −0.0158 (11) 0.0024 (9) 0.0052 (9) C17 0.0816 (18) 0.0817 (17) 0.0276 (11) −0.0288 (14) 0.0023 (11) 0.0087 (11) C18 0.0723 (16) 0.0672 (15) 0.0347 (12) −0.0260 (12) 0.0053 (10) 0.0071 (10) C19 0.0595 (13) 0.0455 (11) 0.0314 (10) −0.0118 (9) 0.0050 (9) 0.0012 (8) C20 0.0570 (12) 0.0366 (10) 0.0351 (10) −0.0118 (9) 0.0066 (9) −0.0009 (8) N51 0.0633 (14) 0.145 (3) 0.0359 (12) −0.0305 (15) −0.0006 (10) −0.0144 (14) C52 0.126 (3) 0.113 (3) 0.0447 (17) −0.011 (2) −0.0056 (17) −0.0286 (17) C53 0.126 (3) 0.0787 (19) 0.0403 (14) −0.0017 (18) −0.0102 (15) −0.0181 (13) C54 0.0499 (12) 0.0755 (16) 0.0286 (11) −0.0147 (11) 0.0009 (9) −0.0067 (10) C55 0.096 (2) 0.100 (2) 0.0419 (15) 0.0139 (17) −0.0143 (14) −0.0037 (14) C56 0.082 (2) 0.135 (3) 0.0486 (17) 0.002 (2) −0.0152 (14) 0.0081 (18) N101 0.0379 (8) 0.0442 (9) 0.0244 (8) 0.0040 (7) −0.0013 (6) −0.0035 (6) C102 0.0418 (11) 0.0434 (11) 0.0358 (10) 0.0092 (8) 0.0059 (8) −0.0024 (8) C103 0.0466 (11) 0.0392 (10) 0.0360 (10) 0.0058 (8) 0.0063 (8) 0.0009 (8) C104 0.0386 (10) 0.0412 (10) 0.0218 (8) 0.0052 (8) −0.0005 (7) −0.0033 (7) C105 0.0376 (11) 0.0436 (12) 0.0707 (15) 0.0078 (9) 0.0018 (10) −0.0042 (10) C106 0.0422 (11) 0.0384 (11) 0.0665 (15) 0.0038 (9) −0.0002 (10) −0.0022 (10) N151 0.157 (4) 0.182 (4) 0.0499 (18) −0.092 (3) −0.022 (2) 0.029 (2) C152 0.267 (7) 0.130 (4) 0.067 (3) −0.043 (4) −0.074 (3) 0.028 (3) C153 0.218 (5) 0.105 (3) 0.064 (2) −0.016 (3) −0.068 (3) 0.026 (2) C154 0.0797 (18) 0.108 (2) 0.0312 (13) −0.0441 (16) −0.0058 (12) 0.0136 (13) C155 0.088 (2) 0.180 (4) 0.0349 (15) −0.029 (2) 0.0018 (13) −0.0193 (18) C156 0.098 (3) 0.210 (5) 0.0446 (19) −0.059 (3) 0.0097 (17) −0.011 (2) N201 0.0714 (14) 0.0754 (15) 0.0616 (13) −0.0299 (12) 0.0153 (11) −0.0022 (11) C202 0.0847 (19) 0.0592 (15) 0.0786 (19) −0.0325 (14) 0.0173 (16) −0.0085 (14) C203 0.0751 (17) 0.0495 (13) 0.0658 (16) −0.0190 (12) 0.0165 (13) −0.0104 (11) C204 0.0627 (13) 0.0454 (11) 0.0340 (11) −0.0135 (10) 0.0057 (9) 0.0004 (8) C205 0.0693 (15) 0.0540 (13) 0.0410 (12) −0.0153 (11) 0.0148 (10) −0.0076 (10) C206 0.0700 (16) 0.0752 (17) 0.0482 (14) −0.0173 (13) 0.0177 (12) −0.0064 (12) ------ -------------- -------------- -------------- --------------- -------------- --------------- Geometric parameters (Å, °) {#tablewrapgeomlong} =========================== --------------------- ------------- ----------------------- ------------- Zn1---N24 2.0684 (15) C19---C20 1.413 (3) Zn1---N21 2.0695 (16) C20---C204 1.483 (3) Zn1---N22 2.0695 (17) N51---C56 1.302 (5) Zn1---N23 2.0747 (16) N51---C52 1.306 (5) Zn1---N101^i^ 2.1385 (16) C52---C53 1.376 (4) N21---C19 1.363 (3) C52---H52 0.9500 N21---C16 1.369 (3) C53---C54 1.369 (4) N22---C11 1.364 (3) C53---H53 0.9500 N22---C14 1.373 (2) C54---C55 1.355 (4) N23---C4 1.369 (3) C55---C56 1.389 (4) N23---C1 1.374 (3) C55---H55 0.9500 N24---C9 1.363 (3) C56---H56 0.9500 N24---C6 1.370 (2) N101---C102 1.333 (3) C1---C20 1.411 (3) N101---C106 1.336 (3) C1---C2 1.443 (3) N101---Zn1^ii^ 2.1385 (16) C2---C3 1.347 (3) C102---C103 1.376 (3) C2---H2 0.9500 C102---H102 0.9500 C3---C4 1.442 (3) C103---C104 1.383 (3) C3---H3 0.9500 C103---H103 0.9500 C4---C5 1.407 (3) C104---C105 1.383 (3) C5---C6 1.404 (3) C105---C106 1.385 (3) C5---C54 1.497 (3) C105---H105 0.9500 C6---C7 1.442 (3) C106---H106 0.9500 C7---C8 1.352 (3) N151---C152 1.294 (7) C7---H7 0.9500 N151---C156 1.307 (6) C8---C9 1.443 (3) C152---C153 1.385 (5) C8---H8 0.9500 C152---H152 0.9500 C9---C10 1.408 (3) C153---C154 1.372 (5) C10---C11 1.407 (3) C153---H153 0.9500 C10---C104 1.489 (3) C154---C155 1.345 (5) C11---C12 1.449 (3) C155---C156 1.413 (5) C12---C13 1.345 (3) C155---H155 0.9500 C12---H12 0.9500 C156---H156 0.9500 C13---C14 1.432 (3) N201---C206 1.320 (3) C13---H13 0.9500 N201---C202 1.330 (4) C14---C15 1.401 (3) C202---C203 1.382 (4) C15---C16 1.405 (3) C202---H202 0.9500 C15---C154 1.500 (3) C203---C204 1.387 (3) C16---C17 1.442 (3) C203---H203 0.9500 C17---C18 1.339 (3) C204---C205 1.388 (3) C17---H17 0.9500 C205---C206 1.382 (3) C18---C19 1.441 (3) C205---H205 0.9500 C18---H18 0.9500 C206---H206 0.9500 N24---Zn1---N21 162.77 (7) C17---C18---H18 126.1 N24---Zn1---N22 88.42 (6) C19---C18---H18 126.1 N21---Zn1---N22 88.84 (7) N21---C19---C20 126.29 (18) N24---Zn1---N23 89.34 (6) N21---C19---C18 109.15 (18) N21---Zn1---N23 87.94 (6) C20---C19---C18 124.46 (19) N22---Zn1---N23 161.70 (7) C1---C20---C19 124.66 (19) N24---Zn1---N101^i^ 95.10 (6) C1---C20---C204 118.30 (18) N21---Zn1---N101^i^ 102.11 (6) C19---C20---C204 116.97 (18) N22---Zn1---N101^i^ 102.00 (6) C56---N51---C52 115.3 (3) N23---Zn1---N101^i^ 96.29 (6) N51---C52---C53 124.6 (3) C19---N21---C16 106.82 (16) N51---C52---H52 117.7 C19---N21---Zn1 126.42 (13) C53---C52---H52 117.7 C16---N21---Zn1 126.72 (14) C54---C53---C52 119.8 (3) C11---N22---C14 106.27 (17) C54---C53---H53 120.1 C11---N22---Zn1 126.10 (13) C52---C53---H53 120.1 C14---N22---Zn1 127.37 (14) C55---C54---C53 115.9 (2) C4---N23---C1 106.45 (16) C55---C54---C5 123.2 (2) C4---N23---Zn1 125.15 (13) C53---C54---C5 120.9 (2) C1---N23---Zn1 126.61 (13) C54---C55---C56 119.9 (3) C9---N24---C6 106.48 (15) C54---C55---H55 120.0 C9---N24---Zn1 126.17 (12) C56---C55---H55 120.0 C6---N24---Zn1 126.59 (13) N51---C56---C55 124.4 (3) N23---C1---C20 124.64 (18) N51---C56---H56 117.8 N23---C1---C2 109.72 (17) C55---C56---H56 117.8 C20---C1---C2 125.64 (19) C102---N101---C106 116.85 (17) C3---C2---C1 106.8 (2) C102---N101---Zn1^ii^ 120.26 (13) C3---C2---H2 126.6 C106---N101---Zn1^ii^ 122.55 (14) C1---C2---H2 126.6 N101---C102---C103 123.56 (18) C2---C3---C4 107.42 (19) N101---C102---H102 118.2 C2---C3---H3 126.3 C103---C102---H102 118.2 C4---C3---H3 126.3 C102---C103---C104 119.61 (18) N23---C4---C5 126.00 (18) C102---C103---H103 120.2 N23---C4---C3 109.56 (18) C104---C103---H103 120.2 C5---C4---C3 124.42 (18) C105---C104---C103 117.31 (18) C6---C5---C4 125.98 (18) C105---C104---C10 122.03 (17) C6---C5---C54 117.43 (18) C103---C104---C10 120.65 (17) C4---C5---C54 116.59 (18) C104---C105---C106 119.38 (19) N24---C6---C5 125.05 (18) C104---C105---H105 120.3 N24---C6---C7 109.78 (17) C106---C105---H105 120.3 C5---C6---C7 125.15 (18) N101---C106---C105 123.3 (2) C8---C7---C6 106.91 (17) N101---C106---H106 118.4 C8---C7---H7 126.5 C105---C106---H106 118.4 C6---C7---H7 126.5 C152---N151---C156 117.0 (4) C7---C8---C9 106.83 (18) N151---C152---C153 123.7 (5) C7---C8---H8 126.6 N151---C152---H152 118.2 C9---C8---H8 126.6 C153---C152---H152 118.2 N24---C9---C10 125.42 (17) C154---C153---C152 120.3 (5) N24---C9---C8 110.00 (16) C154---C153---H153 119.9 C10---C9---C8 124.54 (18) C152---C153---H153 119.9 C11---C10---C9 125.06 (18) C155---C154---C153 116.1 (3) C11---C10---C104 117.12 (16) C155---C154---C15 122.8 (3) C9---C10---C104 117.77 (16) C153---C154---C15 121.1 (3) N22---C11---C10 125.68 (17) C154---C155---C156 120.0 (4) N22---C11---C12 109.87 (17) C154---C155---H155 120.0 C10---C11---C12 124.42 (18) C156---C155---H155 120.0 C13---C12---C11 106.50 (19) N151---C156---C155 122.9 (5) C13---C12---H12 126.7 N151---C156---H156 118.6 C11---C12---H12 126.7 C155---C156---H156 118.6 C12---C13---C14 107.59 (19) C206---N201---C202 115.6 (2) C12---C13---H13 126.2 N201---C202---C203 124.5 (2) C14---C13---H13 126.2 N201---C202---H202 117.8 N22---C14---C15 124.8 (2) C203---C202---H202 117.8 N22---C14---C13 109.76 (18) C202---C203---C204 119.5 (2) C15---C14---C13 125.41 (19) C202---C203---H203 120.3 C14---C15---C16 126.03 (19) C204---C203---H203 120.3 C14---C15---C154 117.7 (2) C203---C204---C205 116.2 (2) C16---C15---C154 116.28 (19) C203---C204---C20 121.7 (2) N21---C16---C15 125.76 (19) C205---C204---C20 122.07 (19) N21---C16---C17 109.5 (2) C206---C205---C204 119.5 (2) C15---C16---C17 124.7 (2) C206---C205---H205 120.3 C18---C17---C16 106.7 (2) C204---C205---H205 120.3 C18---C17---H17 126.6 N201---C206---C205 124.8 (2) C16---C17---H17 126.6 N201---C206---H206 117.6 C17---C18---C19 107.8 (2) C205---C206---H206 117.6 --------------------- ------------- ----------------------- ------------- Symmetry codes: (i) *y*, −*x*+1/2, −*z*+1/2; (ii) −*y*+1/2, *x*, −*z*+1/2. Hydrogen-bond geometry (Å, °) {#tablewraphbondslong} ============================= --------------------- --------- --------- ----------- --------------- *D*---H···*A* *D*---H H···*A* *D*···*A* *D*---H···*A* C7---H7···N151^iii^ 0.95 2.65 3.583 (4) 167\. C17---H17···N51^iv^ 0.95 2.66 3.583 (3) 165\. --------------------- --------- --------- ----------- --------------- Symmetry codes: (iii) *x*, *y*, *z*+1; (iv) *x*, *y*, *z*−1. ###### Selected geometric parameters (Å, °) --------------- ------------- Zn1---N24 2.0684 (15) Zn1---N21 2.0695 (16) Zn1---N22 2.0695 (17) Zn1---N23 2.0747 (16) Zn1---N101^i^ 2.1385 (16) --------------- ------------- --------------------- ------------ N24---Zn1---N21 162.77 (7) N24---Zn1---N22 88.42 (6) N21---Zn1---N22 88.84 (7) N24---Zn1---N23 89.34 (6) N21---Zn1---N23 87.94 (6) N22---Zn1---N23 161.70 (7) N24---Zn1---N101^i^ 95.10 (6) N21---Zn1---N101^i^ 102.11 (6) N22---Zn1---N101^i^ 102.00 (6) N23---Zn1---N101^i^ 96.29 (6) --------------------- ------------ Symmetry code: (i) . ###### Hydrogen-bond geometry (Å, °) *D*---H⋯*A* *D*---H H⋯*A* *D*⋯*A* *D*---H⋯*A* -------------------- --------- ------- ----------- ------------- C7---H7⋯N151^ii^ 0.95 2.65 3.583 (4) 167 C17---H17⋯N51^iii^ 0.95 2.66 3.583 (3) 165 Symmetry codes: (ii) ; (iii) .

Is There Really a Male Bias? {#s1} ============================ The diagnosis of classic autism and Asperger Syndrome (AS), known as Autism Spectrum Conditions (ASC), rests on difficulties in reciprocal social interaction and communication, alongside strongly repetitive behavior and unusually narrow interests [@pbio.1001081-APA1]. The prevalence of ASC is estimated to be 1% [@pbio.1001081-Baird1],[@pbio.1001081-BaronCohen1]. A diagnosis of classic autism, unlike AS, also requires the presence of additional learning difficulties and language delay. ASC is neurobiological, evidenced by atypical brain development in structure and function [@pbio.1001081-Bauman1]. ASC is also genetic [@pbio.1001081-Stodgell1],[@pbio.1001081-Geschwind1] though not without some interaction with environmental influences. ASC is strongly biased towards males [@pbio.1001081-Fombonne1], with ratios of 4∶1 (male∶female) for classic autism [@pbio.1001081-Chakrabarti1] and as high as 11∶1 in individuals with AS [@pbio.1001081-Gillberg1]. The specific factors responsible for the higher male prevalence in ASC remain unclear. ASC is not the only neurodevelopmental condition more common among males---a greater prevalence in males versus females is also seen in Attention Deficit and Hyperactivity Disorder (ADHD), dyslexia, conduct disorder (CD), specific language impairment, Tourette Syndrome, and Learning Difficulties (see [Table 1](#pbio-1001081-t001){ref-type="table"}) [@pbio.1001081-Rutter1]. 10.1371/journal.pbio.1001081.t001 ###### Male biased sex ratios in other neurodevelopmental conditions. {#pbio-1001081-t001-1} ------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- **1** **Attention Deficit Hyperactivity Disorder (ADHD).** The ratio of males to females with ADHD is high in clinic samples (up to 10∶1) [@pbio.1001081-Arnold2],[@pbio.1001081-Biederman1]. However, it drops to 2∶1 to 4∶1 in community samples [@pbio.1001081-Bauermeister1],[@pbio.1001081-Costello1],[@pbio.1001081-Merikangas1] and the majority of studies in adults show no significant effect of sex on prevalence [@pbio.1001081-Simon1]. This suggests that the biased sex ratios observed in ADHD may result from referral bias rather than a biological mechanism. **2** **Conduct Disorder (CD).** Males are two to four times more likely to develop CD than females [@pbio.1001081-Loeber1], though no sex difference was observed in the recent NHANES study [@pbio.1001081-Merikangas1]. This discrepancy probably reflects the observation that while sex differences are not pronounced in adolescent-limited antisocial behavior, the male∶female ratio for early-onset, life-course-persistent antisocial behavior is 10∶1 or greater [@pbio.1001081-Moffitt1]. **3** **Dyslexia/Reading Disability (RD).** Early research suggested that there was a significant excess of males with RD, but this view has been challenged as reflecting referral bias and subjective methods of assessment [@pbio.1001081-Shaywitz1]. It is clear that ascertainment bias does inflate the true prevalence of RD in males, but a review of existing studies suggests that there is a slightly skewed gender ratio, between 1.7 and 2.00 [@pbio.1001081-Liederman1]. **4** **Specific Language Impairment (SLI).** While many early studies reported a male biased sex ratio of between 2∶1 and 3∶1 [@pbio.1001081-Bishop1] for SLI, it has been suggested that this reflects ascertainment bias [@pbio.1001081-Shaywitz1]. Epidemiological studies have identified equivalent numbers of males and females meeting diagnostic criteria [@pbio.1001081-Tomblin1] or increased prevalence in females [@pbio.1001081-Law1]. **5** **Tourette Syndrome (TS).** TS shows a male to female ratio of between 4∶1 and 6∶1 [@pbio.1001081-Kadesjo1]. It is notable that 50%--90% have comorbid ADHD, particularly in clinic populations, which may contribute to the biased ratio. ------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- However, the male bias is much more pronounced in ASC, especially in the case of AS. This male bias could simply reflect the difficulty of diagnosing AS in females. Though classic autism would not be missed in females, AS could be if it presented as some other condition, such as anorexia [@pbio.1001081-Treasure1] or borderline personality disorder [@pbio.1001081-New1], both of which involve the exercise of excessive control over the environment or other people, and a certain degree of a self-centeredness. Equally, AS in females could be under-diagnosed if females are more motivated to learn to conform socially or have better imitation skills that allow them to "pretend to be normal" [@pbio.1001081-HollidayWilley1]. Finally, this male bias might reflect the inability of the widely used diagnostic instruments (the Autism Diagnostic Observation Schedule (ADOS) or Autism Diagnostic Interview-Revised (ADI-R)) to detect the more subtle ways in which AS may present in females. While these explanations of mis- or under-diagnosis may explain part of the male bias, there may also be *biological* reasons for the male bias in ASC. We argue that the bias can be understood as an extreme expression of the psychological and physiological attributes of the male brain; that is, males need only slight psychological and physiological changes to exhibit ASC while females would require more, thus making ASC rarer in females. What factors might favor overdevelopment of male characteristics? One possible biological mechanism could be the masculinizing effect of fetal testosterone (fT). Two other possibilities include the X- and Y-linked theories and the reduced autosomal penetrance theory (which posits that females harbor fewer ASC-related mutations on autosomal chromosomes). Future research will help to resolve the validity or flaws of these theories, which for now remain neither fully confirmed nor refuted. Here, we lay out some of the evidence for these theories in explaining the male bias in ASC. Is ACS an Extreme Expression of the Male Brain? {#s2} =============================================== The Extreme Male Brain (EMB) theory of autism extends the Empathizing-Systemizing (E-S) theory of typical sex differences [@pbio.1001081-BaronCohen2], which proposes that females on average have a stronger drive to *empathize* (to identify another person\'s thoughts and feelings and to respond to these with an appropriate emotion), while males on average have a stronger drive to *systemize* (to analyze or construct rule-based systems). Whilst sociologists still debate if there are any sex differences at all, and if so whether these are purely the result of cultural conditioning, biologists have long known from animal research that sex differences in behavior exist in primates and are influenced by biology as well as the environment. On the Empathy Quotient (EQ) [@pbio.1001081-BaronCohen3] typical females score higher than typical males who score higher than those with ASC [@pbio.1001081-BaronCohen3]. On the Systemizing Quotient (SQ), individuals with ASC score higher than typical males who score higher than typical females [@pbio.1001081-BaronCohen4]--[@pbio.1001081-Auyeung1]. Additional psychological evidence (summarized in [Table 2](#pbio-1001081-t002){ref-type="table"} and in [Text S1](#pbio.1001081.s001){ref-type="supplementary-material"}) shows that---irrespective of the direction of sex difference---people with autism show an extreme of the male profile. Note that the EMB theory does not state that all psychological sex differences will be exaggerated in ASC---only those relating to empathy and systemizing. 10.1371/journal.pbio.1001081.t002 ###### A summary of the psychological evidence for the Extreme Male Brain (EMB) theory (see [Text S1](#pbio.1001081.s001){ref-type="supplementary-material"} for a fuller discussion). {#pbio-1001081-t002-2} Psychological Measure Autism\>Male\>Female Female\>Male\>Autism Key References -------------------------------------------------------- ---------------------- ---------------------- ------------------------------------------------------------------------------------- Adolescent AQ ✓ [@pbio.1001081-BaronCohen11] Adult Autism Spectrum Quotient (AQ) ✓ [@pbio.1001081-BaronCohen9],[@pbio.1001081-Wakabayashi1]--[@pbio.1001081-Hoekstra1] Adult Systemizing Quotient (SQ) ✓ [@pbio.1001081-BaronCohen4] Child AQ ✓ [@pbio.1001081-Auyeung5] Child SQ ✓ [@pbio.1001081-Auyeung6] Childhood Autism Spectrum Test (CAST) ✓ [@pbio.1001081-Scott1]--[@pbio.1001081-Williams2] Embedded Figures Test ✓ [@pbio.1001081-Shah1],[@pbio.1001081-Jolliffe1] Intuitive Physics Test ✓ [@pbio.1001081-Lawson1],[@pbio.1001081-BaronCohen12] Social Responsiveness Scale ✓ [@pbio.1001081-Constantino1],[@pbio.1001081-Constantino2] Quantitative Checklist for Autism in Toddlers (Q-CHAT) ✓ [@pbio.1001081-Allison1] Adult Empathy Quotient (EQ) ✓ [@pbio.1001081-BaronCohen3] Child EQ ✓ [@pbio.1001081-Auyeung6] Faux Pas Test ✓ [@pbio.1001081-BaronCohen13] Friendship and Relationship Questionnaire (FQ) ✓ [@pbio.1001081-BaronCohen14] Reading the Mind in the Eyes ✓ [@pbio.1001081-BaronCohen15] Social Stories Questionnaire (SSQ) ✓ [@pbio.1001081-Lawson1] Sexual Dimorphism in the Human Brain {#s2a} ------------------------------------ Additional support for the EMB theory of ASC comes from evidence of neural sexual dimorphism across development. Some key examples of typical sexual dimorphism reveal an extreme of the typical male profile in the neurodevelopment of ASC [@pbio.1001081-BaronCohen5]. However, one caveat to keep in mind is that just as all *psychological* sex differences do not constitute an exaggerated form of maleness in ASC, neither do all *neural* differences. Indeed, given that the EMB theory is defined at the psychological level, we should expect only a narrow set of neural sex differences will be involved in such hyper-masculinization in ASC. A key finding supporting this prediction is that infant males on average have a larger brain than females [@pbio.1001081-Gilmore1] and children with autism have even larger brains early in life right around the time they would typically receive a diagnosis (2--4 years) [@pbio.1001081-Courchesne1]. In addition, independent of global differences in brain size, the amygdala in typical males tends to be larger than in females [@pbio.1001081-Good1], and early in development the amygdala in autism is even more enlarged than that observed in typical males [@pbio.1001081-Schumann1]--[@pbio.1001081-Mosconi1]. In addition to such *structural* sexual dimorphism in the brain, exaggeration of neural sexual dimorphism extends to brain *function* and corroborates predictions from the EMB theory (see [Table 3](#pbio-1001081-t003){ref-type="table"} and [Text S1](#pbio.1001081.s001){ref-type="supplementary-material"} for fuller discussion) [@pbio.1001081-BaronCohen6]--[@pbio.1001081-Manjaly1]. 10.1371/journal.pbio.1001081.t003 ###### A summary of the evidence consistent with the EMB theory at the neural level (see [Text S1](#pbio.1001081.s001){ref-type="supplementary-material"} for a fuller discussion). {#pbio-1001081-t003-3} Brain Region Autism\>Male\>Female Female\>Male\>Autism Key References --------------------------------------------------------------- ---------------------- ---------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- **Structure** Total brain volume ✓ [@pbio.1001081-Gilmore1],[@pbio.1001081-Redcay1]--[@pbio.1001081-Courchesne2] Amgydala ✓ [@pbio.1001081-Good1]--[@pbio.1001081-Mosconi1],[@pbio.1001081-Cheng1]--[@pbio.1001081-Chen1]. Corpus callosum ✓ [@pbio.1001081-Lenroot1],[@pbio.1001081-Frazier1] Perisylvian language areas (Heschl\'s gyrus/planum temporale) ✓ [@pbio.1001081-Good1],[@pbio.1001081-Rojas1]--[@pbio.1001081-Sowell1] L\>R asymmetry in planum temporale ✓ [@pbio.1001081-Good1],[@pbio.1001081-Rojas2],[@pbio.1001081-Witelson2]--[@pbio.1001081-Gage1] Lateral fronto-parietal cortex ✓ [@pbio.1001081-Cheng1],[@pbio.1001081-Yamasue1],[@pbio.1001081-Goldstein1],[@pbio.1001081-Chen1],[@pbio.1001081-Sowell1],[@pbio.1001081-Im1]--[@pbio.1001081-McAlonan1] **Function** Default Mode Network Connectivity ✓ [@pbio.1001081-Biswal1],[@pbio.1001081-Kennedy1] Embedded Figures fMRI ✓ [@pbio.1001081-Ring1]--[@pbio.1001081-Manjaly1],[@pbio.1001081-Lee1] Reading the Mind in the Eyes task fMRI ✓ [@pbio.1001081-BaronCohen6],[@pbio.1001081-BaronCohen7] The set of striking findings of hyper-masculinization in ASC at three simultaneous levels (cognitive, neuroanatomy, and neural function) raises the question as to which biological mechanism(s) are involved. Two plausible mechanisms that could give rise to sexual dimorphism, hyper-masculinization, and/or the absence of typical sexual dimorphism at the levels of brain, cognition, and behavior are the "organizing" effects of fetal testosterone (fT) [@pbio.1001081-Geschwind2]--[@pbio.1001081-Arnold1] and X- or Y-linked genetic factors. We review these three interesting hypotheses, since these may also have relevance to the sex ratio in ASC. These are not proposed as complete explanations for ASC, since ASC is recognized to be multi-factorial, but they may form an important part of the explanation. What Might Cause an Extreme Male Brain? {#s3} ======================================= The Fetal Testosterone (fT) Theory {#s3a} ---------------------------------- ### Fetal androgens affect the brain: Evidence from animal and human studies {#s3a1} Animal studies, especially in rodents, confirm that early exposure to androgens (such as testosterone) acts on the brain to produce sex differences in behavior, cognition, brain structure, and function (see [Text S1](#pbio.1001081.s001){ref-type="supplementary-material"} for more discussion of work with animals) [@pbio.1001081-Phoenix1]--[@pbio.1001081-DeVries1]. It is widely accepted that fT exposure also affects brain development and behavior in humans. Human males experience a surge in fT between weeks 8 to 24 of gestation [@pbio.1001081-BaronCohen8]--[@pbio.1001081-Hines1], reaching almost pubertal levels. There is also a second surge soon after birth (here called "neonatal testosterone," or nT). Usually the levels remain high and then drop to barely detectable levels by 4--6 months [@pbio.1001081-Smail1], until the third surge at puberty. Whilst the third surge is understood to be controlling the onset of puberty, the function of first surge (fT) is believed to play a major role in brain masculinization. While direct manipulation of hormones as has been conducted in animal studies is unquestionably unethical in human fetuses and infants, alternative research strategies include relating individual variation in amniotic fT exposure to later development [@pbio.1001081-vandeBeek1], or studying people in whom---for medical reasons---the sex hormones are higher or lower than expected for a person\'s sex [@pbio.1001081-Money1], and using proxy measures of fT exposure. Here we review evidence from studies of cognitive traits relevant to ASC and their relationship with amniotic fT. (Evidence from disorders of sexual differentiation and from proxy measures of fT exposure is presented in the [Text S1](#pbio.1001081.s001){ref-type="supplementary-material"}.) ### Fetal androgens affect ASC traits: evidence from amniotic fluid testosterone {#s3a2} fT can be measured in amniotic fluid, obtained during routine amniocentesis. Because amniocentesis is typically performed during the second trimester of pregnancy (usually 14--20 weeks of gestation), when serum testosterone peaks in male fetuses, it offers a unique opportunity to compare fT with ASC traits. There is a well-documented large sex difference in amniotic androgen levels [@pbio.1001081-Dawood1]--[@pbio.1001081-Robinson1]. The origin of androgens in amniotic fluid appears to be the fetus itself, and testosterone obtained in amniotic fluid is thought to be a good reflection of the levels in the fetus [@pbio.1001081-vandeBeek1]. In the Cambridge Fetal Testosterone Project, initiated by our group in 1998, children whose mothers had amniocentesis during pregnancy (but who were otherwise developing normally) have been followed up after birth every year or two and are now approximately 11 years of age [@pbio.1001081-BaronCohen8]. Evidence that amniotic fT affects individual differences in cognitive development in typically developing children (but with clear relevance to ASC) includes the following: fT is *inversely* associated with frequency of eye contact at 12 months old [@pbio.1001081-Lutchmaya1] and with size of vocabulary development at 18 and 24 months [@pbio.1001081-Lutchmaya2]. fT is also *inversely* associated with quality of social relationships at 48 months [@pbio.1001081-Knickmeyer1] and with empathy at 48 and 96 months [@pbio.1001081-Chapman1],[@pbio.1001081-Knickmeyer2]. In contrast, amniotic fT is *positively* associated with narrow interests at 48 months [@pbio.1001081-Knickmeyer1], with "systemizing" at 96 months [@pbio.1001081-Auyeung1], and with performance on the Embedded Figures Test (EFT) as a measure of attention to detail at 96 months [@pbio.1001081-Auyeung2]. These are all behaviors that show sexual dimorphism, but critically, these fT effects are often found within one sex as well as when analyzing the sexes combined. The finding of a consistent inverse correlation between fT and social domains, and a consistent positive correlation between fT and non-social domains, across development, is striking and suggests these are real effects which substantiate the notion that fT plays an "organizational" role in development. In the first study to directly assess if fT affects not just human cognition but also human brain structure, we found that increasing levels of fT are associated with increasing rightward asymmetry in the thickness of one subsection of the corpus callosum, the isthmus [@pbio.1001081-Chura1]. This is interesting since the isthmus projects to posterior parietal and superior temporal cortices, which are integral for language and visuospatial ability and are known to be sexually dimorphic in lateralization, structure, and function (see [Text S1](#pbio.1001081.s001){ref-type="supplementary-material"}). All of the above behavioral domains (eye contact, language development, quality of social relationships, narrow interests, empathy, systemizing, and embedded figures/attention to detail) and brain structure show sexual dimorphism and appear hyper-masculinized in ASC, raising the possibility that fT may play a role in the development of ASC itself. Three recent experiments have confirmed a positive correlation between fT levels and the number of autistic traits a child shows in toddlerhood [@pbio.1001081-Auyeung3] and in later childhood [@pbio.1001081-Auyeung4]. The Cambridge Fetal Testosterone Project has too few children (currently *n* = 635 are enrolled) to test whether fT is elevated in those who later are diagnosed with ASC, but testing for a *direct* association between fT levels and diagnosed ASC will be possible in our ongoing collaboration with the Danish Biobank, which has tens of thousands of amniotic samples, with adequate power to test this hypothesis. Using a different line of evidence, a number of studies have found also *current* androgen dysregulation in ASC or in their relatives, or androgen-related genes being associated with ASC (see [Table 4](#pbio-1001081-t004){ref-type="table"} for a summary of the evidence for the fT/androgen theory). 10.1371/journal.pbio.1001081.t004 ###### Evidence for the effect of sex steroids in autism (see [Text S1](#pbio.1001081.s001){ref-type="supplementary-material"} for a fuller discussion). {#pbio-1001081-t004-4} Evidence Key References ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ----------------------------------------------------------- **From typically developing children** Eye contact is inversely related to fT [@pbio.1001081-Lutchmaya1] Quality of social relationships are inversely related to fT [@pbio.1001081-Knickmeyer1] Vocabulary size is inversely related to fT [@pbio.1001081-Lutchmaya2] Empathy is inversely related to fT [@pbio.1001081-Chapman1],[@pbio.1001081-Knickmeyer2] Autistic traits are positively associated with fT [@pbio.1001081-Auyeung3],[@pbio.1001081-Auyeung4] Restricted interests are positively associated with fT [@pbio.1001081-Knickmeyer1] Systemizing is positively associated with fT [@pbio.1001081-Auyeung1] Rightward asymmetry in the isthmus of the corpus callosum is positively associated with fT [@pbio.1001081-Chura1] **From people with ASC** 10 genes involved in sex steroid synthesis, transport, and/or metabolism associated with AS or AQ or empathy: *HSD11B1*, *LHCGR*, *CYP17A1*, *CYP19A1*, *SCP2*, *CYP11B1*, *ESR1*, *ESR2*, *HSD17B4*, *HSD17B2* [@pbio.1001081-Chakrabarti2] Timing of puberty: Boys with ASC enter puberty earlier. Girls with ASC enter puberty later [@pbio.1001081-Tordjman1]--[@pbio.1001081-Ingudomnukul1] Testosterone related medical conditions in women with ASC and their mothers (e.g., PCOS, breast and ovarian cancers, acne) [@pbio.1001081-Ingudomnukul1] Testosterone related characteristics in women with ASC and their mothers [@pbio.1001081-Ingudomnukul1],[@pbio.1001081-Knickmeyer4] Lower 2D∶4D ratio in ASC, and parents [@pbio.1001081-Manning1]--[@pbio.1001081-deBruin1] SRD5A1 and AR genes associated with ASC [@pbio.1001081-Henningsson1],[@pbio.1001081-Hu1] Decreased expression of RORA gene and aromatase in post-mortem frontal and cerebellar tissue [@pbio.1001081-Sarachana1],[@pbio.1001081-Nguyen2] Females with Congenital Adrenal Hyperplasia (CAH) have elevated AQ [@pbio.1001081-Knickmeyer5] Testosterone levels are elevated in ASC [@pbio.1001081-Schmidtova1] Androstenedione levels are elevated in ASC [@pbio.1001081-Ruta1] Although some studies have failed to support a role for testosterone in ASC (and most of these have not been able to study fT specifically), the studies reported above suggest that fT is implicated in the biased sex ratio seen in ASC. However, alternative models exist which could also explain the excess of males with ASC. In the final part of this article we review the main contender, the X chromosome theory. For completeness, we also briefly review the Y chromosome theory and the reduced autosomal penetrance theory. The X Chromosome Theory {#s3b} ----------------------- The X chromosome contains more genes expressed in the brain than the other chromosomes [@pbio.1001081-Nguyen1]. In addition, more than 10% of people with learning difficulties show an X-linked pattern of inheritance [@pbio.1001081-Laumonnier1], involving mutations in over 90 different X-linked genes [@pbio.1001081-Gecz1],[@pbio.1001081-Ropers1]. Individuals with X-linked learning difficulties may also have ASC, the best-known example being Fragile X Syndrome, where 46% of males and 16% of females carrying the full mutation also have ASC [@pbio.1001081-Bailey1]. On the face of it, the biased sex ratio in ASC would therefore be parsimoniously explained by an X chromosome theory. A problem for this theory is that the majority of linkage and association studies of ASC have failed to find regions of interest on the X chromosome [@pbio.1001081-Consortium1]--[@pbio.1001081-Shao1]. A related problem for this theory is that in the three recent genome-wide studies of copy number variation (CNV) in individuals with ASC that identified mutations affecting the X chromosome, this was only true in a very small minority of cases. This suggests X-linked mutations are only occasionally seen in ASC and therefore cannot account for the large majority of cases. A final problem for the X-linked theory is that other large CNV scans have reported no significant findings on the X chromosome [@pbio.1001081-Szatmari1],[@pbio.1001081-Morrow1]--[@pbio.1001081-Weiss2]. While epigenetic effects on X chromosome genes could affect risk for autism, this hypothesis has not yet been empirically tested. In summary, at present it appears that there are X-linked causes of ASC, but these represent a far smaller percentage of cases than is seen in learning difficulties. Girls with Turner Syndrome (TS) (characterized by the XO karyotype) [@pbio.1001081-Lippe1] are at an increased risk for ASC, which could be the result of an X-linked recessive gene, but this is not clear-cut since XYY and XXYY males are also at increased risk [@pbio.1001081-Tartaglia1]. One study [@pbio.1001081-vanRijn1] has also reported higher autistic traits scores (as measured on the Autism Spectrum Quotient \[AQ\] in XXY males), though this is not always seen [@pbio.1001081-Tartaglia1]. There are other possible versions of the X chromosome theory of ASC. Although females have two X chromosomes, only one of these is generally active. *X chromosome inactivation* (the process by which one X chromosome is suppressed while the other remains active) acts to negate the "dosage" difference in X chromosome genes between males and females. However, 10%--15% of X chromosome genes may continue to be expressed from the supposedly inactive X. Gong and colleagues [@pbio.1001081-Gong1] directly tested this hypothesis and found no evidence for a skewed X chromosome inactivation in a large sample of individuals with and without ASC. X chromosome gene dosage could play a role in sex ratios if the non-silenced genes were protective. However, comparing the incidence of ASC across different sex aneuploidies does not suggest a simple dosage effect, and frequently the ASC occurs in the context of clear learning disabilities, and so could simply be secondary to the latter. It is increasingly recognized that learning difficulties are themselves a risk factor for ASC [@pbio.1001081-Wing1], so any evaluation of the X chromosome theory needs to consider these separately. *Genomic imprinting* (the process by which genetic effects are influenced by whether the genes are transmitted through the father or the mother [@pbio.1001081-Keverne1]) is also of interest. Ordinarily this would not result in sex differences in the rate of a condition, but could do so if the imprinting affects the X chromosome. Skuse [@pbio.1001081-Skuse1],[@pbio.1001081-Skuse2] suggested that an imprinted X-locus could explain sex differences in social and communication skills and the male vulnerability to social and communication impairment. His theory was inspired by the finding that in individuals with TS, the rate of social difficulties varied according to whether their single X chromosome was inherited from the father (X~p~O cases) or the mother (X~m~O cases) (where ~p~ is paternal, and ~m~ is maternal) [@pbio.1001081-Skuse1]. Social problems are greater in X~m~O relative to X~p~O individuals. Typical females always inherit an X chromosome from both parents (X~p~X~m~), but typical males always have only a maternal X (X~m~Y). Skuse hypothesized that a gene expressed on the paternal X acts as a protective factor against the social problems seen in TS and, by extrapolation, as a protective factor against ASC. Creswell and colleagues [@pbio.1001081-Creswell1] subsequently reported five cases of ASC from an unselected sample of 150 subjects with TS. All the cases were X~m~O (or had a structurally abnormal paternal X). All of the cases in that report also had moderate to severe learning difficulties and low verbal IQ scores, despite the fact that intelligence is usually in the average range in TS. This raises the possibility that the kind of ASC observed was related to learning difficulties (i.e., applicable only to classic autism rather than the full autistic spectrum, which includes AS). Also, given that 77% of TS females are X~m~O, while only 23% are X~p~O [@pbio.1001081-Grumbach1], this means that *by chance* one would expect to find ASC more often associated with X~m~O than with X~p~O. No specific X-linked genes have yet been identified which explain these findings, but there is evidence that whichever genes are involved may modulate amygdala circuits which are disrupted in ASC [@pbio.1001081-Skuse3]. Whilst the amygdala has not been directly examined, a study of the whole brain in a mouse model of TS did not identify any paternally expressed X-linked genes, but did identify a maternally expressed gene, *xlr3b*, which was implicated in cognitive flexibility [@pbio.1001081-Davies1]. However, it is unclear if a functioning human orthologue of this gene exists. A recent study searched for imprinted genes in the preoptic area (POA) and medial prefrontal cortex (mPFC) in mouse. No X-linked imprinted genes were identified when using a cut-off of *p*\<0.05, but using a less stringent cut-off of 0.1, a small set of putative X-linked imprinted genes were identified including three paternally expressed genes in the POA and three different paternally expressed genes in the mPFC [@pbio.1001081-Gregg1]. Three of these genes (*cask*, *acsl4*, and *ids*) have human orthologues whose disruption can cause MR. Another intriguing finding from this study was that total levels of expression from X~m~ were increased relative to those of X~p~ in females. This could reflect preferential inactivation of the X~p~ and would act to minimize dosage differences between the sexes. If a screen of females with ASC identified rare mutations or CNVs on the X~p~, this would provide important evidence for the theory. The Y Chromosome Theory {#s3c} ----------------------- Since the XYY and XXYY syndromes have an increased incidence of ASC [@pbio.1001081-Bruining1]--[@pbio.1001081-Tartaglia2], it is important to consider if the male bias in ASC could also result from the male-limited expression of genes on the Y chromosome. This possibility has attracted very little research attention. Such genes should be located in the non-recombining region of the Y. *SRY* (the sex determining gene) is expressed in the medial rostral hypothalamus, as well as the frontal and temporal regions of the human brain [@pbio.1001081-Mayer1]. In vitro assays suggest that *SRY* can increase transcription of tyrosine hydroxylase (the rate-limiting enzyme in dopamine biosynthesis) by binding at a promoter site [@pbio.1001081-Milsted1]. In addition, the knockdown of *SRY* expression in the substantia nigra of the rat decreases tyrosine hydroxylase expression [@pbio.1001081-Dewing1]. This could implicate SRY in the male bias for disorders involving disregulated catecholamines such as ADHD. SRY may also regulate the monoamine oxidase A (*MAO-A*) gene [@pbio.1001081-Wu1]. Other Y-linked genes known to be expressed in human brain include *ZFY* and *PCDH11Y* [@pbio.1001081-Mayer1],[@pbio.1001081-Durand1]. A small candidate gene study failed to find associations between variants in *PCDH11Y* and autism [@pbio.1001081-Durand1], while *ZFY* has not been specifically investigated. One study has reported a missense variant in *NLGN4Y* in a single patient with autism and his father with learning difficulties [@pbio.1001081-Yan1]. Comparison of Y chromosome haplotype groups between cases and controls represents an alternative strategy to identifying Y chromosome effects. Two such studies have been conducted in regard to ASC---one was positive [@pbio.1001081-Serajee1] and one was negative [@pbio.1001081-Jamain1]. Y chromosome effects certainly merit additional research attention, but current evidence is too sparse to evaluate to what extent this mechanism could explain the sex bias in ASC. Reduced Autosomal Penetrance in Females? A Final Theory {#s3d} ------------------------------------------------------- For completeness we briefly mention a final theory, arising from studies of rare CNVs with ASC [@pbio.1001081-Szatmari1],[@pbio.1001081-Sebat1],[@pbio.1001081-Christian1],[@pbio.1001081-Marshall1]. As mentioned earlier, these scans have *not* routinely implicated the X chromosome, but this final model proposes that a significant proportion of ASC cases are the result of dominant de novo mutations (on the autosomes) which have reduced penetrance in females. Statistical analysis of ASC family data has provided supporting evidence [@pbio.1001081-Zhao1]. A problem for this theory, however, is that the majority of studies report that the sex ratio in children with ASC and de novo CNVs is 1∶1. This clearly does not fit with reduced penetrance in females [@pbio.1001081-Beaudet1]. A second problem for this theory is that it does not address *why* penetrance should be reduced in females. However, we agree that it is critical that large-scale linkage and association studies test for sex-specific effects. Not Mutually Exclusive Theories {#s3e} ------------------------------- The X and Y chromosome theories and the fT model offer potential explanations for the biased sex ratio in ASC and warrant further research. While often conceived as competing theories, they need not be mutually exclusive. This is because we cannot rule out the possibility that genes on the X and Y chromosomes may be regulated by fT or have products that affect the production or sensitivity of an individual to fT. X chromosome genes may also regulate Y chromosome genes and vice versa. In addition, it is possible that X or Y chromosome genes and fT exposure are independent risk factors for ASC. The theories do, however, make contrasting predictions for individuals with certain intersex conditions, in particular those with Complete Androgen Insensitivity Syndrome (CAIS), where there is a complete deficiency of working androgen receptors, in the presence of a typical male genetic complement (XY). Given the rarity of this condition, studies using measures of autistic traits (such as the AQ [@pbio.1001081-BaronCohen9]) may be more feasible than studies of diagnosed cases of ASC in CAIS per se. (These contrasting predictions are summarized in [Table 5](#pbio-1001081-t005){ref-type="table"}.) 10.1371/journal.pbio.1001081.t005 ###### Rates of ASC/autistic traits in different medical conditions, as predicted by the X and Y chromosome theories, and the fT theory. {#pbio-1001081-t005-5} Medical Condition Prediction from X-Dosage or X-Linked Recessive Model Prediction from Imprinted X Model Prediction from Y-Chromosome Model Prediction from FT Theory ---------------------------------------------------------- ------------------------------------------------------ ----------------------------------- ------------------------------------ ---------------------------- Complete Androgen Insensitivity Syndrome (CAIS) in males Similar to typical males Similar to typical males Similar to typical males Similar to typical females Congenital Adrenal Hyperplasia (CAH) in females Similar to typical females Similar to typical females Similar to typical females Similar to typical males Turner Syndrome (with a maternal X; X~m~O) Similar to typical males Similar to typical males Similar to typical females Similar to typical females Turner Syndrome (with a paternal X; X~p~O) Similar to typical males Similar to typical females Similar to typical females Similar to typical females Finally, whilst it may be that the psychiatric classification system is "carving nature at its joints," it is also possible that some of the underlying hormonal and genetic mechanisms are involved not just in ASC but are relevant to a broader category of neurodevelopmental conditions (see [Box 1](#pbio-1001081-box001){ref-type="boxed-text"}). **Box 1.** fT and X-linked factors in other neurodevelopmental conditions. **ADHD:** fT has been implicated by several studies using the proxy measure of 2D∶4D (finger) ratio [@pbio.1001081-deBruin1],[@pbio.1001081-McFadden1],[@pbio.1001081-Martel1] and one study of genetic variation at the androgen receptor [@pbio.1001081-Comings1]. An animal model of ADHD suggests that early androgen exposure affects catecholamine innervation of the frontal cortex and cognitive function [@pbio.1001081-King1]. ADHD has also been associated with X-linked genes, in particular monoamine oxidase-B [@pbio.1001081-Jiang1],[@pbio.1001081-Rommelse1] and steroid sulfatase [@pbio.1001081-Brookes1]. The latter has also been implicated in attention deficits in a mouse model of Turner Syndrome [@pbio.1001081-Davies2]. However, genome-wide scans have not implicated the X chromosome in ADHD [@pbio.1001081-Fisher1],[@pbio.1001081-Franke1]. **Conduct Disorder (CD):** Activational effects of gonadal steroids have shown relationships with CD [@pbio.1001081-Pajer1]--[@pbio.1001081-Dorn1], but there is not a simple one-to-one correspondence. In addition, the X-linked gene coding for monoamine oxidase A has been linked to aggression and neural hyperactivity to threat [@pbio.1001081-MeyerLindenberg1]. **Reading Disorder/Dyslexia:** Two studies have failed to find a relation between 2D∶4D (digit) ratio (as a proxy for fT) and dyslexia [@pbio.1001081-Liederman1],[@pbio.1001081-vanGelder1]. One genome-wide linkage analysis suggested a locus on Xq26 [@pbio.1001081-Fisher2]. A nearby susceptibility locus in a single extended family has also been reported [@pbio.1001081-vanGelder1]. **Specific Language Impairment:** The correlation between amniotic fT levels and early vocabulary [@pbio.1001081-Lutchmaya2],[@pbio.1001081-Finegan2] could indicate a role for fT in SLI. Genome-wide linkage studies have not implicated the X chromosome [@pbio.1001081-Bartlett1]--[@pbio.1001081-Newbury1]. **Tourette Syndrome:** Tics in individuals with TS increase in intensity during puberty, suggesting an activational testosterone effect. A role for fT has also been proposed based on a study of gender dysphoria, play preferences, and spatial skills in individuals with TS [@pbio.1001081-Alexander1]. Genome-wide linkage studies have not implicated the X chromosome [@pbio.1001081-State1], but Lawson-Yuen [@pbio.1001081-LawsonYuen1] have reported a pedigree with a NLGN4X deletion which was associated with TS in one family-member. Looking Ahead: Toward a Unified Theory? {#s4} ======================================= For as long as ASC has been recognized, a higher prevalence has been observed in males, yet until 1997, when our group proposed the extreme male brain theory, this potential clue to the etiology of the condition went unexplored [@pbio.1001081-BaronCohen10]. In the early years following the publication of the EMB theory, the majority of the evidence relevant to the theory came from psychological studies, but since 2001 supporting evidence has also come from biology. In the present article we have considered studies that suggest that fetal testosterone is involved in sex differences in key areas of behavior and cognition in the general population (in social development, language development, empathy, systemizing, and attention to detail), as well as in influencing brain structure, and the number of autistic traits an individual possesses. Understanding the relationship between empathy and systemizing will require more research because presenting them as independent ignores the fact that both are related to fT. Nor can we yet extrapolate the fT results to individuals with an ASC diagnosis since this will require much larger collections of amniotic samples than has been possible to date. Strengthening a role for fT in ASC is the recent genetic evidence in which SNPs in key sex steroid genes are associated with either diagnosed AS and/or autistic traits. The main alternatives to the fT theory are the X and Y chromosome theories. Future research could usefully test these theories against each other, or test if all are valid, either independently or because of gene-hormone interactions. Whilst it remains a possibility that the male bias in ASC simply reflects diagnostic difficulties in recognizing ASC in females, the link between ASC and maleness has generated a novel framework for exploring the link between sex and ASC, and a wealth of data relating prenatal hormones to masculinization of the mind and the brain. Supporting Information {#s5} ====================== ###### Supplementary material. (DOC) ###### Click here for additional data file. The authors have declared that no competing interests exist. MRC, Wellcome Trust, Nancy Lurie Marks Foundation, NIHR CLAHRC for Cambridgeshire and Peterborough. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. ADHD : Attention Deficit and Hyperactivity Disorder AQ : Autism Spectrum Quotient AS : Asperger Syndrome ASC : Autism Spectrum Conditions CAIS : Complete Androgen Insensitivity Syndrome CD : conduct disorder CNV : copy number variation EMB : extreme male brain EQ : Empathy Quotient E-S : Empathizing-Systemizing fT : fetal testosterone mPFC : medial prefrontal cortex nT : neonatal testosterone POA : preoptic area SQ : Systemizing Quotient TS : Turner Syndrome [^1]: Unsolved Mysteries discuss a topic of biological importance that is poorly understood and in need of research attention.

INTRODUCTION ============ Surimi, a stabilized myofibrillar protein from fish, is the primary ingredient in fish paste or imitation crab. It is prepared by separation, washing, and mincing to eliminate undesirable blood, lipids, enzymes, and sarcoplasmic proteins ([@b41-ajas-27-1-115-13]; [@b28-ajas-27-1-115-13]). In general, surimi is light in color, bland in odor, low in fat, and extremely functional due to the unique gelling properties of the myofibrillar proteins. These properties make surimi a robust functional ingredient for fabricating new food products ([@b11-ajas-27-1-115-13]; [@b20-ajas-27-1-115-13]). Numerous studies have been conducted on surimi containing fish meat. In addition, application of surimi technology in the production of surimi-based products using the proteins from other animal species could provide a new approach for increasing its utilization and functional properties. Animal meat is of much interest in the development of surimi-based products such as those made from beef, pork and chicken ([@b27-ajas-27-1-115-13]; [@b37-ajas-27-1-115-13]; [@b19-ajas-27-1-115-13]). Pork leg meat, for instance, has been reported to have high myofibrillar protein content allowing increased gel forming capacity in surimi. Thus chicken or pork meat is an attractive substitute for fish meat surimi ([@b17-ajas-27-1-115-13]), with spent laying hens providing a particularly economical source of surimi-based products. Indeed, it was recently shown that spent laying hen meat has potential as filler in surimi-based products ([@b13-ajas-27-1-115-13]). Therefore, the addition of myofibrillar proteins from spent laying hens to surimi-based imitation fish paste (IFP) holds promise in promoting the utilization of such products. It is important for the meat industry to develop new products to satisfy emerging consumer demands for high quality foods. Protein hydrolysates, rich in low molecular weight peptides (di- and tri-peptides, with minimal free amino acids), are a good dietary source due to their high nutritional value and therapeutic properties ([@b6-ajas-27-1-115-13]). In recent years, research has focused on the generation of bioactive peptides from food sources including meat and meat by-products ([@b8-ajas-27-1-115-13]; [@b21-ajas-27-1-115-13]). Peptides have been shown to exert antioxidant, antimicrobial, and antihypertensive effects ([@b31-ajas-27-1-115-13]; [@b14-ajas-27-1-115-13]; [@b22-ajas-27-1-115-13]; [@b9-ajas-27-1-115-13]). Some peptides possess antihypertensive activity through their ability to inhibit Angiotensin I-converting enzyme (ACE) ([@b4-ajas-27-1-115-13]). Also, bioactive peptides can be used as components in functional foods. Protein hydrolysates have been shown to enhance the emulsifying and foaming properties of fish ([@b35-ajas-27-1-115-13]). These peptides need first to be released from the original protein during food processing or digestion in order to exhibit biological activity and little is known about this process in meat and meat products. Mechanically deboned chicken meat (MDCM) has a high content of heme pigments, connective tissue, and fat ([@b44-ajas-27-1-115-13]). It is dark in color, has undesired textural properties, and is susceptible to lipid oxidation. However, MDCM is one of the most common raw materials used to produce processed poultry products. Also, MDCM is widely used in the food industry for upgrading the functional and nutritional properties of proteins, thereby creating value-added products. Thus, MDCM hydrolysis may be used as a potential starting material for the generation of bioactive peptides. The aims of the present investigation were to study the effects of added MDCM hydrolysates on the gel properties, oxidative stability, and angiotensin I-converting enzymes (ACE) inhibitory activities of IFP made from Alaska Pollack and spend laying hen meat. MATERIALS AMD METHODS ===================== Sample preparation ------------------ Frozen Alaska Pollack were purchased from Han-sung Food Co. Ltd. (Pusan, Korea), cut into 500 g blocks while frozen, packed into polyethylene bags, and stored at −20°C until use. Spent laying hens also were obtained at the same time from a commercial slaughterhouse. Spent laying hen myofibrillar protein was collected by the pH adjustment ([@b16-ajas-27-1-115-13]). MDCM hydrolysates were obtained by the protein hydrolysates method. Three batches were collected on different days for experimental replication. IFP samples were divided into three groups: Control (C); composed of commercial IFP containing spent laying hen meat, T1; consisted of IFP sample containing 0.4% MDCM hydrolysate, and T2; consisted of IFP sample containing 0.8% MDCM hydrolysate. The composition of the IFP is presented in [Table 1](#t1-ajas-27-1-115-13){ref-type="table"} and a flow diagram depicting IFP preparation is shown in [Figure 1](#f1-ajas-27-1-115-13){ref-type="fig"}. ### pH method The external fat tissue, bone, and skin were removed from the muscles, and the lean muscle was cut into approximately 3.0×3.0×2.0 cm^3^ cubes and ground through a 3 mm diameter hole using a mincer. The minced samples were combined with six times volume of distilled water and homogenized with a Polytron homogenizer (T25-B, IKA Sdn. Bhd., Malaysia) at 8,000 rpm for 30 s. The pH of the homogenate was then adjusted to pH 11.0 by the addition of 1 mol NaOH and centrifuged at 10,000×*g* for 25 min, after which the top layer of fat and bottom layer of connective tissue were discarded. The middle layer containing the myofibrillar protein was then adjusted to pH 5 by addition of 1 N HCl solution, after which it was centrifuged at 10,000×*g* for 25 min. The resulting sediment was then used for the manufacture of IFP. ### Preparation of the protein hydrolysates In this study, one- and two-stage hydrolysis was employed. The nerves, skin, and visible fat were removed from the meat, which was then fragmented, ground, and homogenized with distilled water (meat:water ratio, 1:3 w/w). The homogenate was heated at 43°C and the pH was adjusted to 7.0 with 2 N NaOH. Five percent of Protarmex) was added and the reaction pH was maintained at a constant pH by the addition of 2 N NaOH. The hydrolytic process was terminated by heating the mixture at 85°C for 20 min, ensuring inactivation of the enzyme. The resulting slurry was centrifuged (Union 5KR, Hanil, Gangneung, Korea) at 8,000 rpm for 10 min to remove insoluble fractions. The hydrolysate slurry was then heated to 50°C and the pH was adjusted to 7.0 with 2 N NaOH. Bromelain (1%) was added to the mixture and the reaction pH was maintained by the addition of 2 N NaOH. After heating at 90°C for 15 min to inactivate the enzyme, the hydrolysate was centrifuged at 8,000 rpm for 20 min to remove insoluble fractions. The degree of hydrolysis (DH) was determined using the 20% (w/v) trichloroactic acid (TCA) method, as described in the semi-micro Kjeldahl procedure. DH was defined as the percentage ratio of the total nitrogen in two-stage hydrolysate (A) to the total nitrogen in one-stage hydrolysate (B), and calculated as (\[A--B\]/B)× 100 (4.36% crude protein and 24.7% DH). The protein hydrolysate was then stored at 4°C until use. Proximate composition --------------------- The proximate composition analysis of IFP batters including moisture, crude protein, crude fat, and crude ash, was performed according to AOAC methods 950.46, 992.15, 985.15, and 920.153 for sausage batter ([@b3-ajas-27-1-115-13]). Moisture, protein, fat, and ash parameters were determined in triplicate from IFP product. Gel characteristics ------------------- The gel characteristics of cooked IFP samples were determined according to the method described by [@b30-ajas-27-1-115-13]. Three cylindrical pieces 3.5 cm wide and 3 cm thick were maintained at 20°C prior to measuring. The breaking force, deformation, gel strength and jelly strength were measured using a texture analyzer (EZ-test, Shimadzu, Tokyo, Japan) equipped with a cylindrical plunger (diameter 5 mm, depression speed 80 mm/min). Water-holding capacity (WHC) ---------------------------- The water-holding capacity of IFP samples was determined by the method described by [@b12-ajas-27-1-115-13]. Samples (approximately 10 g) were placed in 50 mL plastic centrifuge tubes and heated for 15 min in a water bath (90°C). The samples were then cooled to room temperature and centrifuged at 9,000×*g* at 4°C for 20 min (Union 5KR, Hanil, Korea). The supernatant was eliminated and the WHC of the remaining pellets were calculated as follows: WHC (%) = 1−(\[weight of sample before heating-weight of sample after heating and centrifugation\]/total water content in the sample×100). 2,2-diphenyl-1-picryhydrazla hydrate (DPPH) radical scavenging activity ----------------------------------------------------------------------- The DPPH radical scavenging activity measurement was modified according to the method of [@b5-ajas-27-1-115-13]. 500 μL of each peptide fraction was mixed with 500 μL of ethanol and 250 μL of a DPPH solution (0.5 mM 1,1-diphenyl-2-picrylhydrazyl/ethanol). The mixtures were incubated for 30 min in the dark at room temperature and the reduction of DPPH radicals was measured at 517 nm. DPPH radical scavenging activity was calculated as: DPPH radical scavenging activity (%) = (\[absorption of control − absorption of sample\]/absorption of control)×100. The control was conducted in the same manner, with the exception that distilled water was used instead of sample. Lipid oxidation --------------- Lipid oxidation was determined using the thiobarbituric acid reactive substances (TBARS) method ([@b7-ajas-27-1-115-13]). Cooked IFP sample (5 g) was weighed into a 50 mL test tube and homogenized with 15 mL of deionized distilled water using the Polytron homogenizer for 15 s at the highest speed (T25basic, IKA, Selangor, Malaysia). The IFP sample homogenate (2 mL) was transferred to a disposable test tube (13×100 mm) and butylated hydroxyanisole (10%, 50 μL) and thiobarbituric acid/trichloroacetic acid (TBA/TCA) solutions (4 mL) were added. The sample was mixed using a vortex mixer, incubated in a boiling water bath for 15 min to allow color development, and cooled at room temperature. Absorbance was determined at 531 nm against a blank containing 2 mL of deionized distilled water and 4 mL of TBA/TCA solution. The TBARS measure was expressed in mg of malondialdehyde (MDA) per kg of sample. Angiotensin I-converting enzyme (ACE) inhibitory activity --------------------------------------------------------- ### ACE activity assay The activity of ACE was determined using hippuryl-His-Leu (HHL) with the methods reported by [@b15-ajas-27-1-115-13]. The assay was conducted in a Borate buffer (0.1 M, pH 8.3). The assay volume consisted of 150 μL of the ACE enzyme solution and 40 μL of assay sample. All of the solutions were incubated for 10 min at 37°C. 5 mL of HHL (0.3 M) and 100 μL of 0.1 M Borate buffer (pH 8.3) were added and incubated for 30 min. 1 M HCl (150 μL) was added to stop ACE activity. The reaction mixture obtained was used to quantitate the hippuric acid produced due to ACE activity on the substrate. ### HPLC determination of hippuric acid content A reversed-phase C~18~ column (Bonclone C~18~, 10 μM, 50×1.0 mm) protected by a guard column (Bondclone C~18~, 5 μM, 250×4.6 mm, Phenomenex Co., Torrance, CA, USA) was employed. The injection volume used was 20 μL. Elution of hippuric acid was detected by monitoring the absorbance at 228 nm. The control reaction mixture contained 40 μL of buffer instead of the assay sample. The percent inhibition of enzyme activity was calculated as follows: ACE inhibition (%) = (\[hippuric acid of control-hippuric acid of sample\]/hippuric acid of control)×100. The concentration of hydrolysate needed to inhibit 50% of ACE activity was defined as the IC~50~ value. Statistical analysis -------------------- The study was designed as a 3×4 factorial experiment with treatment (control, T1 and T2) and storage time (0, 2, 4, and 6 weeks). The experiment was replicated three times. Data were analyzed using the general linear model (GLM) ([@b34-ajas-27-1-115-13]). Duncan's multiple test was used to determine the statistical significance among the means at a 95% significance level. RESULTS AND DISCUSSION ====================== Proximate composition --------------------- The proximate composition of IFP with added MDCM hydrolysates is shown in [Table 2](#t2-ajas-27-1-115-13){ref-type="table"}. In this study, the IFP showed no significant difference in crude protein compared to the control during storage. However, all treatments with MDCM hydrolysates had lower moisture content compared to control at 0 and 6 weeks storage periods (p\<0.05). Crude fat content in the MDCM hydrolysate-added groups was significantly higher than in the control at week 0 of storage. Likewise, the ash content in the 0.8% MDCM hydrolysate-added sample was significantly lower than in the other groups at 6 weeks of storage (p\<0.05). The functional and textural properties of surimi depend on many factors including thermal conditions and various gelling and non-gelling ingredients ([@b26-ajas-27-1-115-13]). In general, high protein, low fat, and adequate water are required to make high quality surimi-based products ([@b17-ajas-27-1-115-13]), Moisture and fat are critical factors in surimi products ([@b40-ajas-27-1-115-13]) and excessive lipids may adversely affect quality, due to oxidized lipids interacting with proteins ([@b36-ajas-27-1-115-13]). Also, protein concentration greatly affects the gel properties ([@b23-ajas-27-1-115-13]). However, we assumed that the protein contents would not influence the physical properties of IFP in this study because protein showed no consistent trends among the samples. Gel characteristics ------------------- Gel characteristics of IFP with added MDCM hydrolysates are shown in [Table 3](#t3-ajas-27-1-115-13){ref-type="table"}. The breaking force values increased during storage in T1 and T2. The breaking force was lower in T2 as compared to T1 until 2 weeks storage (p\<0.05). The gel strength and jelly strength showed consistent trends among the IFP samples; however, treatment with MDCM hydrolysates imparted higher breaking force values than control group at 4 and 6 weeks. Also, deformation, gel strength, and jelly strength were improved with added MDCM compared to the control group at 4 or 6 weeks (p\<0.05). Protein concentration has a major positive effect on the breaking force ([@b24-ajas-27-1-115-13]). It has also been reported ([@b38-ajas-27-1-115-13]) that the gel-forming ability of surimi increases with decreasing water content, a result of higher myofibril protein concentrations and increased cross-link density. Therefore, our results indicate that the gel characteristics are favorably influenced by changes of proximate compositions and the addition of MDCM to IFP. Water holding capacity (WHC), DPPH radical scavenging activity and lipid oxidation ---------------------------------------------------------------------------------- Water holding capacity (WHC), DPPH radical scavenging activity, and lipid oxidation of IFP with added MDCM hydrolysates are shown in [Table 4](#t4-ajas-27-1-115-13){ref-type="table"}. The WHC was lower when MDCM hydrolysates were added compared to the control group during storage (p\<0.05). During cooking, various meat proteins denature causing structural changes, shrinkage of meat fibers, and gel formation of myofibrillar proteins ([@b39-ajas-27-1-115-13]). Therefore, our current results suggest that the addition of MDCM to IFP may not directly affect WHC, but that with extensive degradation of muscle fibers an apparent WHC reduction occurs. All treatment samples demonstrated increased DPPH radical scavenging activity during storage, while IFP showed significantly decreased DPPH radical scavenging activity after 2 weeks (p\<0.05). DPPH radical scavenging activity was higher in the MDCM hydrolysate treatment groups compared to the control up until 4 weeks. The free radical scavenging activity and antioxidant activity; including the ability to donate hydrogen, to stabilize or terminate radicals, to sequester pro-oxidative metal ions, and to form a physical barrier around fat droplets, is determined by a specific amino acid composition and sequence ([@b18-ajas-27-1-115-13]). Additionally, hydrolysate antioxidant activity depends upon the enzyme used. It has been reported ([@b42-ajas-27-1-115-13]) that DPPH activity improved for mackerel protein hydrolysate generated by Protease N. Another group ([@b43-ajas-27-1-115-13]) found that porcine hemoglobin hydrolysates, prepared through hydrolysis by Alcalase followed by Flavourzyme, exhibited high ferrous ion chelating abilities and DPPH radical scavenging activity. In our research, the DPPH radical scavenging activities of the Alcalase hydrolysates were similar to those reported by [@b43-ajas-27-1-115-13]. The TBARS value increased in correlation with increasing storage periods in all IFP samples. In the treatment groups, TBARS values rapidly increased at 2 weeks (p\<0.05) while they did not differ significantly among IFP samples until 4 weeks. TBARS values were higher in the MDCM hydrolysate treatment groups compared to the control at 6 weeks (p\<0.05). [@b2-ajas-27-1-115-13] reported that differences in fat content, fatty acid composition, and the classes of lipids present effected the lipid oxidation of stored turkey patties. Meat products with a high degree of unsaturation are more susceptible to lipid oxidation ([@b25-ajas-27-1-115-13]). Accordingly, lipid oxidation in the present study was influenced by fat content ([Table 2](#t2-ajas-27-1-115-13){ref-type="table"}). Previously, [@b33-ajas-27-1-115-13] reported reduction of TBARS values in beef homogenate by egg-yolk protein hydrolysates. However, hydrolysates possess a lower antioxidant activity in meatballs ([@b10-ajas-27-1-115-13]). Hydrolysates at doses ranging from 1% to 2% slowed lipid oxidation of pork patties but their activity remained weaker than synthetic antioxidants ([@b29-ajas-27-1-115-13]). Angiotensin I-converting enzyme (ACE) inhibitory activity --------------------------------------------------------- The ACE inhibitory activity in IFP with added MDCM hydrolysates is shown in [Figure 2](#f2-ajas-27-1-115-13){ref-type="fig"}. ACE inhibitory activity of IFP showed a decreasing trend with increasing storage time. However, the ACE activity in the IFP with added treatments was lower than that of control. Addition of 0.8% MDCM hydrolysates was found to be more effective at inhibiting ACE activity of IFP compared to addition of 0.4% MDCM hydrolysates (p\<0.05). ACE inhibitory peptides have been discovered in various animal sources such as porcine and chicken muscle. Specific ACE inhibitory peptides, Met-Asn-Pro-Asn (IC~50~ = 66.6 μM), Asn-Pro-Pro (IC~50~ = 290.5 μM), and Thr-Asn-Pro (IC~50~= 207.4 μM) were discovered in a porcine myosin hydrolysate ([@b4-ajas-27-1-115-13]). Gly-Phe-Hyp-Gly-Thr-Hyp-Gly-Leu-Hyp-Gly-Phe (IC~50~ = 42 μM) was isolated from chicken breast muscle hydrolysate ([@b32-ajas-27-1-115-13]). [@b1-ajas-27-1-115-13] reported that mixing 5% meat hydrolysate from the porcine muscle biceps femoris with normal diet in rats resulted in clear positive effects on a common lifestyle-related disease such as hypertension. ACE plays an important role in the regulation of blood pressure as well as fluid and salt balance in mammals. Therefore, our results support the conclusion that MDCM hydrolysates can be used as a good source of health-promoting constituents in functional foods. This work was supported by Technology Development Program for Ministry for Food, Agriculture, Forestry and Fisheries and the Priority Research Centers Program through the national research foundation of Korea (FRF) funded by the Ministry of Education, Science and Technology, Republic of Korea. Also, following are results of a study on the "Leaders in Industry-university Cooperation" project, supported by the Ministry of Education. {#f1-ajas-27-1-115-13} {ref-type="table"}.](ajas-27-1-115-13f2){#f2-ajas-27-1-115-13} ###### The basic formulation of imitation fish paste Ingredients (%) Control T1 T2 ------------------------------------------------------------------- --------- ------- ------- Alaska Pollack 59.70 59.70 59.70 Spent laying hen surimi 14.93 14.93 14.93 Fresh egg white 4.72 4.72 4.72 Soy protein 0.94 0.94 0.94 Sugar 1.51 1.51 1.51 Salt 1.51 1.51 1.51 Monosodium glutamate 1.26 1.26 1.26 Seasoning mix 0.31 0.31 0.31 Wheat starch 6.30 6.30 6.30 Distilled water 8.82 8.42 8.02 MDCM[1](#tfn1-ajas-27-1-115-13){ref-type="table-fn"} hydrolysates \- 0.4 0.8 Total 100 100 100 Mechanically deboned chicken meat. ###### Proximate composition of imitation fish paste batter with added MDCM hydrolysates Treatments[1](#tfn5-ajas-27-1-115-13){ref-type="table-fn"} (g/100 g) Storage periods (weeks) ---------------------------------------------------------------------- ---- ------------------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------ Moisture C 69.74±0.10[A](#tfn3-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn4-ajas-27-1-115-13){ref-type="table-fn"} 69.22±0.11[A](#tfn3-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn4-ajas-27-1-115-13){ref-type="table-fn"} T1 68.87±0.12[B](#tfn3-ajas-27-1-115-13){ref-type="table-fn"} 68.67±0.30[B](#tfn3-ajas-27-1-115-13){ref-type="table-fn"} T2 69.00±0.10[B](#tfn3-ajas-27-1-115-13){ref-type="table-fn"} 68.73±0.17[B](#tfn3-ajas-27-1-115-13){ref-type="table-fn"} Crude protein C 18.16±0.17 18.82±0.32 T1 18.39±0.52 18.72±0.54 T2 17.28±0.18[b](#tfn4-ajas-27-1-115-13){ref-type="table-fn"} 18.02±0.08[a](#tfn4-ajas-27-1-115-13){ref-type="table-fn"} Crude fat C 0.80±0.02[B](#tfn3-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn4-ajas-27-1-115-13){ref-type="table-fn"} 1.20±0.01[a](#tfn4-ajas-27-1-115-13){ref-type="table-fn"} T1 1.02±0.06[A](#tfn3-ajas-27-1-115-13){ref-type="table-fn"} 1.10±0.20 T2 1.07±0.09[A](#tfn3-ajas-27-1-115-13){ref-type="table-fn"} 1.15±0.05 Ash C 0.76±0.13 0.71±0.03[A](#tfn3-ajas-27-1-115-13){ref-type="table-fn"} T1 0.87±0.17 0.73±0.01[A](#tfn3-ajas-27-1-115-13){ref-type="table-fn"} T2 0.75±0.12[a](#tfn4-ajas-27-1-115-13){ref-type="table-fn"} 0.49±0.01[B](#tfn3-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn4-ajas-27-1-115-13){ref-type="table-fn"} Data are means±standard deviation. *n* =3. Means with different superscript capital letters in a column within each treatment differ significantly (p\<0.05). Means with different superscript small letters in a row within each storage time differ significantly (p\<0.05). Treatments are the same as in [Table 1](#t1-ajas-27-1-115-13){ref-type="table"}. ###### Changes in gel characteristics in imitation fish paste with added MDCM hydrolysates during cold storage Treatments[1](#tfn9-ajas-27-1-115-13){ref-type="table-fn"} Storage periods (weeks) ------------------------------------------------------------ -------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------- Breaking force (g) C 379.67±7.51[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} 408.00±20.07[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} 382.33±20.53[B](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} 357.00±26.46[B](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} T1 374.67±6.43[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[c](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 394.74±10.74[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[bc](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 435.33±24.66[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 417.00±10.00[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[ab](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} T2 341.33±16.77[B](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[c](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 355.00±8.66[B](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[c](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 433.33±12.58[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 398.0±3.61[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} Deformation (mm) C 6.91±0.21[a](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 6.71±0.06[a](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 5.81±0.32[C](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 5.71±0.21[B](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} T1 6.67±0.25 6.47±0.35 6.31±0.12[B](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} 6.61±0.35[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} T2 6.71±0.21 6.51±0.15 6.91±0.23[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} 6.77±0.15[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} Gel strength (g/cm^2^) C 262.32±12.96[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 273.70±15.58[a](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 231.43±37.19[B](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[ab](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 203.36±7.96[B](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} T1 249.93±5.95[AB](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} 262.79±25.52 274.63±18.13[AB](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} 275.63±18.99[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} T2 228.97±14.58[B](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[c](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 231.00±8.49[c](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 299.27±12.61[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 269.56±5.40[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} Jelly strength C 193.62±3.82[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} 207.92±10.22[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} 202.60±23.43 181.18±13.47[B](#tfn7-ajas-27-1-115-13){ref-type="table-fn"} T1 190.16±3.27[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 206.34±8.94[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 221.13±12.56[a](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 212.76±5.09[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} T2 173.39±8.54[B](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[c](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 180.00±4.41[B](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[c](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 220.94±6.40[a](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} 202.99±1.83[A](#tfn7-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn8-ajas-27-1-115-13){ref-type="table-fn"} Data are means±standard deviation. *n =* 3. Means with different superscript capital letters in a column within each treatment differ significantly (p\<0.05). Means with different superscript small letters in a row within each storage time differ significantly (p\<0.05). Treatments are the same as in [Table 1](#t1-ajas-27-1-115-13){ref-type="table"}. ###### Changes in water-holding capacity (WHC), DPPH radical scavenging activity and TBARS in imitation fish paste with added MDCM hydrolysates during cold storage Treatments[1](#tfn13-ajas-27-1-115-13){ref-type="table-fn"} Storage periods (weeks) ------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------- WHC (%) C 69.74±0.10[A](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 69.40±0.10[A](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 69.71±0.02[A](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 69.22±0.11[A](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[c](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} T1 68.82±0.12[B](#tfn11-ajas-27-1-115-13){ref-type="table-fn"} 69.05±0.09[B](#tfn11-ajas-27-1-115-13){ref-type="table-fn"} 68.97±0.12[B](#tfn11-ajas-27-1-115-13){ref-type="table-fn"} 68.67±0.30[B](#tfn11-ajas-27-1-115-13){ref-type="table-fn"} T2 69.00±0.10[B](#tfn11-ajas-27-1-115-13){ref-type="table-fn"} 68.61±0.14[C](#tfn11-ajas-27-1-115-13){ref-type="table-fn"} 68.58±0.25[C](#tfn11-ajas-27-1-115-13){ref-type="table-fn"} 68.73±0.17[B](#tfn11-ajas-27-1-115-13){ref-type="table-fn"} DPPH (%) C 19.70±1.00[C](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 21.76±1.66[C](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 17.63±1.21[B](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[c](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 13.91±1.16[d](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} T1 21.20±0.46[B](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 23.69±1.69[B](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 19.00±1.29[A](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[c](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 14.15±1.11[d](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} T2 23.98±0.91[A](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[b](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 25.44±1.42[A](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 20.65±0.84[A](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[c](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 14.44±0.74[d](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} TBARS (mg/100 g) C 0.51±0.03[c](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 2.27±0.01[b](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 2.62±0.06[a](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 2.49±0.09[C](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[ab](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} T1 0.49±0.04[c](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 2.50±0.02[b](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 2.56±0.04[ab](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 2.62±0.04[B](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} T2 0.43±0.03[d](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 2.50±0.05[c](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 2.60±0.04[b](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} 2.82±0.01[A](#tfn11-ajas-27-1-115-13){ref-type="table-fn"}[a](#tfn12-ajas-27-1-115-13){ref-type="table-fn"} Data are means±standard deviation. *n* = 3. Means with different superscript capital letters in a column within each treatment differ significantly (p\<0.05). Means with different superscript small letters in a row within each storage time differ significantly (p\<0.05). Treatments are the same as in [Table 1](#t1-ajas-27-1-115-13){ref-type="table"}. [^1]: The Department of Internal Medicine, Yale University School of Medicine, New Haven, CT 06510, USA. [^2]: Division of Applied Life Science, Graduate School, Gyeongsang National University, Jinju, Gyeongnam 660-701, Korea. [^3]: Department of Animal Science Institute of Agriculture and Life Science, Gyeongsang National University, Jinju, Gyeongnam 660-701, Korea. [^4]: Swine Science and Technology Center, Gyeongnam National University of Science and Technology, Jinju, Gyeongnam 660-758, Korea. [^5]: These authors equally contributed

Introduction {#s1} ============ Malignant pleural mesothelioma (MPM) is a lethal cancer arising from pleura mesothelial cells, showing a close association with previous exposure to asbestos. This tumor is characterized by long latency period (20--30 years) and slow growth which cause late diagnosis, poor prognosis, and limited effective therapies. It has also been suggested that additional factors besides asbestos may play a role in the tumor pathogenesis, such as SV40 infection [@pone.0058051-Kroczynska1] and genetic predisposition [@pone.0058051-Carbone1]. The problem presented by the disease is exacerbated by the lack of reliable biological markers to be used for early screening, and by its rapid progression following diagnosis, resulting in a median survival time of about 10--12 months [@pone.0058051-Montanaro1]. Despite pre-clinical and clinical efforts, there is currently no effective therapeutic approach to MPM. Decisions of carrying out surgery, radiotherapy, chemotherapy or multimodal procedures are taken on a case-by-case basis, and frequently a palliative treatment is the only choice available [@pone.0058051-West1]. Intrusive surgical procedures, based on extrapleural pneumonectomy and pleurectomy, are not suitable for most of the patients due to locally advanced or unresectable disease [@pone.0058051-Sugarbaker1]. Radiotherapy is mainly used as adjuvant therapy following surgery or for symptom relief [@pone.0058051-Belli1]. In locally advanced or metastatic disease, chemotherapy improves the quality of life and alleviates symptoms. However, the tumor is generally chemoresistant, and most single-agent treatments exhibit low intrinsic activity [@pone.0058051-Ellis1]. Response rates and survival are generally improved by using combination of drugs rather than by single-agent regimens. Combined therapies of cisplatin with antimetabolites are more effective than each single agent alone, and currently represent the standard treatment for MPM [@pone.0058051-vanMeerbeeck1], [@pone.0058051-Scherpereel1]. However, patient response rates by far below 50%, and the prognosis remains poor. Other approaches, including gene therapy, vaccines and molecular target therapies are under evaluation, but the need of new therapies for this malignancy is compelling [@pone.0058051-Zauderer1]. Among alternative remedies for cancer treatment, there is a growing interest in the preventive action of active nutrients, like vitamins [@pone.0058051-ReaganShaw1]. Several studies suggest that these molecules could also be exploited in a pharmacologic way. Vitamin E analogues, like α-tocopheryl succinate, have been reported to selectively trigger mitochondrial apoptosis in tumor cells [@pone.0058051-Constantinou1], while ascorbate, also known as vitamin C, has already been used in clinical trials as an alternative cancer therapy [@pone.0058051-Chen1], [@pone.0058051-Cameron1]. Based on these data, we decided to investigate the effects of combined active nutrients and pharmaceutical drugs on MPM in a pre-clinical model. Antitumor nutrients are generally better tolerated by the organism than chemotherapeutic drugs, and can both increase the efficacy and allow for lower, safer dosages of these drugs. In a previous study, we have shown that ascorbate exerts a cytotoxic action on MPM cells, with a lower effect on normal, non-neoplastic mesothelial cells. Ascorbate administration induces extracellular H~2~O~2~ production coupled with an intrinsic higher level of reactive oxygen species (ROS) in MPM cells [@pone.0058051-Ranzato1]. These results encouraged us to employ ascorbate in our study, in association with other anti-tumor agents. A series of *in vitro* tests on MPM cells has revealed a synergistic cytotoxicity of ascorbate in combination with the conventional tumor drug gemcitabine, and with the green tea polyphenol epigallocatechin-3-gallate (EGCG) [@pone.0058051-Martinotti1]. Gemcitabine is one of the most effective single agents on MPM and is currently used both in combination with chemo/targeted therapy, as a first-line treatment, and as a single agent for second line treatment [@pone.0058051-Kindler1]. EGCG has been found to exert antitumor activity in many cancer models [@pone.0058051-Cooper1], [@pone.0058051-Yang1]. Even though EGCG is generally known as an antioxidant, mounting evidence points a role in enhancing ROS release, which in turn inhibits tumor growth [@pone.0058051-Li1], [@pone.0058051-Azam1]. In line with these findings, we have previously shown *in vitro* that EGCG is more cytotoxic for MPM cells than for normal mesothelial cells, through a mechanism of action based on extracellular H~2~O~2~ production, Ca^2+^ homeostasis loss, and intracellular ROS increase [@pone.0058051-Ranzato2]. In the present preclinical study, we have investigated the *in vitro* interaction of ascorbate with both EGCG and gemcitabine, a triple combined treatment herein defined AND therapy (Active Nutrients/Drug). Thereafter, we have studied the effects of intraperitoneal injections of AND on MPM tumor xenografts growing in the peritoneum of immunodeficient mice. Materials and Methods {#s2} ===================== Reagents and solutions {#s2a} ---------------------- (−)-Epigallocatechin-3-gallate (EGCG) was purchased from Cayman Chemical Co. (Ann Arbor, MI, USA); gemcitabine (Gemzar) was from Ely Lilly Italia S.p.A. (Sesto Fiorentino, Italy); L-ascorbic acid (ascorbate), was from Sigma (St. Louis, MO, USA). All other reagents were from Sigma, unless otherwise specified. Ascorbate and EGCG were directly dissolved in culture medium, pH 7.4, while gemcitabine stock solution was prepared in 0.9% NaCl in ultrapure water. For *in vivo* injections, the compounds were dissolved in sterile 0.9% NaCl and the solutions filter-sterilized under hood. L-ascorbic acid was stored as a powder and dissolved immediately prior to use. In vitro cell culture {#s2b} --------------------- The following human MPM cell lines were available at our laboratory: REN cells are a p53-mutant, inflammatory epithelial subtype [@pone.0058051-Smythe1]; MM98 cells were established from pleural effusion of a sarcomatous MPM [@pone.0058051-Orengo1]; BR95 epithelial cells were obtained from pleural effusions of MPM patients with histologically confirmed malignant mesothelioma [@pone.0058051-Orecchia1]; MPP89 are epithelial mesothelioma cells [@pone.0058051-Orengo1]. In addition, epithelial NCI-H28, having a wild-type p53 [@pone.0058051-Maeda1], were purchased from ATCC (cat. no. CRL-5820™, Rockville, MD, USA). Cells were cultured in DMEM supplemented with 10% foetal bovine serum (FBS, Euroclone, Pero, Italy) and 1% antibiotic mixture (Gibco, Invitrogen Life Technologies, S. Giuliano Milanese, Italy), and maintained at 37°C in a humidified atmosphere with 5% CO~2~. Cytotoxicity assay {#s2c} ------------------ The calcein cytotoxicity assay was carried out by using the lipophilic, nonfluorescent calcein acetoxymethylester (calcein-AM), which penetrates cell membranes and is then cleaved by intracellular esterases, yielding the hydrophilic fluorescent dye. Cells growing in 96-well plates were treated as specified, washed with PBS, and then incubated for 30 min at 37°C with a solution of 2.5 μM calcein-AM in PBS. Plates were read in a fluorescence reader (Infinite 200 Pro, Tecan, Wien, Austria), by using 485-nm excitation and 535-nm emission filters. In vitro drug interaction analysis {#s2d} ---------------------------------- Dose response curves and IC~50~ values, based on the calcein-AM assay at 48 h, were first derived for single compounds (ascorbate, EGCG, or gemcitabine), as described in Martinotti *et al.* [@pone.0058051-Martinotti1]. The concentrations used in these experiments are reported in [Table 1](#pone-0058051-t001){ref-type="table"}. Thereafter, IC~50~ were derived for the AND mixture (ascorbate/EGCG/gemcitabine) by using a constant ratio combination design consisting of serial dilutions of the equipotency concentrations of single compounds ([Table 2](#pone-0058051-t002){ref-type="table"}). After having obtained IC~50~ values, the AND combination was analysed for synergy as described in Martinotti *et al*. [@pone.0058051-Martinotti1], by using Chou and Talalay\'s Combination Index (CI) [@pone.0058051-Chou1].where CI~x~ is the combination index at effect level x% ( =  percent of viability inhibition); (D)~A~, (D)~B~ and (D)~C~ are the doses of drugs A, B and C that combined together inhibit cell viability by x%; (D~x~)~A~, (D~x~)~B~ and (D~x~)~C~ are the doses of drugs A, B and C that inhibit cell viability by x% when used alone. If the CI value is \<1,  = 1, or \>1, then synergism, additivity or antagonism is indicated, respectively. Variations of drug interaction at different levels of inhibition can be visualized by a plot of CI on the y-axis as a function of effect levels f~a~ on the x-axis (f~a~-CI plot). 10.1371/journal.pone.0058051.t001 ###### Concentration series used to derive dose-response curves for single compounds. {#pone-0058051-t001-1} Compound Concentrations (µM) ------------- --------------------- ----- ----- ----- ----- ----- ----- ----- ------ ascorbate 0 50 100 150 200 250 300 500 1000 EGCG 0 1 5 10 15 20 30 50 gemcitabine 0 0.1 1 2.5 5 10 20 30 10.1371/journal.pone.0058051.t002 ###### Concentration series used to derive dose-response curve for AND in REN cells. {#pone-0058051-t002-2} Dilution ratios (µM) --------------------------------- ----------- ------ ------ ---------- -------- ------ ----- ------ Single compounds in mixture ascorbate 4.75 9.5 19 38 76 228 684 EGCG 0.42 0.83 1.7 3.33 6.7 20 60 gem 0.18 0.37 0.74 1.48 2.97 8.9 26.7 Total concentrations in mixture AND 5.35 10.7 21.4 42.8 85.7 257 771 Animals and in vivo experiments {#s2e} ------------------------------- Male NOD-SCID CB13 mice (6--8 weeks old) were purchased from Charles River Laboratories Italia Srl (Calco, Italy), and housed for 3--4 days before experiments. Mice were maintained and handled under aseptic conditions, were allowed access to food and water ad libitum, and received i.p. injections of 10×10^6^ REN cells in 1.0 mL of PBS. Mice injected with cells were randomly divided into different treatment groups. Treatments were carried out every 3^rd^ day by i.p. injections. *In vivo* experiments were done in accordance with institutional animal committee guidelines. Necropsy and histochemical analyses {#s2f} ----------------------------------- At the end of *in vivo* experiments, mice were sacrificed and their abdominal cavity was opened and photographed. Complete necropsy was performed with collection of tumors and major organs and tissues. Necropsied tissues were rapidly frozen in liquid nitrogen and stored at −80°C until use. Part of tumor tissues were fixed in 10% buffered neutral formalin, processed to paraffin and sectioned at 5 µm. Slides were stained with hematoxylin and eosin (H&E) for morphological analysis or used for immuno-histochemistry. Sections were deparaffinized with xylene and graded alcohol, and rehydrated in PBS. Endogenous peroxidases were blocked with 3.0% H~2~O~2~ in PBS. Apoptotic cells were identified on sections using an indirect TUNEL labeling assay (In Situ Cell Death Detection Kit, AP, Roche), according to manufacturer\'s protocol. Cell proliferation was evaluated by PCNA histochemistry (Abcam, Cambridge, UK), using the Vectastain Elite ABC kit (Vector Laboratories, Burlingame, CA, USA), according to manufacturer\'s instructions. Angiogenesis antibody array {#s2g} --------------------------- Angiogenesis factors were quantified using the Human Angiogenesis Antibody Array (Panomics, Inc., Redwood City, CA). The array allows for simultaneous detection of 19 factors and provides positive and negative controls. Tumor samples were lysed [@pone.0058051-Dorrell1] and hybridized to each membrane of an antibody-sandwich angiogenesis array according to manufacturer\'s guidelines. Spots were observed and digitized with the Quantity One Imaging system (ChemiDoc XRS, Bio-Rad, Hercules, CA). Multiplex analysis of phosphorylated proteins {#s2h} --------------------------------------------- The phosphorylation of specific signal transduction proteins was analyzed using the Bio-Plex TM bead suspension array system (Bio-Rad), allowing the assay of multiple proteins in a single well. Tumor samples were homogenized in a lysis solution (Bio-Rad), vortexed, centrifuged at 10,000 *g* for 4 min and the supernatant collected. Lysates were adjusted to 1,000 µg/mL protein for use in an assay for 6 different phosphorylated proteins, including Akt (Ser473), Erk 1/2 (Thr202/Tyr204, Thr185/Tyr187), JNK (Thr183/Tyr185), p38 MAPK (Thr180/Tyr182), p70 S6 kinase (Thr421/Ser424), IκBα (Ser32/Ser36). Samples were prepared according to the manufacturer\'s instructions and sent to Bioclarma srl (Turin, Italy) for fluorescence recording and data analysis. Statistics {#s2i} ---------- ANOVA and post hoc tests were carried out using the Instat Software package (GraphPad Software, Inc.). Survival curves were evaluated by the Kaplan-Meier method and compared by the log-rank test [@pone.0058051-Pepe1]. Results {#s3} ======= AND therapy has a synergistic action in decreasing the viability of MPM cell lines {#s3a} ---------------------------------------------------------------------------------- In a previous study based on the Chou and Talalay\'s combination index (CI) method [@pone.0058051-Chou1], we had shown that ascorbate/gemcitabine and ascorbate/EGCG combinations have synergistic cytotoxic activity on REN cells, an established cell model of MPM [@pone.0058051-Martinotti1]. We used here the same method to investigate *in vitro* effects of the AND combination (ascorbate/EGCG/gemcitabine) on various MPM cell lines, by using the calcein-AM assay at 48 h. Dose-response curves and IC~50~ values were obtained for each compound alone, and these data were then used to derive dose-response curve and IC~50~ for the triple combination ([Table 3](#pone-0058051-t003){ref-type="table"}). A comparison of the IC~50~s of single compounds obtained on different cell lines showed that ascorbate was the least cytotoxic compound, while gemcitabine was the most cytotoxic one, or was similar to EGCG. An exception was made by NCI-H28, for which a low toxicity of gemcitabine was recorded, possibly depending on the lower growth rate of these cells. 10.1371/journal.pone.0058051.t003 ###### Values of IC~50~ (µM) determined on different MPM cells by the calcein-AM assay at 48 h. {#pone-0058051-t003-3} cell type ascorbate EGCG gemcitabine AND ----------- ---------------- ------------- ----------------- ---------------- REN 228 (202--258) 20 (18--22) 8.9 (6.1--12.8) 38 (26--54) MM98 47 (36--63) 25 (24--26) 1.1 (0.6--1.9) 13 (12--14) BR95 189 (175--204) 66 (61--71) 1.6 (0.8--3.1) 83 (77--89) NCI-H28 706 (645--772) 70 (61--80) 915 (801--1045) 192 (148--249) MPP89 176 (159--194) 15 (14--16) 23 (7--77) 37 (30--46) Dose concentration curves showing cell viability (calcein-AM assay) for each single compound and the AND mixture, and for each cell type are shown in [Fig. S1](#pone.0058051.s001){ref-type="supplementary-material"} (Supplementary Information). Dose-response experiments were carried out in duplicate, with a minimum of 6 replicates for each dose. 95% confidence intervals are given in parentheses. Data from single and combined treatments were used to assess synergy by Combination Index (CI) analysis, as described in the Methods. The f~a~-CI plots, depicting CI values vs f~a~ (fraction of inhibited viability), showed the occurrence of synergistic effects (CI \<1) in all MPM cells, although at variable extents in different cell types ([Fig. 1](#pone-0058051-g001){ref-type="fig"}). The strongest synergism at mid-to-high f~a~ was observed in REN cells. We therefore used these cells for *in vivo* experiments, also considering that they have already been documented to be tumorigenic [@pone.0058051-Bertino1]. {#pone-0058051-g001} AND reduces tumor burden, metastasization and tumor-induced hemorrhage in an MPM mouse model {#s3b} -------------------------------------------------------------------------------------------- Pilot tests showed the development of tumor xenografts in all animals injected with REN cells. In final experiments, REN-inoculated animals were randomly stratified into 4 groups of 5 individuals. Each group was exposed to one of the following treatments: (i) gemcitabine alone, used as a reference therapy; (ii) an ascorbate/EGCG mixture, to evaluate the efficacy of active nutrients without a chemotherapeutic drug; (iii) the AND mixture; (iv) a placebo consisting of 0.9% NaCl. Treatments were made every 3^rd^ day, while doses were designed on the basis of literature reports and the results of our preliminary tests. Gemcitabine as single agent was used at 150 mg/kg [@pone.0058051-Amoh1], [@pone.0058051-Bocci1]; ascorbate/EGCG were used at 2,000 mg/kg and 30 mg/kg, respectively [@pone.0058051-Chuu1], [@pone.0058051-Isbrucker1]; the AND therapy consisted of 2,000 mg/kg ascorbate, 30 mg/kg EGCG, and 100 mg/kg gemcitabine. Such doses were previously tested on tumor -- free mice and found to be tolerable for a period of 30 days. After 30 days of treatment, animals were sacrificed and necropsied. Untreated mice showed different symptoms of disease including severe ascites, the development of a main tumor, different small tumor nodules at various locations in the peritoneal cavity (mainly on the right kidney, liver and colon), a diffuse metastasization of diaphragm and other alterations such as splenomegaly and intraperitoneal hemorrhage ([Fig. 2](#pone-0058051-g002){ref-type="fig"}). The main tumor was located more prevalently on intestinal fat localized on the left side close to the stomach. This was the side subjected to cell injection. All treatments significantly reduced the number and weight of tumor masses, and the degree of hemorrhage and diaphragm tumor coverage, evaluated by arbitrary scores ([Fig. 2](#pone-0058051-g002){ref-type="fig"}). However, the strongest reduction in primary tumor development and metastasis was achieved by the AND treatment, along with the total absence of abdominal hemorrhage. Further analyses were therefore focused on this specific treatment. {#pone-0058051-g002} AND inhibits cell growth signaling pathways in tumor cells {#s3c} ---------------------------------------------------------- The effects of AND on angiogenesis were evaluated by using an array including antibodies for the following factors: angiotensin, VEGF, TNF-α, INF-γ, IL-1α, IL-1β, IL-6, IL-8, IL-12, G-CSF, IP-10, leptin, FGF-α, FGF-β, HGF, PIGF, TGF-β, TIMP-1, and TIMP-2. Tumors excised from mice after 30 days of treatment exhibited low expression levels for most of these factors, and no significant variations induced by AND. The only exception was a reduction of tissue inhibitors of matrix metalloproteinases TIMP-1 and -2 ([Fig. 3A](#pone-0058051-g003){ref-type="fig"}). However, a multiplex analysis of signal transduction proteins involved in cell proliferation and growth, including Akt, ERK1/2, IκBα, JNK, p38, and S6K, displayed a general abatement of phosphorylation levels upon AND treatment ([Fig. 3B](#pone-0058051-g003){ref-type="fig"}). {#pone-0058051-g003} AND increases overall survival of xenograft mice {#s3d} ------------------------------------------------ In a second *in vivo* experiment, to assess animal survival, mice were injected with REN cells as above, and then randomly stratified into 2 groups of 6 animals (treated) and one group of 10 animals (control). Treatments started 7 days after cell injection and were made every 3^rd^ day, as above. Due to an expected longer duration of the survival experiment, a dose-escalating treatment was adopted. Mice were treated with 30 mg/kg gemcitabine as single agent; with an AND mixture of 20 mg/kg gemcitabine, 30 mg/kg EGCG, and 2,000 mg/kg ascorbate; or control treated for 14 days. Thereafter, doses were increased to 50 mg/kg for gemcitabine as single agent, and to 30 mg/kg gemcitabine, 30 mg/kg EGCG, and 2,000 mg/kg ascorbate for AND, for an additional period of 14 days. Mice died spontaneously or were sacrificed to avoid excessive suffering (mainly due to heavy intraperitoneal hemorrhage with abdominal enlargment). Kaplan-Meier curves and the log-rank test indicated a significant increase of overall survival in animals treated with AND or gemcitabine, as compared to control-treated ([Fig. 4](#pone-0058051-g004){ref-type="fig"}). Most AND-treated animals survived for a longer time than the gemcitabine-treated ones, but the trend was not statistically significant. However, AND contained lower dosage of gemcitabine. {#pone-0058051-g004} AND inhibits tumor cell proliferation {#s3e} ------------------------------------- Tissues from xenograft tumors were excised, formalin fixed, paraffin embedded, and stained for markers of cell proliferation and apoptosis. The histology of REN-derived tumor masses revealed a more densely cellularized superficial sheath surrounding a central mass of fibrous tissue with scattered cells. According to such tissue organization, histochemistry revealed a prevalence of cell proliferation at the tumor surface (according to PCNA staining, [Fig. 5A](#pone-0058051-g005){ref-type="fig"}), whereas the internal portion was characterized by a prevalence of apoptosis (according to TUNEL assay, [Fig. 5C](#pone-0058051-g005){ref-type="fig"}). These data suggest that tumor growth occurs mainly at the surface while more internal portions progressively transform into fibrous tissue, possibly due to hypoxic conditions. In AND-treated tumors, staining patterns were drastically different. Surface layers showed a drop in cell proliferation accompanied by an increase in apoptosis ([Fig. 5B, D](#pone-0058051-g005){ref-type="fig"}). This suggests that AND treatment diffusing from the surface toward the interior of tumor mass blocks cancer proliferation. {#pone-0058051-g005} Discussion {#s4} ========== Here we show that a triple combined treatment based on EGCG, ascorbate and gemcitabine (AND therapy) reduces mesothelioma growth and metastasization. Due to the lack of side effects, we propose that this combined therapy should be evaluated in other preclinical and clinical models. Ascorbate, a known active nutrient, is well tolerated by the human body and exerts antitumor effects both *in vitro* and *in vivo* [@pone.0058051-Padayatty1], [@pone.0058051-Verrax1]. We therefore used ascorbate on MPM cells *in vitro*, and showed selective cytotoxicity due to a maladaptive redox mechanism of these cells causing strong oxidative stress [@pone.0058051-Ranzato1]. However, given the chemoresistance of MPM, our therapeutic goal was to combine ascorbate with other drugs in order to maximally strengthen the final effect through a synergistic mixture. In an *in vitro* screening of various ascorbate/drug combinations, two compounds showed synergistic effects against MPM cells, viz. the standard antitumor gemcitabine, and EGCG, an active nutrient with antitumor properties [@pone.0058051-Martinotti1]. The next step has been to combine ascorbate with these two compounds in the triple AND treatment and achieve a preclinical assessment of its feasibility as an anti-MPM therapy. Ascorbate and EGCG are mainly known as antioxidants from a nutritional point of view, but investigations regarding their antitumor properties have also pointed out pro-oxidant properties [@pone.0058051-Chen1], [@pone.0058051-Ranzato1], [@pone.0058051-Li1]. This study provided a confirmation of our previous results, showing that when EGCG and gemcitabine are combined together with ascorbate to form the AND mixture, a synergistic effect is obtained. For preclinical investigations we used a murine model of MPM, developed by injecting tumorigenic REN cells within NOD-SCID immunodeficient mice, a strain that is widely used in tumor biology and xenograft research [@pone.0058051-Harris1]. This model showed various symptoms of disease at necropsy, including main tumor masses close to the site of injection, secondary tumor nodules at various abdominal locations and on the diaphragm, and acute abdominal hemorrhage. In previous studies, different combinations of our active nutrients with conventional antitumor drugs have been reported. It has been shown that ascorbate increases the effects of arsenic trioxide, doxorubicin, cisplatin and paclitaxel on human breast cancer cells [@pone.0058051-Dai1], [@pone.0058051-Grad1], of 5-fluorouracil and cisplatin on esophageal cancer cells [@pone.0058051-Nagy1], and of gemcitabine in a preclinical model of pancreatic cancer [@pone.0058051-Espey1]. EGCG reportedly sensitized breast cancer cells to paclitaxel [@pone.0058051-Luo1], overcame resistance to etoposide-induced apoptosis [@pone.0058051-Ermakova1], and increased apoptosis rates induced by gemcitabine, mitomycin C, or 5-fluorouracil in cholangiocarcinoma cells [@pone.0058051-Lang1]. In our study, the synergistic power of the AND mixture was firstly demonstrated by cytotoxicity and combination index analysis. Data from REN cells indicate that at a 50% effect level, induced by AND at 42.8 µM, the mixture is synergistic with individual concentrations of ascorbate, EGCG, and gemcitabine at 38, 3.3, and 1.48 µM, respectively. All these figures are about sevenfold lower than those inducing the same effect when the compounds are used alone. This results in concentrations compatible with those achieved in human plasma [@pone.0058051-Mereles1]--[@pone.0058051-Wang1], while it is likely that these EGCG levels could be tolerated by humans, as suggested by safety assessment on rodents [@pone.0058051-Isbrucker2]. For *in vivo* experiments, we have chosen an intraperitoneal xenograft mouse model of MPM. This model has already been used in various studies, [@pone.0058051-Bertino1], [@pone.0058051-Frizelle1], [@pone.0058051-Feng1] showing that it is well-suited for mesothelioma research, and may be useful for evaluating novel antitumor treatments *in vivo*. Also, as reported in the Results, we have selected doses for each single component of the injected mixtures previously determined to be safe to animals. This was also confirmed by our preliminary tests. By using this experimental design, strong anti-MPM effects of AND have emerged at different observational levels. At population level, Kaplan-Meier analysis showed significant survival increase for gemcitabine-treated mice, but an identical result was achieved with AND treatment, where gemcitabine dosage was about one third lower. At organism level, the AND mixture inhibited tumor onset, metastasis and tumor-related symptoms like internal hemorrhage. In this respect, gemcitabine alone failed to reduce internal hemorrhage, possibly due to adverse collateral effects, thus further arguing for the superiority of the AND therapy. The lower dose of gemcitabine in the AND therapy seems to reduce the drug\'s toxic effects, and indicates that the combined therapy fits the goal of this study. At the cellular level, there was a shift from cell proliferation to apoptosis in the outermost layer of tumor mass, concomitantly with the inactivation of kinases involved in cell growth. Conversely, angiogenesis factors did not seem to be particularly expressed in MPM tumor xenografts, or specifically targeted by the treatment. Taken together, our data indicate that the AND treatment inhibits tumor growth and invasiveness. The mechanism of action is likely to involve redox processes, as suggested by our previous data about the effects of ascorbate or EGCG on MPM cells [@pone.0058051-Ranzato1], [@pone.0058051-Ranzato2]. However, the AND synergism also indicates that the combined effect is not a mere sum of its single constituents. In conclusion, in this study we have proposed a new possible therapy for MPM, based on a novel, synergistic combination of active nutrients/drug, all used at pharmacological doses. Data provided the following pieces of evidence. 1. The AND treatment showed *in vitro* synergistic anti-MPM activity. 2. *In vivo* experiments on a murine MPM model showed that AND vigorously inhibited the development of disease and exerted a better therapeutic action at reduced dosages of gemcitabine. 3. Data indicated a shift from cell proliferation to apoptosis, blocking tumor growth and invasiveness. Based on these data, we propose the AND therapy as a possible new treatment to be tested on MPM patients in clinical trials. Supporting Information {#s5} ====================== ###### **Dose concentration curves showing cell viability (calcein-AM assay) for each single compounds and the AND mixture, and for each cell type. Vertical dotted line: IC50; vertical continuous line: IC05. Horizontal lines: 95% CI.** (DOCX) ###### Click here for additional data file. We are grateful to Amy Morrison for text revision. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: VV ER SB BB. Performed the experiments: VV ER SM SG MVR. Analyzed the data: VV ER SM LM SB BB. Contributed reagents/materials/analysis tools: SB BB. Wrote the paper: LM SB BB. [^3]: Current address: New York University, School of Medicine, New York, New York, United States of America

All relevant data are within the paper and its Supporting Information files. Introduction {#sec001} ============ Globally, thousands of mushroom poisonings are reported each year \[[@pone.0231781.ref001]--[@pone.0231781.ref009]\]. Approximately 80% of the mushroom poisonings involve unknown mushroom species. The poisonous mushrooms are often classified based on the toxins involved and the clinical signs they elicit \[[@pone.0231781.ref010]\]. Most of the lethal cases are attributed to mushrooms that contain amatoxins. Amatoxins are a family of bicyclic octapeptides that are not inactivated by extreme temperatures, pH, cooking, or digestive enzymes in humans. The principal toxins responsible for toxicity are the amanitins (here, amatoxins; [Fig 1](#pone.0231781.g001){ref-type="fig"}), most prominently α-amanitin (α-AMA), β-AMA and γ-AMA. They are potent inhibitors of RNA polymerase II, essentially halting protein synthesis in eukaryotes. The human LD~50~ for active amatoxins (estimated as the total content of the major toxic amanitins) in a fresh mushroom is considered to be \~ 0.1 mg/kg \[[@pone.0231781.ref011]\]. When α-AMA, β-AMA, and γ-AMA were tested individually in mice (via ip injection), the LD~50~s ranged from 0.2--0.8 mg/kg \[[@pone.0231781.ref012], [@pone.0231781.ref013]\]. Amatoxin-containing mushrooms include a few species from the genera *Amanita*, *Galerina*, and *Lepiota* \[[@pone.0231781.ref011]\]. {#pone.0231781.g001} In addition, there is another class of structurally related cyclopeptide toxins, the phallotoxins. These are produced mainly by *Amanita* species, and debatably by a single *Conocybe* species \[[@pone.0231781.ref011], [@pone.0231781.ref014]\]. Phalloidin, the most well-studied phallotoxin, tightly binds filamentous actin, which prevents depolymerization and ultimately leads to cell death in eukaryotes. Though toxic to eukaryotic cells, phallotoxins are not absorbed through the gastrointestinal tract and thus do not seem to play a role in human mushroom intoxication \[[@pone.0231781.ref013]\]. Both the amatoxins and phallotoxins are encoded by the cycloamanide gene family and are biosynthetically produced on the ribosome \[[@pone.0231781.ref015]\]. Ongoing research continues to explore this pathway to understand more about toxin production and regulation. For expert mycologists, current techniques to identify toxic mushroom species are based on extensive morphological evaluations of the mushroom and knowledge of its habitat. Mushrooms of the same species can vary in appearance at different growth stages and can appear different due to environmental and genetic factors. Many poisonous mushrooms resemble edible wild mushrooms and all genera that contain poisonous mushrooms also include many non-poisonous and edible mushrooms \[[@pone.0231781.ref016]\]. For instance, *A*. *velosa* is a highly desirable edible wild mushroom, but it can produce pure white forms, which to amateur mycologists may appear similar to the pure white *A*. *phalloides* var. *alba* \[[@pone.0231781.ref017]\]. The poisonous white mushroom, *A*. *ocreata*, also emerges in California in the same spring season as the edible *A*. *velosa*. Both associate with oak trees and could be confused by the untrained eye. Mature toxic *Amanita* species can also be misidentified as edible *Volvariella volvacea* (paddy straw mushroom) \[[@pone.0231781.ref018]\] or for edible *Amanita* speciess (i.e., *A*. *hemibapha* and *A*. *princeps*) naturally found in Southeast Asia \[[@pone.0231781.ref019]\]. Due to the lethality of amatoxins, there is a great need for a field-portable, simple and accurate chemical test to determine the presence of amatoxins in mushrooms or diagnostic samples. Early attempts for a rapid, chemical assay of amatoxins in mushrooms used the Meixner-Wieland test \[[@pone.0231781.ref020]\]. The Meixner-Wieland test is a simple procedure wherein juice from a fresh mushroom is rubbed onto lignin-containing paper. In the presence of a concentrated acid, a blueish-green color is observed. However, the test also reacts with hydroxylated indoles and therefore is not specific for amatoxins \[[@pone.0231781.ref021]\]. False positives were reported 19% (63 out of 335) of the time \[[@pone.0231781.ref022]\]. For mushroom analysis, instrumental methods (e.g., liquid chromatography-mass spectrometry (LC-MS)) are highly sensitive and selective, but require extensive sample pre-treatment and expensive equipment \[[@pone.0231781.ref023]--[@pone.0231781.ref026]\]. Immunoassays (e.g., enzyme-linked immunosorbent assays, ELISAs) are sensitive and selective, but still require specialized reagents and equipment, and take a few hours to perform \[[@pone.0231781.ref027]--[@pone.0231781.ref032]\]. However, these same immunoreagents used in an ELISA can be transferable to a lateral flow immunoassay (LFIA) format, which often significantly reduces the assay time and the need for specialized equipment. Previous attempts to generate a LFIA for amatoxin detection utilized a recombinant single chain variable fragment antibody and was used to evaluate spiked mushroom samples \[[@pone.0231781.ref033]\]. Recently, we generated new high-affinity monoclonal antibodies (mAbs) for the detection of amatoxins. We also demonstrated that rapid (\<1 min) extraction of amatoxins from mushrooms was feasible with simple aqueous-based solutions \[[@pone.0231781.ref032]\]. In this study, we incorporated the new mAbs into a competitive LFIA. After optimization, the LFIA was characterized to determine analyte sensitivity and selectivity, and product shelf-life. We then used this assay to detect toxins from mushroom extracts and, for the mushrooms tested, compared those results to previous descriptions in the scientific literature (i.e., contains amatoxins or not), while a few selected specimens were screened by LC-MS. Materials and methods {#sec002} ===================== Reagents and components {#sec003} ----------------------- Monoclonal antibody (AMA9G3; American Type Culture Collection Accession number PTA-125922) and hapten-protein conjugates (PERI-AMA-BSA and LB-AMA-BSA) were produced as described earlier \[[@pone.0231781.ref032], [@pone.0231781.ref034]\]. Colloidal gold (40 nm), goat-anti-mouse IgG, PVC backing cards, nitrocellulose membranes, Ahlstrom 243 wick pad, Ahlstrom 8964 sample pad, and Ahlstrom 8980 glass conjugate release pad (Helsinki, Finland) were provided by DCN Diagnostics Inc (Carlsbad, CA, USA). The nitrocellulose membranes consisted of MDI 150 and MDI 90 (Advanced Microdevices, Pvt. Ltd, India), FF120 and FF80 (GE Healthcare, Pittsburgh, PA, USA), and CN95 and CN140 (Sartorius Stedim Biotech, Concord, CA, USA). Solutions were dispensed using an XYZ3060 Dispensing Platform (BioDot, Irvine, CA, USA) equipped with a Frontline contact dispenser for the antigen and an AirJet dispenser for the antibody-gold conjugates. The inhibitors tested were α-AMA (≥95%, Funite, Ann Arbor, MI, USA), β-AMA (≥98%, Funite), γ-AMA (≥90%, Enzo Life Sciences, Farmingdale, NY, USA), microcystin-LR (≥95%, Enzo), nodularin (≥95%, Enzo), phalloidin (\>90%, Enzo), phallacidin (≥85%, Sigma, St. Louis, MO, USA), pysilocybin (\>99%, Cerilliant, Round Rock, TX, USA), muscimol (\>99%, Abcam, Cambridge, MA, USA), and ibotenic acid (\>98%, Abcam). The remaining reagents were purchased from Fisher (Waltham, MA, USA) or Sigma, unless specified. All wild mushroom samples were collected from the Point Reyes National Seashore (\#PORE-2017-SCI-0054), obtained from local fungal fairs, or provided by generous mycologists. Most (all but 6) of the mushrooms sampled in this study have been deposited in the UC Berkeley Herbarium for future research access. Preparation of the conjugate pad {#sec004} -------------------------------- Anti-amatoxin mAb AMA9G3 was conjugated to 40 nm colloidal gold. A checkerboard titration of both pH (6, 7, 8 and 9) and antibody concentration (0, 1, 2, 4, 6, 8, 10 and 12 μg/mL) was used to determine the optimal amount of antibody required to stabilize colloidal gold particles. First, a solution of colloidal gold (OD~540~ = 1) was prepared in borate buffer (10 mM) at each pH (6, 7, 8, and 9) and added to the wells (0.2 mL/well) of a low protein binding microtiter plate. Next, for each pH level, aliquots of antibody were added to achieve the desired final concentrations and then incubated for 5 min at room temperature. A solution of 10% NaCl (20 μL/well) was added and the change in color was assessed. The wells exhibiting no color change provided a stable conjugate. The conditions that permitted the lowest antibody concentration to stabilize the gold were used to produce a larger batch of antibody-gold conjugates. After conjugation, the particles were blocked with 10% bovine serum albumin (BSA) for 30 mins at room temperature, and then centrifuged at 15,000 x g for 20 mins at 4 °C. The pellet was resuspended in borate buffer (50 mM borate, 1% BSA, pH 9) and adjusted to a final OD~540~ of 10. When needed for half strip testing, 5 μL of particles were added to 45 μL of phosphate buffered saline (PBS; 10 mM phosphate, 138 mM NaCl, 2.7 mM KCl, pH 7.4) containing 1% BSA and 0.25% Tween-20, adjusted to pH 8. When used for spraying onto the conjugate release pad, sucrose (10% final) and trehalose (2% final) were added. For preparation of the conjugate release pad, conjugate pads were first blocked (50 mM Borate, 1% BSA, and 0.25% Tween-20, pH 8) by complete immersion into solution to allow saturation and then dried for 2 hours at 40 °C. Antibody-gold conjugate was sprayed onto the pad at 10 μL/cm and dried for 1 hour at 40 °C. Immobilization of the capture reagents onto nitrocellulose membranes {#sec005} -------------------------------------------------------------------- Half strips, consisting of a nitrocellulose membrane and a wick adhered to a backing card, were constructed to determine the ideal antigen and nitrocellulose combination. Two different antigens (PERI-AMA-BSA and LB-AMA-BSA conjugates) were dispensed as test lines onto six different nitrocellulose membranes. PERI-AMA-BSA was coated at 11 mg/mL and LB-AMA-BSA was coated at 1 mg/mL in PBS. Control lines were coated with goat-anti-mouse polyclonal antibodies at 1 mg/mL in PBS. The different nitrocellulose membranes were: MDI 150, GE FF120, GE FF80, MDI 90, Sartorius CN95, and Sartorius CN140. The membranes were dried for 1 hour at 40 °C and when assembled, the wicking pad (21 mm) overlapped the nitrocellulose membranes (25mm) by \~2mm. To visualize, equal aliquots of antibody-gold nanoparticles were placed into the bottom of test tubes and each membrane type was dropped into the solution and run for approximately 10 minutes. Preparation and assembly of the lateral flow strips {#sec006} --------------------------------------------------- Full strips were assembled using CN95 coated with antigen LB-AMA-BSA at 0.5 mg/mL. The antigen was applied at 10 μL/cm and then dried for 1 hour at 40 °C. Full strips (4 mm in width) consisted of a 60 mm backing card, a 15 mm sample pad, 10 mm conjugate pad, 25 mm nitrocellulose membrane, and a 21 mm wicking absorbent pad. Fully assembled strips were stored at room temperature in sealed pouches with desiccant, until needed. Full strips were tested both inside and outside of a cassette and no aberrant reactions were observed with each format. For all remaining experiments, full test strips were tested in round-bottom glass test tubes or in wells of a 96 well microtiter plate without the use of a cassette. Analytical detection of α-AMA, β-AMA, and γ-AMA by lateral flow immunoassay (LFIA) {#sec007} ---------------------------------------------------------------------------------- The analytical cut-off value was defined as the amount of toxin that just causes complete disappearance of the test line. To determine the cut-off value for α-AMA, β-AMA, and γ-AMA, a set of eight solutions ranging from 0.1 to 10 ng/mL were prepared in PBS. For β-AMA, additional concentrations were tested ranging from 1 to 2000 ng/mL. For each test concentration and the blank containing only buffer, 100 μL of the solution was added to the test strip at the conjugate pad. Each sample was tested in triplicate. The intensity of the lines was resolved by 10 minutes. If no control line appeared, the test was determined to be invalid. The strips were visualized by two independent readers recording a visual score of the test line intensity (0--6; 0 = no color, 1 = barely visible (faint), 2 = weak color, 3 = moderate color, 4 = moderately strong color, 5 = strong color, 6 = very strong color) and by taking a digital photograph of the test strips. Photographs were acquired by a Nikon SLR camera equipped with an LED ring light (B&H Foto and Electronics Corps, New York, NY, USA) for even lighting. The digital image was analyzed with ImageJ software (NIH, Bethesda, MD, USA). Images were contrast enhanced (default setting of 0.3%) and boxes of consistent size were used to integrate the test line's pixel value. Pixel values were inverted by subtracting the measured value from the maximum possible (i.e., 255). The strips were tested in triplicate and the values were expressed as mean ± standard error. The data was plotted using a 4-parameter logistic equation (GraphPad Prism 7; La Jolla, CA, USA). Analytical selectivity of the LFIA {#sec008} ---------------------------------- The LFIA test strips were tested with a panel of near neighbor chemicals, such as phallotoxins, other cyclic peptides, and other chemicals known to exist in mushrooms, to determine the selectivity of the assay. The chemicals tested were phalloidin, phallacidin, microcystin-LR, nodularin, pysilocybin, muscimol, and ibotenic acid. Each purified chemical was dissolved in deionized H~2~O, then diluted into PBS at relatively high concentrations. Aliquots of these samples were assessed in triplicate. If cross-reactivity (i.e., a disappearance of the test line intensity) was observed, samples were diluted and re-tested at lower concentrations. A visual qualitative reading of either YES (+, positive test) or NO (--, negative test) was performed by two individuals and a digital image of the strip was acquired as described previously. Cross-reactivity (%) was calculated as follows: (\[cut-off value of α-AMA\] / \[cut-off value of the test inhibitor\] x 100. Shelf-life testing of the LFIA {#sec009} ------------------------------ The performance of the test strips over time was assessed via accelerated stability studies to simulate enhanced degradation of the product. The assembled strips packaged in foil pouches with desiccant bags were incubated at 45 and 55 °C with ambient humidity. These conditions were selected as they fall within the typical temperature range for testing in vitro diagnostic products \[[@pone.0231781.ref035]\]. Testing was performed at 0, 4, 7, 15, 22, 26, 37, 44, and 87 days for the strips kept at 45 °C and at 0, 1, 4, 8, 14, 17, 21, 25, 37 and 52 days for the strips kept at 55 °C. On each of the indicated days, a 100 μL aliquot of PBS, 1 ng/mL of α-AMA in PBS, and 10 ng/mL of α-AMA in PBS, was tested in triplicate for each concentration. Visual score readings were performed by one of three independent readers randomly varied by day. Digital analyses were performed as described previously. Mean values of triplicate measurements from the same dose concentration were compared to the first day values by using a one-way analysis of variance (GraphPad Prism) and a post hoc test (Holm-Sidak method). P-values of less than 0.05 were considered statistically significant. The conversion of accelerated time to standard day was calculated using the Arrhenius equation using a Q~10~ factor of 2.6 \[[@pone.0231781.ref035]\]. Mushroom analysis {#sec010} ----------------- Whole mushroom specimens were identified by expert mycologists and then dried at 45 °C for 24 hours. The specimens included those that were known to contain amatoxins (*A*. *bisporigera*, *A*. *ocreata*, *A*. *phalloides*, *A*. *marmorata*, *Galerina marginat*a, and *Lepiota subincarnata*) and several that were known to not contain amatoxins, but were either closely related (*A*. *augusta*, *A*. *calyptratoides*, *A*. *constricta*, *A*. *gemmata*, *A*. *magniveracuta*, *A*. *novinupta*, *A*. *pantherina*, *A*. *protecta*, and *A*. *velosa*), locally foraged (*Boletus edulis*, *Cantharellus californicus*, *Galerina sideroides*, *Pholiotina gracilenta*, *Pholiotina utricystidiata*, and *Volvariella volvacea*), or contained other gastrointestinal irritants or hallucinogenic toxins *(A*. *muscaria*, *Agaricus californicus*, and *Ag*. *xanthodermus*). Small portions of the cap of dried specimens were weighed (\~10--200 mg) and then placed into a 15 mL Falcon tube containing 1 mL of PBS. The solutions were briefly (\<1 min) swirled by hand and then a 100 μL aliquot of the extract was immediately applied to the sample pad of the LFIA test strip, in triplicate. Each sample produced a visual qualitative reading of either YES (+) or NO (--) which was performed by two individuals, and a digital image of the strip was acquired. The absence of a test line indicated the presence of amatoxins or amatoxin-like compounds, while a visible test line indicated no amatoxins were present in the extract. A visible control line indicated the gold-labeled antibody flowed along the test strip and performed appropriately. To increase the number of mushroom species tested with this LFIA method, an herbarium collection (dried samples, collected up to 20 years ago) was utilized to sample a large repertoire (n = 86) of wild foraged mushrooms. As before, small portions of the dried mushrooms were briefly mixed with 1 mL of PBS, and a 100 μL aliquot of the extract was immediately applied to the sample pad of the LFIA test strip. The line intensity was interpreted and recorded within 10 minutes and a digital image was also acquired for each test strip. To confirm the presence or absence of α-AMA, LC-MS analysis was conducted on species known to contain amatoxins (*A*. *bisporigera*, *A*. *ocreata*, *A*. *phalloides*, *A*. *marmorata*, *Galerina marginat*a, and *Lepiota subincarnata*) and on four closely related species that were known to not contain amatoxins (*A*. *constricta*, *A*. *gemmata*, *A*. *muscaria*, *and A*. *pantherina)*. Extraction was performed using dried mushroom tissue extracted using methanol-water-0.01 M HCl (5:4:1, v/v/v) at a ratio of 100 mg of dried mushroom to 1 mL of extraction buffer. The tissue was incubated with shaking for 30 minutes at room temperature, and then centrifuged at 10,000 x g for 10 mins. The supernatant was removed and analyzed by LC-MS/MS/MS for α-AMA and, for one specimen, by ultra-high pressure liquid chromatography-high resolution accurate mass spectrometry (UHPLC-HRAMS) for phalloidin and phallacidin. Mushroom extracts were analyzed for α-AMA according to a previously described LC-MS/MS/MS method with slight modifications \[[@pone.0231781.ref036]\]. In brief, the samples were analyzed using a Thermo Velos Pro linear ion trap mass spectrometer interfaced with a Dionex Ultimate 3000 UHPLC system (Thermo, San Jose, CA, USA). The HPLC was fitted with a 2.1 x 50 mm, 1.8 μm Agilent Zorbax SB-C18 column (Agilent, Santa Clara, CA, USA). Mobile phases were water (A) and acetonitrile (B), each containing 0.1% formic acid. Gradient elution was used, initially set at 5% B, held for 1.5 minutes, then increased to 30% B at 7 minutes and then 90% B at 9 minutes. At 9.1 minutes the solvent composition was set back to 5% B and the column re-equilibrated for 6 minutes. The column flow rate was 0.35 mL/min and the injection volume was 2.0 μL. Mass spectrometer ionization conditions and ion transitions were as per the previously published method \[[@pone.0231781.ref036]\]. Results were reported as positive if the retention time on the total ion chromatogram and the MS fragmentation aligned with the standard solution of α-AMA. One extract (*A*. *marmorata*) was analyzed for phalloidin and phallacidin using a Thermo Q-Exactive high resolution accurate mass spectrometer (Thermo) interfaced to a Dionex Ultimate 3000 UHPLC. The HPLC was fitted with a 2.1 x 100 mm, 1.7 μm Agilent Eclipse Plus C-18 column (Agilent). Mobile phases were water (A) and acetonitrile (B), each containing 0.1% formic acid. Gradient elution was used, initially set at 1% B, held for 1.5 minutes, then increased to 98% B at 9.5 minutes. It was held at 90% B until 13.5 minutes and then set back to 1% B and re-equilibrated for 4 minutes. The flow rate was 0.35 mL/min and injection volume was 20 μL. Positive electrospray ionization was used. Parallel reaction monitoring was used to provide three scan functions. The first collected full scan spectra from *m/z* 75--1125 with 70,000 mass resolution at *m/z* 200. The second was used to collect MS/MS fragment ion spectra of *m/z* 789, the \[M+H\]^+^ ion for phalloidin. The third collected MS/MS fragment ion spectra of *m/z* 847, the \[M+H\]^+^ ion for phallacidin. Both MS/MS scan functions used 17,500 mass resolution at *m/z* 200 and stepped collision energy at 35, 45, and 55 eV. Results were reported as positive if the retention time on the total ion chromatogram and the MS fragmentation aligned with the standard solution of phalloidin or phallacidin. The PBS-based extracts obtained from the *A*. *marmorata* and *A*. *bisporigera* samples were diluted 1000-fold and 100,000-fold in PBS and analyzed by LFIA. This was performed in order to evaluate if the diluted sample would dilute out the detection of the phallotoxins and amatoxins, respectively. Results and discussion {#sec011} ====================== The LFIA for amatoxin detection was developed and performed in a competitive inhibition assay format. A schematic of the test strip, along with an example of a negative and positive test, is shown in [Fig 2](#pone.0231781.g002){ref-type="fig"}. The sample to be tested is added to the sample pad, which interacts with and rehydrates the gold-labeled antibody pre-loaded on the conjugate pad. A competitive assay works such that if amatoxins are present at a high enough concentration in the sample, the antibodies will bind to the amatoxins, thus not allowing the antibodies to bind to the antigen immobilized at the test line, which results in no visible line. As a control to ensure the test is valid, the gold-labeled antibodies will bind to the anti-mouse antibody immobilized at the control line, thus producing a visible control line. {#pone.0231781.g002} Optimal concentrations of antibody-gold conjugation and immobilized capture reagents {#sec012} ------------------------------------------------------------------------------------ The optimal conditions required to stabilize the colloidal gold particles with mAb AMA9G3 antibody protein were to perform the conjugation at a pH of 8 or greater and using 2 μg/mL of antibody or greater. Since the assay would be a competitive format wherein the toxin is meant to displace the antibody binding, we used this lowest acceptable antibody loading of 2 μg/mL. Preliminary testing established that immobilizing goat anti-mouse IgG using a solution at 1.0 mg/mL was sufficient for a visible control line. For the test line, two conjugates were tested in a half strip format, PERI-AMA-BSA coated at 11 mg/mL and LB-AMA-BSA coated at 1 mg/mL, both on 6 different nitrocellulose membrane types. The line intensity for the test line coated with the LB-AMA-BSA antigen was considerably higher than the test lines coated with PERI-AMA-BSA ([Fig 3](#pone.0231781.g003){ref-type="fig"}). Therefore, the LB-AMA-BSA antigen was the preferred coating antigen used for the remaining tests. Based on line morphology and membrane background, CN95 was the preferred membrane and was used for the remaining experiments. In addition, because there was evidence of a darker leading edge on the test line, to make the coloration appear more uniformly distributed, the antigen coating concentration was reduced down to 0.5 mg/mL for the full strip production used for the remaining experiments. {#pone.0231781.g003} Analytical detection of α-, β-, and γ-amanitin by lateral flow immunoassay (LFIA) {#sec013} --------------------------------------------------------------------------------- To generate a standard calibration curve of the LFIA for the three most common amanitins, solutions of different concentrations of α-AMA, β-AMA, and γ-AMA in PBS were assessed ([Fig 4](#pone.0231781.g004){ref-type="fig"}). Digitally-acquired pixel values correlated extremely well with the subjective visual scoring on a scale of 0--6 for α-AMA and γ-AMA ([Fig 4a and 4c](#pone.0231781.g004){ref-type="fig"}), and moderately so for β-AMA ([Fig 4b](#pone.0231781.g004){ref-type="fig"}). For the β-AMA plot, the misalignment seems to be driven by the visual score data point (blue triangle) at 10 ng/mL, while the remaining visual score points trend with the pixel values, and thus the misalignment is likely due to the subjective scoring by eye. In order to remove ambiguity in reporting results, we defined the analytical cut-off value as the concentration in which the test line is completely absent due to the competitive inhibition by the toxin in a sample solution competing with the gold-labeled antibody. The cut-off value for α-AMA and γ-AMA was 10 ng/mL (0.1 μg toxin/g mushroom) and the cut-off for β-AMA was 2000 ng/mL. These results corroborate what we observed when using this mAb in an ELISA format wherein mAb AMA9G3 exhibited a lower IC~50~ for α-AMA and γ-AMA than for β-AMA \[[@pone.0231781.ref032]\]. Based on the digitized pixel values (shown as the red lines in [Fig 4](#pone.0231781.g004){ref-type="fig"}), the limit of detection (LOD; defined as three times the standard deviation of a sample without amanitin) is 0.3 ng/mL for α-AMA and γ-AMA and 30 ng/mL for β-AMA. The LFIA's cut-off value for α-AMA is comparable to the LOD for LC-MS methods used for α-AMA detection in mushroom analysis \[[@pone.0231781.ref024]--[@pone.0231781.ref026]\]. {#pone.0231781.g004} For competitive LFIAs, it is often hard to discern if the line is simply fainter (and therefore partially inhibited) due to the presence of toxin or possibly from lighting conditions, age of the strips, time of reading, or other unknown or unanticipated variables. This uncertainty is observed in the only other published LFIA for amatoxins, wherein the authors note that at 2, 10 and 20 ng/mL of α-AMA the line is still present, although decreased visual intensity than from the "no toxin" test line \[[@pone.0231781.ref033]\]. In our experience, if our LFIA were read hours (or even days) after development (instead of the suggested 10 mins), a faint line would appear for the standards containing 5 ng/mL of α-AMA or less, yet no line appears for 10 ng/mL of α-AMA standard. Thus, defining the cut-off value for this LFIA at 10 ng/mL for α-AMA gives greater confidence, and less ambiguity in the interpretation of consistent results. Analytical selectivity of the LFIA {#sec014} ---------------------------------- To ensure that the LFIA is accurate and selective for amatoxin detection, chemical standards of closely related compounds and other cyclic peptides were tested for cross-reactivity ([Table 1](#pone.0231781.t001){ref-type="table"}). No detection was observed for mushroom toxins psilocybin, muscimol, and ibotenic acid, nor for cyclic peptides microcystin-LR or nodularin. In contrast, α-AMA and γ-AMA have similar LODs, which are lower than the LOD for β-AMA. The cross-reactivity for β-AMA by LFIA is 0.5%. Although this is a small value, given the large quantity of β-AMA in known mushroom specimens (approximately 1--2 mg/g (dried) \[[@pone.0231781.ref037]\]), it would be detectable in a typical extraction (1 mL per approximately 100 mg of dried tissue) and detectable at up to a 100-fold dilution of that extract. 10.1371/journal.pone.0231781.t001 ###### LFIA test results for pure chemical toxin standards of chemicals from associated mushrooms or are other peptide toxins. {#pone.0231781.t001g} Toxin Concentration tested (μg/mL) LFIA result (n = 3) Cross Reactivity[^a^](#t001fn001){ref-type="table-fn"} (%) -------------------- ------------------------------ --------------------- ------------------------------------------------------------ **α-amanitin** 0.01 \+ + + 100 **γ-amanitin** 0.01 \+ + + 100 **β-amanitin** 2 \+ + + 0.5 **phalloidin** 200 \+ + + 0.005 20 -- -- -- 2 -- -- -- **phallacidin** 200 \+ + + 0.005[^b^](#t001fn002){ref-type="table-fn"} 20 \+ -- -- 2 -- -- -- **psilocybin** 100 -- -- -- nd[^c^](#t001fn003){ref-type="table-fn"} **microcystin-LR** 20 -- -- -- nd **nodularin** 10 -- -- -- nd **ibotenic acid** 200 -- -- -- nd **muscimol** 200 -- -- -- nd ^a^ Cross-reactivity (%) = (\[cut-off value of α-AMA\] / \[cut-off value of the test inhibitor\] x 100. ^b^ value estimated from the LFIA reading with the highest majority ^c^ nd = not determined because the analyte was not detected. The LFIA cross-reacts with the phallotoxins (phallacidin and phalloidin) at 0.005%, or a concentration of 200 μg/mL. This was not seen in our previously developed ELISA using the same mAb AMA9G3 \[[@pone.0231781.ref032]\], because the highest concentrations tested for these analytes in our earlier study were lower than 2 μg/mL. These phallotoxins are often found in *Amanita* species at approximately 1--2 mg/g of dried mushroom \[[@pone.0231781.ref017], [@pone.0231781.ref037]\], which are at comparable concentrations to the amatoxins. At the current extraction volume described here, a positive result could be due to the phallotoxins. There are numerous other chemicals within the classes of amatoxins and phallotoxins for which chemical standards are not currently commercially available, such as ε-amanitin, amanin, amaninamide, amanullin, amanullinic acid, and proamanullin, as well as phallacin, phallisin, phalloin, prophalloin, and phallisacin. While they could not be tested for cross-reactivity in this assay, their concentrations and distributions in mushrooms are also not well-described in the literature. Nonetheless, since an antibody binds molecules based on molecular shape and not exact chemical composition, it is conceivable that any of these molecules might be present in a sample. Furthermore, without standards at this time, those samples cannot be definitively confirmed by techniques, such as LC-MS. And thus, the LFIA might produce a positive result although instrumental LC-MS methods cannot validate it at this time. Shelf-life testing of the LFIA {#sec015} ------------------------------ The shelf-life of a product can be estimated by performing an accelerated stability study. Each day that a product is held at an elevated temperature equates to a presumed stability for an equivalent duration of standard days at room temperature \[[@pone.0231781.ref035]\]. We stored test strips at 45 °C and at 55 °C, for up to 87 days and 52 days respectively. Sets of test strips were removed periodically and tested using three different concentrations of α-AMA (0, 1, and 10 ng/mL) in PBS. Overall, no statistically significant loss in signal was observed for the first 44 accelerated days for the strips held at 45 °C and for the first 25 accelerated days for the strips held at 55 °C ([Fig 5](#pone.0231781.g005){ref-type="fig"}). The stability at these accelerated days equate to a minimum shelf-life of approximately 360 standard days (1 year) and 540 standard days (1.5 years), respectively. {#pone.0231781.g005} Testing of different α-AMA concentrations provided us a way to identify if sensitivity was impacted along with overall signal intensity. The consistency of the signal intensity over time was observed in the experiments when 0 ng/mL of α-AMA (only PBS) was used. A decrease in signal intensity was observed using strips from accelerated day 87 held at 45 °C and at accelerated days 37 and 52 for the strips held at 55 °C ([Fig 5](#pone.0231781.g005){ref-type="fig"}). The study was not maintained longer than the latter of those accelerated time points. In addition, the signal intensity also dropped statistically significantly (compared to the signal intensity from day 0) for one other time point, accelerated day 52, for the strips held at 55 °C and tested with 1 ng/mL of α-AMA. These observed drops in signal intensity were no more than 2 points on the 6-point visual score. No signal was ever observed for the strips tested with 10 ng/mL of α-AMA, at either temperature, which was expected since this amount of α-AMA should eliminate the presence of the test line completely. Although signal intensity decreased over time in this accelerated stability study, when the LFIA was exposed to elevated temperatures, the entire signal was not completely diminished. Thus, the LFIA still produced reliable qualitative results for all the conditions tested. The decrease in signal intensity after 1--1.5 standard years could serve as an internal product monitor to know when a batch of strips may need to be replaced. Furthermore, a year or more shelf-life is desirable for a product like this in which the appearance of mushrooms and their related poisonings typically occur seasonally each year. Detection of amatoxins in foraged wild mushrooms {#sec016} ------------------------------------------------ We tested 110 foraged mushrooms, comprised of 96 different species ([Table 2](#pone.0231781.t002){ref-type="table"}) for the presence or absence of amatoxins. The mushrooms were dried specimens collected anywhere between 1 day to 20 years prior to performing this testing. Most of the mushrooms were identified to species by expert mycologists using morphology. For some of the mushrooms that are difficult to differentiate beyond the genus level, species identification was confirmed by DNA sequencing of the internal transcribed spacer (ITS) region \[[@pone.0231781.ref038], [@pone.0231781.ref039]\]. The DNA sequence of the ITS region was then BLAST searched in the NCBI database to assign a species based on the highest percent match. 10.1371/journal.pone.0231781.t002 ###### Species names and UC Herbarium codes for wild mushrooms sampled in this study and mentioned in [Fig 6](#pone.0231781.g006){ref-type="fig"}. A 'Y" in the "ITS" (internal transcribed spacer) column indicates that the ITS region was sequenced and assigned a species based on the highest percent match when BLAST searched in the NCBI database. Bold text indicates those species that are known to contain amatoxins. Those marked with an \* were subjected to chemical analysis by LC-MS. {#pone.0231781.t002g} --------- ------------------------------------ ----------------------- --------- --------- ----------------------------------- ----------------------- --------- No. Species identification UC Herbarium code ITS No. Species identification UC Herbarium code ITS 1 *Agaricus californicus* UC 2060385 56 *Gastroboletus turbinatus* UC 1998587 2 *Agaricus xanthodermus* UC 2060384 Y 57 *Gastroboletus vividus* UC 1860875 3 *Agrocybe pediades* UC 1998617 Y 58 *Gastroboletus vividus* UC 1998577 4 *Amanita augusta* UC 2060350 59 *Gyromitra gigas* UC 199122 5 ***Amanita bisporigera\**** UC 2060392 60 *Handkea subcretacea* UC 1998567 6 *Amanita calyptratoides* UC 2060368 Y 61 *Hemimycena delectabilis* UC 1998640 7 *Amanita constricta\** UC 2060356 62 *Homophron spadiceum* UC 1999292 8 *Amanita gemmata\** UC 2060365 63 *Hypholoma fasciculare* UC 1998522 9 *Amanita magniverrucata* UC 2060358 64 *Hypholoma fasciculare* UC 1998517 10 ***Amanita marmorata****\** UC 2060363 65 *Ischnoderma resinosum* UC 1998572 11 *Amanita muscaria\** UC 2060362 Y 66 *Kuehneromyces vernalis* UC 1998746 12 *Amanita muscaria* 67 *Lactarius deliciosus* UC 1860851 Y 13 *Amanita novinupta* UC 2060389 68 *Leccinum manzanitae* UC 1998720 14 ***Amanita ocreata\**** UC 2060355 69 *Lepiota aspera* UC 2060096 15 ***Amanita ocreata*** 70 *Lepiota castaneidisca* UC 1999327 16 *Amanita pachycolea* UC 2060372 71 *Lepiota cf*. *cristata* UC 2060101 Y 17 *Amanita pantherina\** UC 2060395 72 *Lepiota flammeatincta* UC 2060193 18 ***Amanita phalloides****\** UC 2060369 73 *Lepiota luteophylla* UC 2060057 19 ***Amanita phalloides*** 74 *Lepiota rhodophylla* UC 2060056 20 *Amanita protecta* UC 2060370 75 *Lepiota sequoiarum* UC 2050032 21 *Amanita sylvicola* UC 2060375 76 *Lepiota spheniscispora* UC 2060100 22 *Amanita velosa* UC 2060361 77 ***Lepiota subincarnata****\** UC 2060054 23 *Boletus appendiculatus* s.l. UC 1998735 78 ***Lepiota subincarnata*** UC 2060095 24 *Boletus edulis* UC 2060353 79 *Lepiota* sp. sect. *Stenosporae* UC 2060030 25 *Boletus fibrillosus* UC 1998721 80 *Leucoagaricus erythrophaeus* UC 1999375 26 *Boletus fibrillosus* UC 1998574 81 *Leucocoprinus brebissonii* UC 2060403 27 *Boletus rex-veris* UC 1998729 82 *Morchella* sp. UC 1999063 28 *Boletus rubripes* UC 1861056 Y 83 *Melanoleuca angelesiana* UC 1998764 29 *Butyriboletus abieticola* UC 1998732 84 *Melanoleuca melaleuca* UC 1998913 30 *Calbovista subsculpta* UC 1998863 85 *Melanoleuca robertiana* UC 1998614 Y 31 *Calbovista subsculpta* UC 1998751 86 *Mycena nivicola* UC 1998796 32 *Caloboletus frustosus* UC 1860877 Y 87 *Myxomphalia maura* UC 1999137 33 *Caloboletus roseipes* UC 1860855 Y 88 *Nolanea verna* UC 1998642 Y 34 *Caloscypha fulgens* UC 1999117 89 *Peziza repanda* UC 1998869 35 *Caloscypha fulgens* UC 1998915 90 *Pholiotina gracilenta* Y 36 *Cantharellus californicus* UC 2060357 91 *Pholiotina utricystidiata* Y 37 *Chroogomphus albipes* UC 1861050 92 *Phyllotopsis nidulans* UC 1999138 38 *Chroogomphus pseudovinicolor* UC 1861026 93 *Phyllotopsis nidulans* UC 1998641 39 *Chrysomphalina aurantiaca* UC 1860175 94 *Plicaria endocarpoides* UC 1861196 40 *Citocybe glacialis* UC 1998610 95 *Psathyrella piluliformis* UC 1998613 41 *Clitocybe nuda* UC 1998524 96 *Rhodocollybia maculata* UC 2060373 42 *Clitocybe squamulosa* UC 1998763 97 *Rhodophana nitellina* UC 1998616 43 *Clitocybe* sp. UC 1999055 98 *Rubroboletus haematinus* UC 1861053 44 *Connopus acervatus* UC 1999132 99 *Russula favrei* UC 1860891 Y 45 *Cortinarius cephalixus* UC 1998661 Y 100 *Sarcosphaera* cf. *coronaria* UC 1998862 46 *Cortinarius cyanites* UC 1999129 101 *Spongiporus leucospongia* UC 1860874 Y **No.** **Species identification** **UC Herbarium code** **ITS** **No.** **Species identification** **UC Herbarium code** **ITS** 47 *Cortinarius gentilis* UC 1999046 102 *Spongiporus leucospongia* UC 1860895 Y 48 *Cortinarius rubicundulus* UC 1999317 103 *Suillellus amygdalinus* UC 1998733 49 *Cortinarius subalpinus* UC 1998860 104 *Tapinella atrotomentosa* UC 1999002 50 *Cortinarius* subgenus *seriocybe* UC 1998569 Y 105 *Tricholomopsis rutilans* UC 1998579 Y 51 *Cortinarius* sp. UC 1999036 106 *Tricholomopsis rutilans* UC 1860853 Y 52 *Entoloma trachyspermum* UC 1999311 107 *Volvariella volvacea* UC 2060349 53 *Fomitopsis pinicola* UC 1998908 108 *Xerocomus subtomentosus* UC 1998765 54 ***Galerina marginata****\** UC 2060366 Y 109 *Xeromphalina campanella* UC 1998761 55 *Galerina sideroides* Y 110 *Xeromphalina campanella* UC 1998609 --------- ------------------------------------ ----------------------- --------- --------- ----------------------------------- ----------------------- --------- For amatoxin detection, a small (\<200 mg) piece of dried mushroom was extracted into PBS and swirled for a few seconds. An aliquot (100 μL) of the extract was then placed onto the sample pad of the LFIA. The results were clearly visible by 5 mins, but for experimental consistency strips were read at 10 mins. The specimens that tested positive by LFIA, resulting in a complete absence of the test line, were *Amanita bisporigera*, *A*. *marmorata*, *A*. *ocreata* (both specimens), *A*. *phalloides* (both specimens), *Galerina marginata*, and *Lepiota subincarnata* (both specimens) (a subset of strips is shown in [Fig 6](#pone.0231781.g006){ref-type="fig"}). All of these specimens (at least one specimen from each species) were confirmed for the presence of α-AMA by LC-MS analysis ([S1 Table](#pone.0231781.s001){ref-type="supplementary-material"}) except *A*. *marmorata*. The other 90 mushroom species sampled by LFIA were negative for amatoxins (a subset of strips is shown in [Fig 6](#pone.0231781.g006){ref-type="fig"}). Four of the other *Amanita* specimens (*A*. *constricta*, *A*. *gemmata*, *A*. *muscaria*, and *A*. *pantherina*) were confirmed negative for α-AMA by LC-MS ([S1 Table](#pone.0231781.s001){ref-type="supplementary-material"}). {ref-type="table"}.](pone.0231781.g006){#pone.0231781.g006} Although, in this study, one *A*. *marmorata* specimen, that was positive by LFIA and did not contain detectable α-AMA by LC-MS, the presence of phallotoxins were confirmed by LC-MS analysis ([S2 Table](#pone.0231781.s002){ref-type="supplementary-material"}). This result demonstrates that this specimen does make cyclopeptide toxins and thus possesses the cycloamanide gene family \[[@pone.0231781.ref011], [@pone.0231781.ref015], [@pone.0231781.ref040]\]. Variability in toxin production (i.e., some specimens within this species has produced detectable amounts of amatoxins and/or phallotoxins) has been observed in *A*. *bisporigera*, *A*. *marmorata*, and *A*. *suballiacea* \[[@pone.0231781.ref011], [@pone.0231781.ref041]\]. Upon further evaluation, the 1000-fold dilution of extracts from *A*. *marmorata* and *A*. *bisporigera* were also positive by LFIA. The 100,000-fold extracts from both specimens tested negative by LFIA, which is expected as this would dilute the amatoxins to below detectable amounts. In theory, given the low cross-reactivity with phallotoxins, a 10-fold dilution of the extract would be sufficient to dilute the phallotoxins to non-detectable amounts. However, antibody-based detection is unique in that all of the amatoxins and phallotoxins (even those for which analytical standards aren't available) bind cumulatively and present as a single result---the simple presence or absence of a line. While the LFIA does minimally (0.005%) cross-react with phallotoxins, we cannot exclude the possibility that a false positive result for *A*. *marmorata* is due to phallotoxins alone. A complete set of chemical standards are needed to establish a conclusion. Thus, the LFIA is a useful screening tool, which is identifying species producing cyclopeptides. Further research with appropriate chemical standards would help to provide definitive experimental evidence to validate which particular cyclopeptides are present. To our knowledge, this is the first demonstration of a LFIA for the detection of amatoxins in authentic amatoxin-containing mushroom samples. The speed of extraction and detection (\~10 mins), along with the accuracy of identifying amatoxin-containing mushroom species obtained by this LFIA test is remarkably faster than current antibody-based or LC-MS methods, which take a minimum of an hour to obtain a result \[[@pone.0231781.ref024]--[@pone.0231781.ref026], [@pone.0231781.ref031], [@pone.0231781.ref042]\]. A previously reported LFIA for amatoxins, testing amanitin-spiked mushroom samples, utilized a 90 minute extraction procedure using a methanol-water solution and the extracts required dilution in order to reduce matrix effects \[[@pone.0231781.ref033]\]. Since the matrix effects in their assay were likely due to the presence of methanol, sample extraction and dilution could probably be simplified using the extraction procedure described in our work. For mushroom analysis, LC-MS, ELISA, and our LFIA method exhibit comparable analytical LOD in the ng/mL range \[[@pone.0231781.ref024]--[@pone.0231781.ref026], [@pone.0231781.ref031], [@pone.0231781.ref032], [@pone.0231781.ref034], [@pone.0231781.ref042]\]. In addition, most of the amatoxin-containing specimens contain 2--4 mg/g of total amatoxins per dried (cap) tissue \[[@pone.0231781.ref024], [@pone.0231781.ref037], [@pone.0231781.ref043]--[@pone.0231781.ref045]\]. Together this means that our extracts of dried amatoxin-containing mushrooms can undergo a 10,000-fold dilution and still be detectable. For LFIA detection along with the extraction method described in this paper, the extraction volumes that could be used while still detecting amatoxins from approximately 10 mg of dried mushroom cap tissue can range from 1 mL to 1 L. In addition, our LIFA has worked on fresh specimens extracted using the same rapid protocol. Fresh specimens contain around 90% water, and therefore toxins are 10-fold more concentrated in a dried specimen. Nonetheless, this large range of suitable extraction weights and volumes is desirable for field testing where precise measurements can be avoided. Rapid amatoxin detection can be used to augment existing techniques used by mycologists when describing new species of mushrooms. To date, it is reported that over 10,000 mushroom species have been named and fully described, although this is likely only 1% of the total species of fungi in the world \[[@pone.0231781.ref046]\]. This test would be particularly helpful when distinguishing mushrooms with relatively few diagnostic features, such as *Galerina* or *Conocybe* species. The misidentification of mushrooms by conventional mycological evaluation (i.e., spore print, habitat, morphological characteristics) can lead to unintended detrimental outcomes. For instance, this LFIA test would be especially useful when collecting *Amanita* species, of which there are choice edibles (e.g., *A*. *hemibapha* and *A*. *princeps* in Southeast Asia, *A*. *velosa* in the USA) as well as deadly poisonous amatoxin-containing species \[[@pone.0231781.ref019]\]. A tool like LFIA could help alleviate confusion. For those with scientific resources, as a rapid chemical test, this LFIA could be paired with other technologies using DNA analysis \[[@pone.0231781.ref047]\]. Furthermore, toxin production may be evident in future sample identifications due to improved analytical technologies and interest. Of the medical cases referred to the US Poison Control Centers, greater than 90% of the time the species of mushroom is unknown \[[@pone.0231781.ref048]\]. If a mushroom was available, most health care facilities would typically request the assistance of an expert mycologist. However, the mushroom may not be recognizable or retain its prominent characteristics needed to determine if it is a species that contains amatoxins. The LFIA test could be a valuable tool in health care settings to aid clinicians in identifying potential amatoxin poisonings. This tool is not intended to determine edibility as there are numerous other toxins that can be present for which this test does not detect. For instance, *A*. *muscaria* contains hallucinogenic compounds (i.e., muscimol and ibotenic acid), while the *Agaricus* species tested contain unknown gastrointestinal irritants. None of these individual compounds or mushroom species cross-reacted with this assay, and therefore would not protect a person from becoming ill. Conclusions {#sec017} =========== This LFIA is a simple tool that detects amatoxins and does not require the use of harmful chemicals. The extraction of the mushroom tissue is performed in an aqueous buffer solution and is completed in less than a minute. Compared to ELISA formats, this LFIA has all of the immunoreagents pre-embedded in the design such that no additional reagents are needed at the time of testing aside from the sample extract. In addition, unlike both ELISA and LC-MS methods, the LFIA is a single step procedure from the point of sample addition and requires no washing steps. The total incubation time is 10 minutes and the result is simply identified by the presence or absence of the test line, without the need for specialized equipment. Furthermore, samples can be run simultaneously, whereas with LC-MS methods, each sample is run sequentially. This LFIA is a simple, sensitive, selective, portable, rapid, and accurate tool to detect amatoxins, which can aid in mushroom identification. Supporting information {#sec018} ====================== ###### Total ion chromatograms (top) and mass spectrum (bottom) from the LC-MS analysis of mushroom extracts for the presence of α-amanitin. (DOCX) ###### Click here for additional data file. ###### Total ion chromatograms (top) and mass spectrum (bottom) from the LC-MS analysis of the *A*. *marmorata* mushroom extract for the presence of phalloidin and phallacidin. (DOCX) ###### Click here for additional data file. We thank Tom Bruns (University of California, Berkeley), Richard Ransom (Funite, LLC), and Debbie Viess and David Rust (Bay Area Mycological Society) for their generous donation of identified mushrooms, as well as Paula Do and George Song for their technical assistance. We are grateful to Michael Filligenzi and Robert Poppenga (California Animal Health and Food Safety Laboratory System) for their technical expertise. [^1]: **Competing Interests:**The authors have declared that no competing interests exist.